June 9, 2026 • 53 mins
We’re talking with Leslie O’Flahavan, owner of E-WRITE and writing expert, about how customer support teams can write better help docs in the AI era!
AI is changing the way customers find answers, but that does not mean we should forget about the humans reading those answers. Leslie explains why great knowledge base content still comes down to plain language, clear structure, and accurate information, whether the reader is a person or an AI tool pulling from your help center.
We discuss:
- How AI uses help articles
- Why outdated or vague documentation can create bigger problems faster
- Ways support teams can improve their knowledge bases
Leslie also shares practical writing tips you can use right away! These tips include how to write clearer article titles, when to use summaries, why “notes” and “read more” links can cause confusion, and how to revise your most-used articles without overwhelming your team.
So, if you write help docs, edit AI-generated support content, or manage a knowledge base, this episode will help you make your writing more useful for customers and more reliable for AI to pull from!
Links mentioned in this episode:
- Leslie's LinkedIn Profile
- E-WRITE
- Help Scout article/webinar: Why Analytical Reading Is a Must-Have Skill
- LinkedIn Learning
We want to hear from you! Share your support stories and questions with us at happytohelp@buzzsprout.com!
To learn more about Buzzsprout visit Buzzsprout.com.
Thanks for listening!
Listen
Watch
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Priscilla (00:00):
Welcome to Happy to Help, a podcast about customer
support from the people ofBuzzsprout.
I'm your host, PriscillaBrooks.
In today's episode, we'refocusing on our writing skills
in the AI era.
Specifically, we'll talk abouthow AI is changing the way we
develop our help guides andknowledge bases and how to
balance writing for AI withoutforgetting your human readers.
(00:20):
Thanks for joining us.
Let's get into it.
So, Jordan, there is so muchconversation these days about
AI.
And I know we've done a coupleepisodes about AI and how to use
AI to make yourself moreproductive, how to automate
tasks.
But today we're gonna look atAI from a different angle, which
I'm really excited about.
(00:40):
Yeah.
We're gonna kind of look at howto write our content to help AI
instead of how to use AI tohelp us.
Oh, I love this.
So I'm really excited becausejoining us today is Leslie
O'Flahavan.
She is a writing expert and theowner of eWrite, a consultancy
that she's led for like 30 yearssince 1996.
(01:02):
She specializes in helpingcustomer support teams improve
how they communicate withcustomers across email and chat
and social media.
And she is working with them tocreate even better customer
experiences.
She has worked with Fortune 500companies, government agencies,
and organizations around theworld.
She is a LinkedIn learninginstructor whose courses have
(01:23):
reached over a million learners.
She has worked with over amillion people to help them
write better.
Priscilla, you didn't tell mewe had a celebrity on this
podcast.
No, I know.
I was looking, I was like, holycow, what that number is crazy.
She brings this reallypractical, hands-on approach to
helping teams write better.
And I saw something on a biothat I read that said something
(01:46):
about how she likes to take anyand all writers and make them
excellent writers.
And it doesn't matter youreducation, it doesn't matter
your writing status, but you canbe a great writer and she can
teach you how to do that.
And I thought that was soencouraging, especially for so
many people who feel inadequatewhen it comes to writing.
So anyway, we're talking aboutyou like you're not here, but
you are here, Leslie.
Thank you for coming.
(02:06):
Hey, hey, big group hug.
Leslie (02:08):
Thank you.
I mean, a million people, overa million people.
Over a million people.
Yeah.
Jordan (02:14):
I can't even believe that.
Leslie (02:16):
I know.
I gotta say thanks to LinkedInLearning, though, because they
know how to put the right coursein front of the right learner
at the right time.
And oh my gosh, I have heardfrom so many of these people.
And I love knowing that for alot of them, the training I
offered slipped in right whenthey were getting promoted, or
they were just about to getpromoted, or they wanted a new
(02:39):
job.
I love knowing that somelearning offerings that I had
helped them get to the nextplace in their career.
Priscilla (02:46):
That's incredible.
That really is incredible, thetiming of that and just how
impactful writing and writingskills can be on your career.
So before we jump into all ofthe writing fun questions I have
for you today, we always liketo kick our episodes off with
some positivity.
So who has made your dayrecently, Leslie?
Leslie (03:03):
Well, I have to tell you who has made my day is my
grandson's first grade teacher.
Okay.
And I'm I she's quite a privateperson, so I'm just going to
call her Mrs.
C.
But um I'm a volunteer in theirclassroom, and I've been with
her and with my grandsonnumerous times this school year.
He's in first grade.
Here we are, April, firstgrade, you know.
(03:26):
And I am just marveling at theholy magic of teaching a person
who doesn't know how to read howto read.
Yeah.
There's a method, obviously,there's a pedagogy, there's
numerous practical activitiesthat lead to this outcome.
And then there's the magic.
And, you know, I have been withpeople in their stages of
(03:49):
learning for my whole workinglife.
I was a high school Englishteacher for nine years, I was a
college writing teacher forabout three or four years, and
then I've owned my own business,as you said, for 30 years
helping people learn to write.
But I'm telling you, thesepeople, these little people with
the dimples in their elbows,they start first grade and they
(04:10):
cannot read, and they end firstgrade and they can read.
Yeah, they can decode a magichieroglyph, they can read.
And it's happening to mygrandson.
I saw him take his littlepencil out and he read
something, and then he wrotedown Hulk Smash, and he spelled
both words correctly.
(04:31):
Oh, and you know, because hecan read, and it's amazing.
Priscilla (04:35):
That's so precious, and it's because his teacher is
taking the time to shepherd himthrough that and to teach him
how to read.
Leslie (04:43):
Yes, and teaching him how to read in a way that makes
sense to him, and then the other16 kids need a different way.
And I I just I just have tohold her in my esteem every day,
but especially because youasked me.
Priscilla (04:57):
Teachers deserve all the positive encouragement and
kudos.
Their job is so important, butit's so hard.
Leslie (05:05):
Yeah, it is hard.
It is, and the children are um,many of them are absolutely off
the rails.
There's there's one child,there's one child who must be
held in all all during theschool day.
She sits on the teacher's lapall during the school day.
And the teacher's like kid.
Yeah.
Jordan (05:23):
I think my daughter's in your grandson's class.
Right.
Leslie (05:27):
Well, if you need to be held, I'm gonna hold you.
Jordan (05:30):
It's amazing what teachers do.
Leslie (05:32):
Uh just amazing.
Yeah.
Yeah.
Priscilla (05:34):
So, Leslie, when I first heard about you, it was a
couple years ago.
Our success team was going anddoing like a team meetup.
So we were all gettingtogether.
We rented a beach house.
It was great.
We were all doing someprofessional development stuff
together as a team.
And we watched a webinar thatyou did with Matt Patterson over
at Help Scout about analyticalreading.
And it was such a greatwebinar.
(05:57):
And we then immediately turnedaround and did like a little
exercise as a team and did someanalytical reading like
exercises.
And it was so encouraging.
And I think everyone who wasthere would say that that really
made them a better writer.
And we spent maybe three hourson it total.
It was so cool.
And it really has impacted howall of us write now and how we
(06:20):
read emails when we're workingin the queue.
So you've spent years workingwith support teams and helping
them communicate.
How did you first find your wayinto this?
And what makes you passionateabout writing?
Leslie (06:32):
I found my way into the role I have helping support
teams write well to customersthrough just being in business
as a writing training andconsulting company.
So my business partner, who hassince passed away, we formed
eWrite in 1996, which you knowbecause there's a hyphen after
the E.
Jordan (06:51):
Yes.
Leslie (06:51):
No one uses a hyphen after the E.
But we did and we're stickingto it.
Yeah.
So we formed this uh companybecause in 1996, workplace
writing was undergoing arevolution of sorts because we
were using desktop computers towrite and previously had not.
And so we knew people wouldneed help switching their
(07:13):
writing skills to meet the needsof the job, especially writing
web content, because web contentis hyperlinked together and the
order is kind of uhuser-created.
The sequence is user-created,and that was a big, big
difference.
But somewhere around Y2K, weboth realized that some of the
(07:33):
least empowered writers in allthe workforce, namely frontline
customer service or supportagents, were being asked to
switch from phone service toemail service.
And those folks work in what Iwould call as an assembly line
of communication.
You know, the elitecommunicators, you know, like
(07:53):
the head of PR, doesn't work inan assembly line.
Yeah.
But a frontline customerservice agent who is writing
important message afterimportant message does.
And being a phone agent doesnot prepare you for being an
email agent.
And so we knew these folks aregoing to need help.
And as a matter of our moralcompass pointing north and
saying, some of the mostdistressed and least paid
(08:18):
amongst us need help.
We can help.
And we can show their managersthat they can learn.
Yeah.
And that they they mustn't becalled on to do the job without
training.
So that's how it started.
Really, probably one of us,Marilyn, my partner, or me, one
of us, you know, called tocomplain and then realized we
could email to complain.
And then we started to think,who's answering this?
(08:40):
And that's that's how we that'show we made that switch.
And it was a long time ago.
It was 2000, 2001.
Priscilla (08:47):
It's such an important intersection to focus
on this idea that you havepeople who are really good at
communicating in a verbal senseor really, you know, they're
people pleasers, they are goodwith that emotional connection,
but might not be as strong withcommunicating that via email or
writing things down.
I have seen times where I havewritten an email that I thought
(09:09):
conveyed my feeling and it wasnot taken that way.
And I went, man, I don't feelthe way that they've taken it,
but I clearly didn't do a goodjob of translating that
correctly.
Leslie (09:19):
I wouldn't draw that conclusion.
Yeah, the email is a terriblechannel for conveying tone.
Yeah.
The channel itself is frail.
Priscilla (09:28):
Yes.
You can have all thepersonality and the feeling and
the empathy in your person, butyou have to be able to translate
it in a written way to thepeople you're working with in
order for it to succeed.
And so I think it's cool thatyou come into that place where
people might not have the strongwriting skills and help them
develop those so they can betterrepresent themselves in their
company.
And I think that's a reallyimportant place to be.
(09:50):
And so kudos to you for doingit.
I I think it's reallywonderful.
And I know a lot of people whowould benefit from it and have
benefited from it.
So it's really great.
Leslie (09:59):
It is.
And there's a lot of reasonswhy that switch from phone
service to email service doesn'tgo smoothly or from phone
service to written service inany channel doesn't go smoothly.
And partly because people whoself-select to talk to unhappy
customers on the phone as theirjob, those people like to talk.
(10:19):
Uh-huh.
And they're good at talking andlistening.
They're like baseball pitcherswith one really big arm.
You know, they're they'retalking, their talking and
listening skills areoverdeveloped.
But let's think about that.
Who chooses to talk and listenfor their work?
It's people who don't like towrite.
Yeah.
It's a choice, it's theselection.
(10:41):
And so when we ask them tostart writing, we need to be
ready to help them.
It's it's immoral to putsomeone in a job that they're
not prepared to do.
Priscilla (10:50):
Yeah.
So let's talk about AI,everyone's favorite topic these
days.
But I feel like over the lasttwo or three years, when I am in
these conversations about AI,one of the big things we're
talking about is how AI can dothe writing for you.
And how now, not only do youneed to know how to write well,
but you also need to know how toedit well because you're
getting these drafts that youneed to now edit, or you need to
(11:13):
know how to write specificallyfor AI models to get them the
information they need.
So there's this shift of I needto know how to communicate well
with a customer to I need toknow how to write the best help
doc so that my AI assistantknows how to find the right
information.
And then I can go in and edittheir response.
So that's kind of what I wantto focus on today.
How have you, Leslie, seen withthe rise of AI, how have you
(11:37):
seen that change the way we'rewriting in support?
Leslie (11:40):
As you were laying it out, Priscilla, two images uh
popped into my mind.
One was a hall of mirrorsbecause we are writing for AI
and we are writing with AI.
Both.
Yeah.
Right.
So that was the hall ofmirrors.
And then I don't know if youcan remember in Dr.
Doolittle, you know, that thatcreature called the push me pull
you.
It was like a llama with twoheads.
That's another image that Ihave, you know, of both this
(12:05):
bi-directional way that we'reusing AI.
The written products insupport, such as knowledge base
articles, other kinds ofdocumentation, macros, email
templates, those kinds ofthings.
These kinds of writing that Ithink about as ingredients in a
pantry almost.
We are writing them for humans,but we are also writing them to
(12:27):
power AI.
So that is a big change.
Because if we were talking fiveor six years ago, you know, my
mentor would have been theknowledge article, you know,
KB00736 is only a good knowledgearticle if the people who need
to use it can understand it andact on it.
And by the way, those twomeasures, can they understand it
(12:50):
with a reasonable, notextraordinary amount of effort?
Can they understand and canthey act on it?
Those are the main principlesof plain language.
We know a piece of writing iswritten in plain language when
the intended reader canunderstand it and act on it.
But we were only talking abouthumans.
Now we're not.
We're talking about humans andAI tools, right?
(13:11):
But the blessing is mostlyhumans and AI tools want the
same thing.
Thank goodness.
I'm wiping the sweat from mybrow.
Yeah.
Thank goodness they want thesame thing.
Yeah.
So I'm, I think we might be insomething of a golden era here.
(13:31):
Or I may, let's make it allabout me.
I may be in a golden erabecause I have been nagging for
30 years straight that you haveto write well for that human
user.
And many, many supportorganizations are like, yeah,
yeah, they can suck it up.
Yeah.
Or yeah, yeah, if they have tocontact us twice because the
(13:52):
help content was wrong, oh well,they'll contact us twice.
But now, if the help content iswrong or difficult for the AI
to read, it's gonna superquickly and repeatedly retrieve
wrong as results, wrong answers,which we have the world's most
comical euphemism for.
We call them hallucinations.
I think that's yes, hilarious.
(14:14):
That's like calling a fart, Idon't know, like a broccoli wind
or something.
It's just it's just ridiculous.
Hallucination is just it gaveyou the wrong information
because something that the tooldoes that you could have
contributed to more accurately,you didn't contribute to.
And now it's giving wronginformation.
Priscilla (14:35):
So let's dig into that a little bit.
How are those AI models usingthis content?
Leslie (14:41):
I'm gonna tell you in the um the way that won't pass
the AI sophistication accuracytest.
I'm gonna tell you the way Iunderstand it, which is pretty
good, but also I can't protectyou from an eye roll from an
expert.
That may happen.
Jordan (14:55):
That's okay.
You're speaking our language,though.
Yes.
Leslie (14:58):
So the we know that AI search tools like ChatGPT use
they're trained on naturallanguage.
They're trained on not thekeyword style searching that we
were doing with Google severalyears ago, where if we wanted a
campsite, wanted to book acampsite, we would say campsite
old gunpowder state park July2026.
(15:22):
We would put that in.
Now we ask, we type, how can Ibook a campsite at Old Gunpowder
State Park this summer?
That's how we would type ourquestion.
Well, if you think about it,where at Old Gunpowder State
Park's website does thatlanguage appear?
It appears in an FAQ or itappears in a knowledge base
(15:45):
article.
So the reason that the AIsearch tools want to match the
way the human user searched itto that kind of content,
generally called help content orknowledge content, is partly
because the queries are sosimilar.
And then the answer in the KBor in the FAQ, the FAQ is going
(16:06):
to be shorter, how to book acampsite.
The FAQ answer will be shorterthan the page that the state
park has on how to book acampsite.
So already it's going to preferthe writing that's shorter,
especially with knowledgearticles, which we're not likely
to see a knowledge article onthe example I gave, but maybe
(16:26):
how do I change the accountpermissions in my online account
or something like that.
If we go to a knowledgearticle, those are segmented.
They're often presented indiscrete sections that don't
overlap, and that really helps.
So there's a title, there'soften a kind of a summary or an
overview in a knowledge article,then there's a procedure, then
(16:49):
there's where to get extrainformation.
And so the fact that theknowledge is structured that way
helps the search tool retrieve,it helps them match more
easily.
Yeah.
And retrieve, and because it'sgoing to present the search
results as accurate even whenthey're wrong.
Because it's got, you know, asurfeit of swagger.
(17:09):
It has way too much.
Agreed.
You know?
Yeah.
Yeah.
It just thinks it's right.
So that's partly why the AIsearch tool relies on this kind
of modular and more explicitcontent than it does on other
types of content.
For example, if your CEO writesa blog post about how supply
(17:32):
chains are harming your abilityto deliver your product, or if
there's some much shortercontent about, you know, if you
put it in your FAQ, which is aweird place, you know, UX-wise
to put information about supplychain problems, but if the the
AI search tool will find iteasier and summarize it better
(17:54):
there.
Priscilla (17:54):
I think that's really interesting.
I was having a conversationwith someone the other day, and
we were talking about just that.
Like, um, how much should we bewriting for AI?
Is it about just putting asmuch information out there as
possible?
Is it, you know, keeping thingssuccinct?
And it sounds like what you'resaying is sometimes the succinct
and efficient responses aresometimes better for AI to grab
(18:17):
than a long, drawn-out, superdetailed, rich response or
article.
Leslie (18:23):
Can I introduce a metaphor that I've used in my
web writing trainings for many,many years?
And actually, some giftedlibrarian in Canada made a
Wikipedia page about this.
It's the bite, snack, and mealapproach to writing content.
And I made this up in 1996 whenwe were talking about writing
web content for human readers.
(18:44):
Is you can't know whether yourreader needs just a summary or
just a tiny bit, a bite, asnack, a summary, or a meal, all
the content.
You can't know.
Yep.
First of all, you don't knowevery single reader, and second
of all, even an individualreader changes.
Sometimes I just want a bite,just the headline, and sometimes
(19:05):
I want all the detail on thevery same topic.
So we present all three sizes.
And I think the AI tools needall three sizes.
But when you present the meal,you have to put a whole lot more
signposts in it than you couldfor a single human individual
reader whose very high-levelinterest you've assumed, right?
(19:28):
Because the AI tool needs awhole lot more headings, lists,
tables, because we want it to beable to see each unit, the
smallest unit of meaning at atime.
Priscilla (19:41):
Yeah.
So what's at risk for companiesthat are not thinking about
writing for AI, that aren'tchanging the way that they write
help articles?
What's at risk there for thosecompanies that aren't adapting?
Leslie (19:52):
Well, it would be like having had a really, really slow
waitress who was ready to go toretirement who was serving
awful food loaded with E.
coli.
And now you have like threeathletes who are super strong
and super fast, and they'reserving awful spoiled food
that's loaded with E.
coli.
(20:13):
If because the AI tools are somuch faster, so so pervasive,
they cross all of your systems,you know.
And so they're just able todeliver bad or wrong or
hard-to-read information faster.
And that's awful.
Yeah.
I think that's maybe if youlook it up, I think that's the
(20:34):
seventh circle right there.
I think so.
Priscilla (20:38):
Well, and it it is a good reminder that with
knowledge bases and with helparticles, it is even more
important now that that is 100%accurate or as accurate as you
can make it be, because you'renot just talking about people
coming and looking for aspecific question, but you're
now talking about content thatis being scraped into AI and
used to teach and for them tolearn.
(20:59):
And you go, it really needs tobe right because it can get to
so many people.
And if it's wrong, that canreally be a problem.
Leslie (21:06):
Indeed.
And we know anyone who managesa knowledge base is struggling
to keep it accurate and current.
And I actually hope that thespeed of AI's ability to deliver
knowledge stored in a knowledgebase will help people scale
their knowledge bases back, orthey'll get better at archiving
information where they, youknow, they put it in the storage
(21:28):
unit that they rented at theU-Haul.
You know, they they just aregonna get it out of live play.
Because I always want to thinkabout knowledge content as
having physical weight, youknow.
And if you think of it thatway, then those articles that,
you know, the the top threearticles are getting a thousand
hits a week, but the bottomthree articles are getting three
(21:53):
hits a year.
And if they have physicalweight, just jettison them, you
know, and if you can't deletethem because it makes you
scared.
Archive them, get them out ofthe AI's site.
Priscilla (22:03):
Yeah.
Jordan (22:04):
I think this happened to me recently because I was
having an issue with mydaughter's um Apple pencil.
And I just I went to ChatGBTbecause it's great for
troubleshooting things.
And it gave me this like crazyoutdated information.
And I looked at like the sourceand I realized it was from one
of those like forums that Applehas, and it like provided an
answer, but the answer was likeyears ago.
Leslie (22:25):
Yeah.
Jordan (22:26):
And like the operating system's not even the same.
Leslie (22:28):
Yeah.
Take it down.
Yeah.
Put it in, put it in the U-Haulstorage unit.
Yeah.
Priscilla (22:33):
Well, and that's a great point, Jordan, because and
I didn't even think of this,but all of the community forums
that are public that people canwrite any kind of information,
whether it's right or wrong, andthat can be pulled in as fact
into, you know, AI and bedelivered up as a response.
Right.
Something from years ago orsomething that was never true to
(22:53):
begin with because it was acommunity forum.
And how do you then go?
So now you can't just go andpost a response that says this
isn't accurate.
You need to find a way to makesure that only the accurate
stuff is staying online andavailable to people.
So, you know, we talked aboutthis a little bit.
When it comes to writing forAI, when I first reached out to
you about doing this episode,you said something that I
(23:14):
thought really hit the nail onthe head.
You said the phrase withoutinconveniencing your human
readers.
And I thought, yeah, that's it.
It's not that you want to writefor AI and fully forget them.
Obviously, you don't want to dothat.
But you also don't want to makeit written in a way that makes
it hard for them to understandand take action, like you were
saying before.
And so there's this balancethat you have to have when
you're going in to write for AIand for humans.
(23:36):
And so I want to talk aboutthat balance a little bit.
And you were talking about howthey really are not that
different, which is good newsfor us.
But I've seen so many companiesrecently with this AI first
strategy that they put outthere.
They say AI first, everythingis AI first.
Is that the right move?
Is are there any risks if yougo full in on that AI first?
(23:57):
Or is it something that becausethey're so similar, you don't
have to worry too much aboutthat?
Leslie (24:02):
Well, they're similar in heaven, you know, they're
similar in philosophy, but AIfirst, I don't even really know
what that means.
You know, I think that meanswe're getting pressure to quote,
implement AI.
So we've spent a lot of moneyand for the near term, for the
next 18 months, we're investingall our efforts on supporting
(24:26):
this AI tool.
But but what does that reallymean?
If the AI tools draw on theinformation in your knowledge
base that was to date probablycreated by humans, but
ultimately meant to serveinformation to human users, then
I mean, I don't the sourcematerial is the same.
(24:47):
You know, I think the metaphorof your of your knowledge base
is like a pantry.
I think that's a good metaphor.
So we we still have to go inthere and get the ingredients
out to help people.
So I think we need to keeptalking about humans.
I think that's probably thegreatest harm of an AI first
mentality is that we're forhowever short or long, we're
(25:09):
going to stop talking abouthuman needs.
That's the risk to me.
It's not, as I said, I thinkmost of the ways that we need to
write for AI tools also servehuman users.
There's a few that might not,but most, most of the ways do
work for human users, but wehave to still talk about their
needs.
Priscilla (25:28):
Yeah.
I think that's one of thethings that when AI was first,
you know, on the scene andbecoming this really big talking
point a couple of, you know,three or four years ago, one of
the things that I kept seeingwas this AI first.
We're AI first, our company isAI first.
And I kind of always felt ickyabout that because it makes me
feel like what you're saying isAI is the most important thing.
(25:50):
But at the end of the day, thecustomers are the most important
thing.
And what you're doing is usingAI to better serve the
customers.
So in my mind, it feels weirdbecause I still see these like
AI first in the slogan ofthings.
And I want to say, well, it'susing AI to put the customers
first.
That's what you're doing.
Right.
Um, what does that really mean,AI first?
(26:11):
Aren't we talking aboutcustomers here and keeping the
human aspect of at the end ofthe day, we are trying to find
better ways to serve humans.
We're not trying to betterserve AI.
Leslie (26:22):
Right, right, exactly.
Even though our conversation isin some ways about serving
information to AI, which is alegitimate thing to talk about,
without, I mean, I think it's umI've I've watched this kind of
industry-wide, I'm not going tocall it hysteria.
I guess it'd be likeuncontrolled excitement or
(26:43):
something like this.
I've watched it.
This is what I feel is my thirdchapter.
So chapter three is AI, chaptertwo was social media, and
chapter one was uh desktop webpublishing.
Because when I started mycompany, there were, you know,
somebody was using front page tocreate the website, and the one
(27:04):
server was under the IT guy'sdesk, you know.
But this is different.
This is different in scope, andit's taking us, I think, longer
to calm down because it isdifferent in scope.
But I'm ready for the calmdown.
Priscilla (27:19):
So I want to talk a little bit about like the
writing tips that you have forall of us who are listening who
want to be better at writinghelp articles, who want to be
better at editing thoseresponses in the inbox.
What are some of the specificwriting techniques that you can
give us that can improve helpdocuments and make them more AI
friendly?
Leslie (27:38):
Sure.
So if you're writing aknowledge article, I can give
you at least four.
And once I get going, probably24 tips for writing a knowledge
article.
The first thing is you shouldwrite a clear title, probably a
title that has a verb in it,because verbs drive meaning in a
way that nouns and noun stringscan't.
So don't call your knowledgearticle something like admin
(28:04):
account user.
You call it add an adminaccount user, or how do I add an
admin account user?
The verb add, it will, youknow, help the AI tool unite the
search to that content, butit's also a great practice for
human users.
Priscilla (28:23):
Those kind of strategies are like exactly what
I love to hear.
Like the very specific use averb in your title.
Leslie (28:30):
Yes.
The practice of adding a verbto the title, whether your
title's written in question formor whether the verb is in
gerund form, like adding anadmin account user.
This helps the human writer, orif you're having the AI tool
write the knowledge articles, ithelps the tool figure out what
the scope of the article will bebecause your adding a user may
(28:51):
be different from changing youraccount status, right?
So you don't want any of thatto slip in.
You named it adding an adminaccount user, and that's all
that it should be.
The next thing is to write thatmeal.
So the heading is the byte, andthe first couple sentences
should be a summary of theinformation that follows.
(29:12):
This is a common practice inknowledge-based writing.
This is already a practice inplace.
Some organizations do it betterthan others, but you use a
sentence or two that summarizeswhat the article will be about.
And that is, you know, we wesee those summaries march right
off of the source content, rightinto the AI search results.
(29:34):
But you have to write one inthere if you want it to uh show
up in the AI search results, orthe AI will make one for you and
it's not likely to be what youwanted.
Some organizations write anintroduction instead of a
summary, but I'm saying no, thenext best practice is to write a
summary at the beginning andkeep it short, a couple
(29:55):
sentences or three sentences.
Then you're moving into thesteps on how to do this.
Use technical writing, bestpractices from 1954 to write
those steps.
So if there's steps in asequence, number them.
Begin each step with a verb,and this is really hard to do,
(30:16):
but really important to do.
Scrub out notes or that pointwhere you say important colon
because those are throatclearings.
Those are the person writingthe procedure, not quite sure of
where to put that information.
So don't do that.
That is that is a very vaguename for a little piece of info,
(30:39):
note.
Yeah.
What does that tell the humanreader?
Not much.
And and it won't tell the AImuch either.
And the AI is dumber than us,right?
Yeah.
So we don't want to give it anote.
It doesn't know what to make ofa note.
So you have to figure out isthat note a step?
Does that cue you that you needanother section in the
procedure?
And it's actually two sets ofsix steps, not one set of 12.
(31:04):
But you know, you you say toyourself, I'm drafting this or
I'm using an AI tool to draftthis.
If the note lingers there,that's fine.
But before a human or a botuses it, I have to fix it and
get rid of those notes.
Priscilla (31:17):
I'm laughing because our help guide is riddled with
notes.
And I know this because I wrotea lot of those articles and I
put a lot of notes in there anda lot of pro tips.
Right.
You know, I'm like, oh, I don'twant them to forget this little
piece of information.
So I'm gonna put a pro tip downhere or a note, like, because I
want to make sure they knowthat you have to be on this
(31:37):
certain plan in order to usethis certain thing.
So I'm gonna throw it in as anote.
And we use those all throughoutour help articles, and now I'm
just my wheels are spinninggoing, okay, I gotta go and
revise some of these help docsand get those notes clarified.
Well, which is better, note orpro tip?
I don't know.
You tell me.
It's 50-50.
Go for it, guess.
(31:58):
I mean, I feel like they'reequally bad, is what I would
know.
Is that what I would okay?
Leslie (32:03):
A tip is, you know, if we asked 10 people on the street
now, what is a tip?
The tip is a little bit ofadvice about how to do something
or how to get what you want.
Okay.
So so I think a tip is probablybetter, but what is a note?
Yeah, what is it?
It's too generic.
And actually, your articlesmight have some kind of internal
(32:24):
strong architecture or theknowledge base might have this
kind of internal strongarchitecture.
If every time you wrote a note,it was the same type of
information.
And then I would say, awesome,leave them in there, but let's
give them a different name.
Yeah.
You know, so maybe every timeyou wrote a note, it was about
(32:45):
whether such a change wouldcause a charge.
Yeah.
Note if you use this app, itwill you have to be on the
silver plan, not the bronzeplan.
Let's and if you did that everytime, then I'd be like, first
of all, what an orderly brainshe has.
And then second of all, let'sjust call it something
(33:05):
different.
It then we're only talkingabout the word now.
Priscilla (33:08):
Yeah, like pricing consideration or something more
specific.
Right.
It's funny, I feel like I canfeel the people listening to
this who manage the helpdocuments.
I mean, I think of the personon my team who manages them, but
then I also think of people whoare listening to this going,
okay, I got some projects infront of me.
I got some things I gottaupdate because it's just so it's
(33:28):
just so practical.
And it's it'll take time, butit's an easy thing to go in and
update and change, and it'sgonna make a big difference for
the way people access yourinformation.
Leslie (33:39):
Yes, it is going to make a big difference.
Yeah.
Absolutely.
And then the next section youwant, and most knowledge
articles have, is uh access toother resources, or the the very
system sets up links to otherresources.
So all we'd want to be sure isthat it's named properly, right?
So read more is not a goodname, and this is definitely not
(34:03):
a good name for AI.
Right.
They are not smart enough andthey're not inferring, but they
have seen read more often enoughto know what it leads to.
Yeah.
Right.
But for human users, we want tosay something like more
information on adding users orrelated information to adding an
(34:23):
admin user.
So that's the best practiceright there.
And that's helpful for humansand for AI tools also.
Priscilla (34:30):
Yeah, I was gonna say one of the first things.
So when I started, I was thefirst support person here at
Buzzsprout, our head ofmarketing was the one who was
really training me because hehad been doing support and
marketing.
And so he was giving me sometraining.
And one of the things that hetold me on my first week, which
I remember it has likesolidified in my head, was when
you're in an email and you'regonna link to a resource, make
(34:53):
sure that the words that you'relinking are describing the
resource.
So it's not just saying clickhere or learn more, but it's
saying check out this videoabout adding users, and then you
link to the video.
And it's one of those thingsthat's stuck with me for the
last decade is that theimportance of being descriptive
(35:14):
when you're linking out, becausepeople are much more, I mean,
humans are much more likely toclick on a link if they know
what they're going to.
Leslie (35:22):
Right.
And a link is like a littledoor.
Yeah.
It's like a door in thecontent.
And if you're gonna walk out adoor, you should know where
you're going.
Yeah.
You know.
Priscilla (35:30):
So what else?
Are there any other strategies?
We've gone through resources.
So that's the end.
When you've got the resources,you link out to those other
places, the other ways to learn.
One thing we do at the end ofall of our help articles, and
you can tell me if this is agood strategy or not, is that we
always put our email at thebottom.
We say, if you still havequestions, here's our email.
It's in every single articlebecause we never want people to
(35:52):
finish an article and not knowhow to get in touch with us.
Yeah.
Do you think that's overkill?
Leslie (35:56):
That's a good thing.
No, no, that's wonderful.
If I had to lay out acontinuum, how are we going to
serve human users who did notget the help they need?
Over at the far right side ishere's our email, contact us,
what you're doing.
And over at the far left sideis those stupid thumbs up,
thumbs down.
(36:17):
Did this help you?
I hate those.
Yeah.
They're so stupid.
They don't help anybody, theydon't help the team get better.
Yeah.
What does it mean, thumbs down?
Priscilla (36:26):
And and when you're in the middle of trying to find
an answer to a question, you'rereally not concerned with
letting someone know that theirhelp article was not helpful.
You're like, well, no, thiswasn't helpful, but I'm I'm
still looking for my answer.
So I'm gonna keep looking.
I'm not gonna answer yoursurvey right now.
Like that's how I feel when Isee those.
So yeah, I'm right there withyou.
Yeah.
We always put there, if youstill have questions, or if this
(36:48):
didn't answer your initialquestion, here's an email
directly to us.
And so yes, I think that's agood strategy for making sure
things stay human.
Leslie (36:56):
I want to add one more point that's I think really
interesting.
It's very interesting to me asa lifelong writing teacher.
The tone of the articles andthe tone of the content that we
write for humans and for AIshould be a conversational tone.
So what's because the searchqueries are coming in in a
conversational way.
Yeah.
(37:16):
You know, nobody's querying aspursuant to my lease.
You know, what addenda should Ireview?
You know, nope.
So nobody's doing that.
So all the plain languagepractitioners who are in my
cohort of life, we have beenurging people to use a more
conversational tone in workplacewriting for many, many years.
(37:37):
And by conversational, I simplymean business writing that is
more similar to the way we speakthan it is to some of the kind
of vaunted academic and legaltones for writing that are
dissimilar from the way wespeak.
I'm not saying write the wayyou talk to your friends, I'm
saying move closer to the wayyou talk.
(38:00):
So use contractions, throw in afragment every once in a while,
because when we're talking toeach other, we sometimes say no
way.
Yeah.
Yeah.
Or we say, what now?
You know, and those kinds ofphrasings can go into our
business writing without harmingour authority or our
authenticity.
So the tools are trained ondiction, conversational diction,
(38:24):
because so much of theinformation they imbibe is
conversational.
So our writing in the knowledgearticle should also be
conversational.
Priscilla (38:33):
Yeah.
I think every single personI've trained on my team over the
last decade, I have had to talkto them about not talking in a
way that they think they'resupposed to talk, but leaning
way more into the easyconversation kind of talking.
And I feel like I see it themost in people who are a little
insecure about their writingbecause they're trying to write
(38:55):
in a way that they think they'resupposed to write to sound
smart.
And I have to go, hey, I'drather hear how you talk than
hear you use words that peopledon't use in their regular life.
Leslie (39:09):
This is a little different challenge in knowledge
articles than it is in emailresponses because the incoming
email may be from somebody who'squarrelsome or upset.
So sometimes the agent, thesupport person, kind of draws
themselves up and like a pufferfish, you know, I'm gonna make
myself big, I'm gonna makemyself sound stern and
(39:30):
authoritative.
So you'll stop asking me thesame question I told you no
already.
But in knowledge articles,that's not the relationship we
have with the reader.
And so conversational writing,even warm writing is appropriate
there.
Priscilla (39:46):
Are there any strategies that you used to
teach people that now in the AIera you don't teach anymore?
You leave out.
Leslie (39:54):
I thought about this question a lot, and I'm gonna
say something's gonna get mekicked out of the English
teacher world.
Like, I would be like, son, ifyou don't know how to use an
apostrophe now, you have graystreaks in your beard and you
don't know how to use anapostrophe now, it's over.
Don't learn it.
Yeah, because AI can do thatfor you.
AI is so much better thanMicrosoft Editor, you know.
(40:17):
It's a good copy editor, yeah.
Yeah, you it really is a goodcopy editor.
And how did you get to be 36and not know how to use an
apostrophe?
There's lots of reasons.
I don't know your particularreason, but life is short.
Give up.
Give up.
Give up.
Stop worrying about it.
But make sure that you know,you can tell the tool.
(40:39):
You know, you couldn't tellMicrosoft Editor, I'm terrible
at apostrophes, but you can tellthe tool, I'm terrible at
apostrophes.
So true.
So check specially forapostrophes.
Or you could tell the tool, I'ma terrible proofreader.
For example, I have a clientwho was telling me that the
proofreading they're doing,contact enter is slipping
(41:02):
through when they mean contactcenter, because of course enter
is a word.
Yeah.
Oh.
Right?
Yeah.
So you can tell the tool, youknow, don't let me make a
mistake where both words arespelled correctly but don't make
sense.
Yeah.
Jordan (41:17):
I feel like this is in the same vein as like math
teachers saying that you arenever, you're not gonna have a
calculator with you at alltimes, and so you have to learn
math.
And like it feels like we're inthat same vein here where it's
like you have to learnpunctuation because you're not
gonna have like an editor withyou all the time.
It's like, well, yeah, what canI do?
(41:38):
Yeah, yeah, you've gotcalculators on your phone.
Leslie (41:40):
Now I got sweaty, Jordan.
I didn't, I said if you havegray in your beard and you don't
know apostrophes, stop worryingabout it.
I did not see it.
Priscilla (41:49):
You're advocating for some take it out.
Leslie (41:52):
Yeah, no, I am not.
I absolutely am not.
Priscilla (41:55):
Yeah.
Leslie (41:56):
You need to know them, but for whatever reason, sir,
you slip through.
Yeah.
Priscilla (42:00):
So well, and I think the same goes for their, there,
and there, and your and your,you know, these same ideas of
like, if you've gotten to bethis age and you don't know it,
then AI can help you.
Yeah.
Leslie (42:14):
Yes.
And and the reason you don'tknow it, if you've gotten to be
in your 30s and you don't knowit, the reason you don't know it
is very likely because you havesome language-related learning
disability and you're you'reabsolutely coping a hundred
percent of the time.
And the fact that you're uhkind of a fluent, even charming
(42:37):
writer, and your spelling andpunctuation uh is atrocious,
right?
Props to you.
Yeah.
Use the tool.
Yeah, you know, use the tool.
Or you had an inadequateeducation for whatever reason.
Right.
But that I'm very happy about.
That I'm really happy about.
Priscilla (42:54):
Because it can take writers that are good at the
communication, but maybe don'thave all the technical details,
and it can make them so muchbetter.
And that really is a veryencouraging thing for all of us.
Leslie (43:06):
Especially in our industry.
Especially in our industry,because our industry on the pro
side is an industry where peoplewith less than conventional
academic credentials can getpromoted.
That's one thing I love aboutour industry.
It's like this person was agreat frontline agent for three
and a half years.
(43:26):
Now what?
Manager?
Now what?
Director?
Oh my God, that's great.
But a lot of times people inour industry flatline because
they don't have, you know,business professional writing
skills.
And they they the exposurecosts them that next promotion.
And AI tools can smooth thatout.
(43:48):
And that's really beautiful.
It is.
Priscilla (43:50):
Mm-hmm.
So you talk about this a littlebit in your strategies, but I
had a question on here, and Iwant to make sure we fully flesh
it out.
Things like headers and thatstructure that you talked about,
that is all.
The very important in AI erabecause the AI tools are looking
at the headings and they'relooking at the subheads and the
notes and all of that and usingthat as they are delivering
(44:12):
these responses.
Leslie (44:13):
Yes, absolutely.
And they want the H1, H2, H3.
They want the formatting andthe structure to be built a
certain way so that they canfind and prioritize the
headings.
And they want modular contentwritten briefly in a
conversational style that'saccurate, you know.
Priscilla (44:34):
You know, you talked a little bit about keywords and
how we used to be so focused onkeywords.
And now is it not as importantto have keyword-rich articles?
Leslie (44:44):
Again, I'm gonna give my answer, but I want to give it
modestly because, you know,imagine that a marketing person
is listening to this and amarketing person who developed
their chops, a marketing person,an SEO pro who can turn to a
prospective client and say, Yes,I can get you on the first page
of Google, right?
That's not me.
Yeah.
That's not me.
(45:04):
AI tools don't search keywordsin the same way that we used to,
you know, thinking back intothe keyword stuffing era and how
bad it was for human readers.
Terrible for human reasons.
You know, where the page wouldsay, our cashmere sweaters have
more cashmere than any othercashmere sweater that a cashmere
(45:25):
loving shopper could buy, youknow, that kind of thing.
So um I I think we're inanother era where uh natural
language and uh human sequencedstrings of words are uh how
we're searching.
Jordan (45:39):
I am looking forward to that hitting the uh recipe
blogs.
Like in where it's just theinformation's at the top and I
don't have to like scrollthrough like their family
history.
Leslie (45:49):
Yeah, oh my God.
Uh and those are definitely AIwritten, and they were slop when
people created them themselves,and they're just slop right
now.
Priscilla (45:58):
It's so hard.
Leslie (45:59):
Jump to recipe.
Jump to recipe.
Priscilla (46:01):
Yes.
Leslie (46:03):
Yeah.
Priscilla (46:04):
So we're talking about creating these help
articles, but we're also kind oftalking about using AI to
create the articles.
So, what are your thoughts onthe help article managers?
So the people who are managingthose knowledge bases using AI
to create the articles that aregoing to be used by AI.
Is there danger in that?
(46:24):
Like we've talked a little bitabout them, but what kind of
strategies do you recommend forpeople who are using AI to
develop the articles that AI isgoing to then use?
Leslie (46:33):
Hall of Mirrors and So I'm an English teacher, I'm a
writing teacher, and I just wantto say on behalf of my
profession, we need a hug.
Yeah.
We need a hug.
It's a little bit like being abank teller the day they
installed the ATM.
We're worried.
(46:54):
Yeah, we're worried.
So um, but also I'm over thatbecause there's so much
opportunity for people to dogreat work as writers using AI
tools.
So, yes, of course, write thearticle using AI and then serve
it to AI.
But here's the thing um that II worry a lot about.
(47:16):
Um, when you are a person usingAI tools to write for humans or
for other AI tools, you becomesomething like the food tester
to the queen.
So the AI tool made the draftknowledge article, and there you
are, and you are responsible.
You are the quality control.
(47:38):
You're the one who stops it andsays, Don't let the queen take
a bite of this, it'll kill her,or you're the one who says, Go
ahead.
And if you were born an analogwriter like me, like both of
you, and like nearly, but noteveryone we know who's listening
to this, if you were born ananalog writer, you have
(47:58):
developed the skills to evaluatea piece of writing by producing
it.
Yeah.
But we're looking at aworkforce, the youngest amongst
us in the workforce, who haven'tand won't be called on to
produce a piece of writing, butthey're still the food taster to
the queen.
They still have to be able todecide whether it's ready for
the human reader and for the AIreader also, the AI tool.
(48:23):
I have a great way that allwriters, but we'll talk about
writers in support and service,can be helped to build skills
that will enable them tocritique and improve an
AI-generated draft.
Okay.
Okay.
And that is that they have tobe helped to know more of their
(48:43):
customers and have theircustomers' reactions to the
written products of theircompany something that they have
access to.
So what I dream of is some kindof, I don't know, customer
expression group and not surveyresults, absolutely not survey
results, videos or live contactwith customers where the
(49:07):
customers talk about what is itlike to read this knowledge base
article?
What is it like for me?
Or where they talk about howthey liked or didn't like that
email response.
Because I do believe if an AItrained human writer can imagine
the customer's reaction, itwill strengthen their abilities
(49:29):
to edit what the AI tool hasgenerated.
But there's no replacementbecause if we're not learning
it, if we're not learning ourediting skills by editing, then
what we need is to make thereader come alive to us in a way
that it's kind of expensive forcompanies, but there's no
substitute for it.
(49:49):
Yeah.
Priscilla (49:50):
Yeah, I think that's really, I think that's really
interesting.
What is one takeaway that youwant people listening to this
who are maybe those people whoare managing their knowledge
bases?
What's one takeaway you wantthem to take from this as they
go and start editing andrevising their articles?
What do you want to them tofocus on?
Leslie (50:09):
I want them to focus on the top 10 or 7% of their
articles.
So I don't even want to whelmthem, much less overwhelm them.
I want to empower them and tosupport them.
So if they're sitting there andthinking, God, I wish I hadn't
listened to that.
Now I have an insurmountableto-do list and I feel like I
(50:30):
want to quit my job, you know.
No, no, that's not what I'msaying.
Those knowledge articles, anykinds of stored knowledge, are
used frequently or not at all,right?
Yeah.
Some are used way frequentlyand some are used not at all.
So if you focus on the onesthat are used really, really
frequently and you improve thewriting of the top three most
(50:51):
frequently used, you can teachthe AI tool to write like that.
And the AI tool can probablyedit a lot of the other ones.
So I would say don't getoverwhelmed.
Nothing I've said is somethingsomeone hasn't heard before.
You know, I'm just saying itwith more enthusiasm.
(51:12):
I don't know.
And just focus on the contentthat affects the quality of the
experience the most.
And we know what that is.
We have all the reporting weneed.
Yes.
Priscilla (51:22):
You can find that information and you can tackle
those first three first.
Indeed.
Indeed.
And I love that.
There's no need to get whelmedor overwhelmed.
Right.
Yeah.
Overwhelmed.
Right.
I love it.
I love it.
Well, thank you so much fortaking the time to be here,
Leslie.
I mean, I feel like we talkedabout a lot of things in this,
but I think we have some really,really incredibly practical
(51:42):
takeaways from this conversationthat I'm really excited to
implement on my team.
And I got to go get those notesout of there.
That's right.
But I just I think there's somereally great information here.
So thank you so much for comingand sharing that with us.
For anyone listening who wantsto level up their writing and
they're like, okay, how do Ibecome the next person in your
(52:03):
in one of your courses?
How do they get in touch withyou?
How do they get access to thatteaching?
Leslie (52:08):
Well, they can always come join me on LinkedIn.
They can definitely take theLinkedIn Learning Trainings.
And I don't know if folks knowthis, but if you have a public
library online account in theUS, if you can log into your
public library online, you canget LinkedIn Learning Training
for free.
Jordan (52:23):
Oh, I didn't know that.
I did not know that either.
I love libraries.
Oh my gosh.
Leslie (52:27):
Yeah, it's called LinkedIn Learning Patron or
something like that.
LinkedIn Learning, and you canget the classes for free.
Cool.
But I would be more than happyto look at a sample of anyone's
knowledge article.
And if they want to say, justgive me some feedback, you know,
including you, Priscilla, ifyou want me to look at your
notes, I would be thrilled to dothat.
And then one other thing isevery month I do a LinkedIn Live
(52:50):
broadcast with my colleague uhKim Saido Campbell, who's a
professor at the University ofNorth Texas.
And the broadcast is called FixThis Writing.
Two experts show you how tomake bad writing better.
And I'd love it if you join.
Priscilla (53:03):
I love that.
That's so great.
Well, we'll link to that in theshow notes.
So anyone who's listening,they'll have access and links to
those things and quick accessto get in touch with you.
Well, thank you again, Leslie.
This was really wonderful.
So if you have any questions ora topic that you would like us
to cover in a future episode,you can shoot us a text or a
voicemail using the Send Us FanMail link in the description.
(53:26):
As always, if you liked thisepisode, please share it with
someone who works in customersupport or someone who's just
trying to level up their writingskills or leave us a review on
Apple Podcasts.
We would love that.
Thank you again to LeslieO'Flehaven for joining us today
on today's episode.
Thank you all for listening.
Now go and make someone's day.
Welcome to Happy to Help, a podcast about customer
support from the people ofBuzzsprout.
I'm your host, PriscillaBrooks.
In today's episode, we'refocusing on our writing skills
in the AI era.
Specifically, we'll talk abouthow AI is changing the way we
develop our help guides andknowledge bases and how to
balance writing for AI withoutforgetting your human readers.
(00:20):
Thanks for joining us.
Let's get into it.
So, Jordan, there is so muchconversation these days about
AI.
And I know we've done a coupleepisodes about AI and how to use
AI to make yourself moreproductive, how to automate
tasks.
But today we're gonna look atAI from a different angle, which
I'm really excited about.
(00:40):
Yeah.
We're gonna kind of look at howto write our content to help AI
instead of how to use AI tohelp us.
Oh, I love this.
So I'm really excited becausejoining us today is Leslie
O'Flahavan.
She is a writing expert and theowner of eWrite, a consultancy
that she's led for like 30 yearssince 1996.
(01:02):
She specializes in helpingcustomer support teams improve
how they communicate withcustomers across email and chat
and social media.
And she is working with them tocreate even better customer
experiences.
She has worked with Fortune 500companies, government agencies,
and organizations around theworld.
She is a LinkedIn learninginstructor whose courses have
(01:23):
reached over a million learners.
She has worked with over amillion people to help them
write better.
Priscilla, you didn't tell mewe had a celebrity on this
podcast.
No, I know.
I was looking, I was like, holycow, what that number is crazy.
She brings this reallypractical, hands-on approach to
helping teams write better.
And I saw something on a biothat I read that said something
(01:46):
about how she likes to take anyand all writers and make them
excellent writers.
And it doesn't matter youreducation, it doesn't matter
your writing status, but you canbe a great writer and she can
teach you how to do that.
And I thought that was soencouraging, especially for so
many people who feel inadequatewhen it comes to writing.
So anyway, we're talking aboutyou like you're not here, but
you are here, Leslie.
Thank you for coming.
(02:06):
Hey, hey, big group hug.
Leslie (02:08):
Thank you.
I mean, a million people, overa million people.
Over a million people.
Yeah.
Jordan (02:14):
I can't even believe that.
Leslie (02:16):
I know.
I gotta say thanks to LinkedInLearning, though, because they
know how to put the right coursein front of the right learner
at the right time.
And oh my gosh, I have heardfrom so many of these people.
And I love knowing that for alot of them, the training I
offered slipped in right whenthey were getting promoted, or
they were just about to getpromoted, or they wanted a new
(02:39):
job.
I love knowing that somelearning offerings that I had
helped them get to the nextplace in their career.
Priscilla (02:46):
That's incredible.
That really is incredible, thetiming of that and just how
impactful writing and writingskills can be on your career.
So before we jump into all ofthe writing fun questions I have
for you today, we always liketo kick our episodes off with
some positivity.
So who has made your dayrecently, Leslie?
Leslie (03:03):
Well, I have to tell you who has made my day is my
grandson's first grade teacher.
Okay.
And I'm I she's quite a privateperson, so I'm just going to
call her Mrs.
C.
But um I'm a volunteer in theirclassroom, and I've been with
her and with my grandsonnumerous times this school year.
He's in first grade.
Here we are, April, firstgrade, you know.
(03:26):
And I am just marveling at theholy magic of teaching a person
who doesn't know how to read howto read.
Yeah.
There's a method, obviously,there's a pedagogy, there's
numerous practical activitiesthat lead to this outcome.
And then there's the magic.
And, you know, I have been withpeople in their stages of
(03:49):
learning for my whole workinglife.
I was a high school Englishteacher for nine years, I was a
college writing teacher forabout three or four years, and
then I've owned my own business,as you said, for 30 years
helping people learn to write.
But I'm telling you, thesepeople, these little people with
the dimples in their elbows,they start first grade and they
(04:10):
cannot read, and they end firstgrade and they can read.
Yeah, they can decode a magichieroglyph, they can read.
And it's happening to mygrandson.
I saw him take his littlepencil out and he read
something, and then he wrotedown Hulk Smash, and he spelled
both words correctly.
(04:31):
Oh, and you know, because hecan read, and it's amazing.
Priscilla (04:35):
That's so precious, and it's because his teacher is
taking the time to shepherd himthrough that and to teach him
how to read.
Leslie (04:43):
Yes, and teaching him how to read in a way that makes
sense to him, and then the other16 kids need a different way.
And I I just I just have tohold her in my esteem every day,
but especially because youasked me.
Priscilla (04:57):
Teachers deserve all the positive encouragement and
kudos.
Their job is so important, butit's so hard.
Leslie (05:05):
Yeah, it is hard.
It is, and the children are um,many of them are absolutely off
the rails.
There's there's one child,there's one child who must be
held in all all during theschool day.
She sits on the teacher's lapall during the school day.
And the teacher's like kid.
Yeah.
Jordan (05:23):
I think my daughter's in your grandson's class.
Right.
Leslie (05:27):
Well, if you need to be held, I'm gonna hold you.
Jordan (05:30):
It's amazing what teachers do.
Leslie (05:32):
Uh just amazing.
Yeah.
Yeah.
Priscilla (05:34):
So, Leslie, when I first heard about you, it was a
couple years ago.
Our success team was going anddoing like a team meetup.
So we were all gettingtogether.
We rented a beach house.
It was great.
We were all doing someprofessional development stuff
together as a team.
And we watched a webinar thatyou did with Matt Patterson over
at Help Scout about analyticalreading.
And it was such a greatwebinar.
(05:57):
And we then immediately turnedaround and did like a little
exercise as a team and did someanalytical reading like
exercises.
And it was so encouraging.
And I think everyone who wasthere would say that that really
made them a better writer.
And we spent maybe three hourson it total.
It was so cool.
And it really has impacted howall of us write now and how we
(06:20):
read emails when we're workingin the queue.
So you've spent years workingwith support teams and helping
them communicate.
How did you first find your wayinto this?
And what makes you passionateabout writing?
Leslie (06:32):
I found my way into the role I have helping support
teams write well to customersthrough just being in business
as a writing training andconsulting company.
So my business partner, who hassince passed away, we formed
eWrite in 1996, which you knowbecause there's a hyphen after
the E.
Jordan (06:51):
Yes.
Leslie (06:51):
No one uses a hyphen after the E.
But we did and we're stickingto it.
Yeah.
So we formed this uh companybecause in 1996, workplace
writing was undergoing arevolution of sorts because we
were using desktop computers towrite and previously had not.
And so we knew people wouldneed help switching their
(07:13):
writing skills to meet the needsof the job, especially writing
web content, because web contentis hyperlinked together and the
order is kind of uhuser-created.
The sequence is user-created,and that was a big, big
difference.
But somewhere around Y2K, weboth realized that some of the
(07:33):
least empowered writers in allthe workforce, namely frontline
customer service or supportagents, were being asked to
switch from phone service toemail service.
And those folks work in what Iwould call as an assembly line
of communication.
You know, the elitecommunicators, you know, like
(07:53):
the head of PR, doesn't work inan assembly line.
Yeah.
But a frontline customerservice agent who is writing
important message afterimportant message does.
And being a phone agent doesnot prepare you for being an
email agent.
And so we knew these folks aregoing to need help.
And as a matter of our moralcompass pointing north and
saying, some of the mostdistressed and least paid
(08:18):
amongst us need help.
We can help.
And we can show their managersthat they can learn.
Yeah.
And that they they mustn't becalled on to do the job without
training.
So that's how it started.
Really, probably one of us,Marilyn, my partner, or me, one
of us, you know, called tocomplain and then realized we
could email to complain.
And then we started to think,who's answering this?
(08:40):
And that's that's how we that'show we made that switch.
And it was a long time ago.
It was 2000, 2001.
Priscilla (08:47):
It's such an important intersection to focus
on this idea that you havepeople who are really good at
communicating in a verbal senseor really, you know, they're
people pleasers, they are goodwith that emotional connection,
but might not be as strong withcommunicating that via email or
writing things down.
I have seen times where I havewritten an email that I thought
(09:09):
conveyed my feeling and it wasnot taken that way.
And I went, man, I don't feelthe way that they've taken it,
but I clearly didn't do a goodjob of translating that
correctly.
Leslie (09:19):
I wouldn't draw that conclusion.
Yeah, the email is a terriblechannel for conveying tone.
Yeah.
The channel itself is frail.
Priscilla (09:28):
Yes.
You can have all thepersonality and the feeling and
the empathy in your person, butyou have to be able to translate
it in a written way to thepeople you're working with in
order for it to succeed.
And so I think it's cool thatyou come into that place where
people might not have the strongwriting skills and help them
develop those so they can betterrepresent themselves in their
company.
And I think that's a reallyimportant place to be.
(09:50):
And so kudos to you for doingit.
I I think it's reallywonderful.
And I know a lot of people whowould benefit from it and have
benefited from it.
So it's really great.
Leslie (09:59):
It is.
And there's a lot of reasonswhy that switch from phone
service to email service doesn'tgo smoothly or from phone
service to written service inany channel doesn't go smoothly.
And partly because people whoself-select to talk to unhappy
customers on the phone as theirjob, those people like to talk.
(10:19):
Uh-huh.
And they're good at talking andlistening.
They're like baseball pitcherswith one really big arm.
You know, they're they'retalking, their talking and
listening skills areoverdeveloped.
But let's think about that.
Who chooses to talk and listenfor their work?
It's people who don't like towrite.
Yeah.
It's a choice, it's theselection.
(10:41):
And so when we ask them tostart writing, we need to be
ready to help them.
It's it's immoral to putsomeone in a job that they're
not prepared to do.
Priscilla (10:50):
Yeah.
So let's talk about AI,everyone's favorite topic these
days.
But I feel like over the lasttwo or three years, when I am in
these conversations about AI,one of the big things we're
talking about is how AI can dothe writing for you.
And how now, not only do youneed to know how to write well,
but you also need to know how toedit well because you're
getting these drafts that youneed to now edit, or you need to
(11:13):
know how to write specificallyfor AI models to get them the
information they need.
So there's this shift of I needto know how to communicate well
with a customer to I need toknow how to write the best help
doc so that my AI assistantknows how to find the right
information.
And then I can go in and edittheir response.
So that's kind of what I wantto focus on today.
How have you, Leslie, seen withthe rise of AI, how have you
(11:37):
seen that change the way we'rewriting in support?
Leslie (11:40):
As you were laying it out, Priscilla, two images uh
popped into my mind.
One was a hall of mirrorsbecause we are writing for AI
and we are writing with AI.
Both.
Yeah.
Right.
So that was the hall ofmirrors.
And then I don't know if youcan remember in Dr.
Doolittle, you know, that thatcreature called the push me pull
you.
It was like a llama with twoheads.
That's another image that Ihave, you know, of both this
(12:05):
bi-directional way that we'reusing AI.
The written products insupport, such as knowledge base
articles, other kinds ofdocumentation, macros, email
templates, those kinds ofthings.
These kinds of writing that Ithink about as ingredients in a
pantry almost.
We are writing them for humans,but we are also writing them to
(12:27):
power AI.
So that is a big change.
Because if we were talking fiveor six years ago, you know, my
mentor would have been theknowledge article, you know,
KB00736 is only a good knowledgearticle if the people who need
to use it can understand it andact on it.
And by the way, those twomeasures, can they understand it
(12:50):
with a reasonable, notextraordinary amount of effort?
Can they understand and canthey act on it?
Those are the main principlesof plain language.
We know a piece of writing iswritten in plain language when
the intended reader canunderstand it and act on it.
But we were only talking abouthumans.
Now we're not.
We're talking about humans andAI tools, right?
(13:11):
But the blessing is mostlyhumans and AI tools want the
same thing.
Thank goodness.
I'm wiping the sweat from mybrow.
Yeah.
Thank goodness they want thesame thing.
Yeah.
So I'm, I think we might be insomething of a golden era here.
(13:31):
Or I may, let's make it allabout me.
I may be in a golden erabecause I have been nagging for
30 years straight that you haveto write well for that human
user.
And many, many supportorganizations are like, yeah,
yeah, they can suck it up.
Yeah.
Or yeah, yeah, if they have tocontact us twice because the
(13:52):
help content was wrong, oh well,they'll contact us twice.
But now, if the help content iswrong or difficult for the AI
to read, it's gonna superquickly and repeatedly retrieve
wrong as results, wrong answers,which we have the world's most
comical euphemism for.
We call them hallucinations.
I think that's yes, hilarious.
(14:14):
That's like calling a fart, Idon't know, like a broccoli wind
or something.
It's just it's just ridiculous.
Hallucination is just it gaveyou the wrong information
because something that the tooldoes that you could have
contributed to more accurately,you didn't contribute to.
And now it's giving wronginformation.
Priscilla (14:35):
So let's dig into that a little bit.
How are those AI models usingthis content?
Leslie (14:41):
I'm gonna tell you in the um the way that won't pass
the AI sophistication accuracytest.
I'm gonna tell you the way Iunderstand it, which is pretty
good, but also I can't protectyou from an eye roll from an
expert.
That may happen.
Jordan (14:55):
That's okay.
You're speaking our language,though.
Yes.
Leslie (14:58):
So the we know that AI search tools like ChatGPT use
they're trained on naturallanguage.
They're trained on not thekeyword style searching that we
were doing with Google severalyears ago, where if we wanted a
campsite, wanted to book acampsite, we would say campsite
old gunpowder state park July2026.
(15:22):
We would put that in.
Now we ask, we type, how can Ibook a campsite at Old Gunpowder
State Park this summer?
That's how we would type ourquestion.
Well, if you think about it,where at Old Gunpowder State
Park's website does thatlanguage appear?
It appears in an FAQ or itappears in a knowledge base
(15:45):
article.
So the reason that the AIsearch tools want to match the
way the human user searched itto that kind of content,
generally called help content orknowledge content, is partly
because the queries are sosimilar.
And then the answer in the KBor in the FAQ, the FAQ is going
(16:06):
to be shorter, how to book acampsite.
The FAQ answer will be shorterthan the page that the state
park has on how to book acampsite.
So already it's going to preferthe writing that's shorter,
especially with knowledgearticles, which we're not likely
to see a knowledge article onthe example I gave, but maybe
(16:26):
how do I change the accountpermissions in my online account
or something like that.
If we go to a knowledgearticle, those are segmented.
They're often presented indiscrete sections that don't
overlap, and that really helps.
So there's a title, there'soften a kind of a summary or an
overview in a knowledge article,then there's a procedure, then
(16:49):
there's where to get extrainformation.
And so the fact that theknowledge is structured that way
helps the search tool retrieve,it helps them match more
easily.
Yeah.
And retrieve, and because it'sgoing to present the search
results as accurate even whenthey're wrong.
Because it's got, you know, asurfeit of swagger.
(17:09):
It has way too much.
Agreed.
You know?
Yeah.
Yeah.
It just thinks it's right.
So that's partly why the AIsearch tool relies on this kind
of modular and more explicitcontent than it does on other
types of content.
For example, if your CEO writesa blog post about how supply
(17:32):
chains are harming your abilityto deliver your product, or if
there's some much shortercontent about, you know, if you
put it in your FAQ, which is aweird place, you know, UX-wise
to put information about supplychain problems, but if the the
AI search tool will find iteasier and summarize it better
(17:54):
there.
Priscilla (17:54):
I think that's really interesting.
I was having a conversationwith someone the other day, and
we were talking about just that.
Like, um, how much should we bewriting for AI?
Is it about just putting asmuch information out there as
possible?
Is it, you know, keeping thingssuccinct?
And it sounds like what you'resaying is sometimes the succinct
and efficient responses aresometimes better for AI to grab
(18:17):
than a long, drawn-out, superdetailed, rich response or
article.
Leslie (18:23):
Can I introduce a metaphor that I've used in my
web writing trainings for many,many years?
And actually, some giftedlibrarian in Canada made a
Wikipedia page about this.
It's the bite, snack, and mealapproach to writing content.
And I made this up in 1996 whenwe were talking about writing
web content for human readers.
(18:44):
Is you can't know whether yourreader needs just a summary or
just a tiny bit, a bite, asnack, a summary, or a meal, all
the content.
You can't know.
Yep.
First of all, you don't knowevery single reader, and second
of all, even an individualreader changes.
Sometimes I just want a bite,just the headline, and sometimes
(19:05):
I want all the detail on thevery same topic.
So we present all three sizes.
And I think the AI tools needall three sizes.
But when you present the meal,you have to put a whole lot more
signposts in it than you couldfor a single human individual
reader whose very high-levelinterest you've assumed, right?
(19:28):
Because the AI tool needs awhole lot more headings, lists,
tables, because we want it to beable to see each unit, the
smallest unit of meaning at atime.
Priscilla (19:41):
Yeah.
So what's at risk for companiesthat are not thinking about
writing for AI, that aren'tchanging the way that they write
help articles?
What's at risk there for thosecompanies that aren't adapting?
Leslie (19:52):
Well, it would be like having had a really, really slow
waitress who was ready to go toretirement who was serving
awful food loaded with E.
coli.
And now you have like threeathletes who are super strong
and super fast, and they'reserving awful spoiled food
that's loaded with E.
coli.
(20:13):
If because the AI tools are somuch faster, so so pervasive,
they cross all of your systems,you know.
And so they're just able todeliver bad or wrong or
hard-to-read information faster.
And that's awful.
Yeah.
I think that's maybe if youlook it up, I think that's the
(20:34):
seventh circle right there.
I think so.
Priscilla (20:38):
Well, and it it is a good reminder that with
knowledge bases and with helparticles, it is even more
important now that that is 100%accurate or as accurate as you
can make it be, because you'renot just talking about people
coming and looking for aspecific question, but you're
now talking about content thatis being scraped into AI and
used to teach and for them tolearn.
(20:59):
And you go, it really needs tobe right because it can get to
so many people.
And if it's wrong, that canreally be a problem.
Leslie (21:06):
Indeed.
And we know anyone who managesa knowledge base is struggling
to keep it accurate and current.
And I actually hope that thespeed of AI's ability to deliver
knowledge stored in a knowledgebase will help people scale
their knowledge bases back, orthey'll get better at archiving
information where they, youknow, they put it in the storage
(21:28):
unit that they rented at theU-Haul.
You know, they they just aregonna get it out of live play.
Because I always want to thinkabout knowledge content as
having physical weight, youknow.
And if you think of it thatway, then those articles that,
you know, the the top threearticles are getting a thousand
hits a week, but the bottomthree articles are getting three
(21:53):
hits a year.
And if they have physicalweight, just jettison them, you
know, and if you can't deletethem because it makes you
scared.
Archive them, get them out ofthe AI's site.
Priscilla (22:03):
Yeah.
Jordan (22:04):
I think this happened to me recently because I was
having an issue with mydaughter's um Apple pencil.
And I just I went to ChatGBTbecause it's great for
troubleshooting things.
And it gave me this like crazyoutdated information.
And I looked at like the sourceand I realized it was from one
of those like forums that Applehas, and it like provided an
answer, but the answer was likeyears ago.
Leslie (22:25):
Yeah.
Jordan (22:26):
And like the operating system's not even the same.
Leslie (22:28):
Yeah.
Take it down.
Yeah.
Put it in, put it in the U-Haulstorage unit.
Yeah.
Priscilla (22:33):
Well, and that's a great point, Jordan, because and
I didn't even think of this,but all of the community forums
that are public that people canwrite any kind of information,
whether it's right or wrong, andthat can be pulled in as fact
into, you know, AI and bedelivered up as a response.
Right.
Something from years ago orsomething that was never true to
(22:53):
begin with because it was acommunity forum.
And how do you then go?
So now you can't just go andpost a response that says this
isn't accurate.
You need to find a way to makesure that only the accurate
stuff is staying online andavailable to people.
So, you know, we talked aboutthis a little bit.
When it comes to writing forAI, when I first reached out to
you about doing this episode,you said something that I
(23:14):
thought really hit the nail onthe head.
You said the phrase withoutinconveniencing your human
readers.
And I thought, yeah, that's it.
It's not that you want to writefor AI and fully forget them.
Obviously, you don't want to dothat.
But you also don't want to makeit written in a way that makes
it hard for them to understandand take action, like you were
saying before.
And so there's this balancethat you have to have when
you're going in to write for AIand for humans.
(23:36):
And so I want to talk aboutthat balance a little bit.
And you were talking about howthey really are not that
different, which is good newsfor us.
But I've seen so many companiesrecently with this AI first
strategy that they put outthere.
They say AI first, everythingis AI first.
Is that the right move?
Is are there any risks if yougo full in on that AI first?
(23:57):
Or is it something that becausethey're so similar, you don't
have to worry too much aboutthat?
Leslie (24:02):
Well, they're similar in heaven, you know, they're
similar in philosophy, but AIfirst, I don't even really know
what that means.
You know, I think that meanswe're getting pressure to quote,
implement AI.
So we've spent a lot of moneyand for the near term, for the
next 18 months, we're investingall our efforts on supporting
(24:26):
this AI tool.
But but what does that reallymean?
If the AI tools draw on theinformation in your knowledge
base that was to date probablycreated by humans, but
ultimately meant to serveinformation to human users, then
I mean, I don't the sourcematerial is the same.
(24:47):
You know, I think the metaphorof your of your knowledge base
is like a pantry.
I think that's a good metaphor.
So we we still have to go inthere and get the ingredients
out to help people.
So I think we need to keeptalking about humans.
I think that's probably thegreatest harm of an AI first
mentality is that we're forhowever short or long, we're
(25:09):
going to stop talking abouthuman needs.
That's the risk to me.
It's not, as I said, I thinkmost of the ways that we need to
write for AI tools also servehuman users.
There's a few that might not,but most, most of the ways do
work for human users, but wehave to still talk about their
needs.
Priscilla (25:28):
Yeah.
I think that's one of thethings that when AI was first,
you know, on the scene andbecoming this really big talking
point a couple of, you know,three or four years ago, one of
the things that I kept seeingwas this AI first.
We're AI first, our company isAI first.
And I kind of always felt ickyabout that because it makes me
feel like what you're saying isAI is the most important thing.
(25:50):
But at the end of the day, thecustomers are the most important
thing.
And what you're doing is usingAI to better serve the
customers.
So in my mind, it feels weirdbecause I still see these like
AI first in the slogan ofthings.
And I want to say, well, it'susing AI to put the customers
first.
That's what you're doing.
Right.
Um, what does that really mean,AI first?
(26:11):
Aren't we talking aboutcustomers here and keeping the
human aspect of at the end ofthe day, we are trying to find
better ways to serve humans.
We're not trying to betterserve AI.
Leslie (26:22):
Right, right, exactly.
Even though our conversation isin some ways about serving
information to AI, which is alegitimate thing to talk about,
without, I mean, I think it's umI've I've watched this kind of
industry-wide, I'm not going tocall it hysteria.
I guess it'd be likeuncontrolled excitement or
(26:43):
something like this.
I've watched it.
This is what I feel is my thirdchapter.
So chapter three is AI, chaptertwo was social media, and
chapter one was uh desktop webpublishing.
Because when I started mycompany, there were, you know,
somebody was using front page tocreate the website, and the one
(27:04):
server was under the IT guy'sdesk, you know.
But this is different.
This is different in scope, andit's taking us, I think, longer
to calm down because it isdifferent in scope.
But I'm ready for the calmdown.
Priscilla (27:19):
So I want to talk a little bit about like the
writing tips that you have forall of us who are listening who
want to be better at writinghelp articles, who want to be
better at editing thoseresponses in the inbox.
What are some of the specificwriting techniques that you can
give us that can improve helpdocuments and make them more AI
friendly?
Leslie (27:38):
Sure.
So if you're writing aknowledge article, I can give
you at least four.
And once I get going, probably24 tips for writing a knowledge
article.
The first thing is you shouldwrite a clear title, probably a
title that has a verb in it,because verbs drive meaning in a
way that nouns and noun stringscan't.
So don't call your knowledgearticle something like admin
(28:04):
account user.
You call it add an adminaccount user, or how do I add an
admin account user?
The verb add, it will, youknow, help the AI tool unite the
search to that content, butit's also a great practice for
human users.
Priscilla (28:23):
Those kind of strategies are like exactly what
I love to hear.
Like the very specific use averb in your title.
Leslie (28:30):
Yes.
The practice of adding a verbto the title, whether your
title's written in question formor whether the verb is in
gerund form, like adding anadmin account user.
This helps the human writer, orif you're having the AI tool
write the knowledge articles, ithelps the tool figure out what
the scope of the article will bebecause your adding a user may
(28:51):
be different from changing youraccount status, right?
So you don't want any of thatto slip in.
You named it adding an adminaccount user, and that's all
that it should be.
The next thing is to write thatmeal.
So the heading is the byte, andthe first couple sentences
should be a summary of theinformation that follows.
(29:12):
This is a common practice inknowledge-based writing.
This is already a practice inplace.
Some organizations do it betterthan others, but you use a
sentence or two that summarizeswhat the article will be about.
And that is, you know, we wesee those summaries march right
off of the source content, rightinto the AI search results.
(29:34):
But you have to write one inthere if you want it to uh show
up in the AI search results, orthe AI will make one for you and
it's not likely to be what youwanted.
Some organizations write anintroduction instead of a
summary, but I'm saying no, thenext best practice is to write a
summary at the beginning andkeep it short, a couple
(29:55):
sentences or three sentences.
Then you're moving into thesteps on how to do this.
Use technical writing, bestpractices from 1954 to write
those steps.
So if there's steps in asequence, number them.
Begin each step with a verb,and this is really hard to do,
(30:16):
but really important to do.
Scrub out notes or that pointwhere you say important colon
because those are throatclearings.
Those are the person writingthe procedure, not quite sure of
where to put that information.
So don't do that.
That is that is a very vaguename for a little piece of info,
(30:39):
note.
Yeah.
What does that tell the humanreader?
Not much.
And and it won't tell the AImuch either.
And the AI is dumber than us,right?
Yeah.
So we don't want to give it anote.
It doesn't know what to make ofa note.
So you have to figure out isthat note a step?
Does that cue you that you needanother section in the
procedure?
And it's actually two sets ofsix steps, not one set of 12.
(31:04):
But you know, you you say toyourself, I'm drafting this or
I'm using an AI tool to draftthis.
If the note lingers there,that's fine.
But before a human or a botuses it, I have to fix it and
get rid of those notes.
Priscilla (31:17):
I'm laughing because our help guide is riddled with
notes.
And I know this because I wrotea lot of those articles and I
put a lot of notes in there anda lot of pro tips.
Right.
You know, I'm like, oh, I don'twant them to forget this little
piece of information.
So I'm gonna put a pro tip downhere or a note, like, because I
want to make sure they knowthat you have to be on this
(31:37):
certain plan in order to usethis certain thing.
So I'm gonna throw it in as anote.
And we use those all throughoutour help articles, and now I'm
just my wheels are spinninggoing, okay, I gotta go and
revise some of these help docsand get those notes clarified.
Well, which is better, note orpro tip?
I don't know.
You tell me.
It's 50-50.
Go for it, guess.
(31:58):
I mean, I feel like they'reequally bad, is what I would
know.
Is that what I would okay?
Leslie (32:03):
A tip is, you know, if we asked 10 people on the street
now, what is a tip?
The tip is a little bit ofadvice about how to do something
or how to get what you want.
Okay.
So so I think a tip is probablybetter, but what is a note?
Yeah, what is it?
It's too generic.
And actually, your articlesmight have some kind of internal
(32:24):
strong architecture or theknowledge base might have this
kind of internal strongarchitecture.
If every time you wrote a note,it was the same type of
information.
And then I would say, awesome,leave them in there, but let's
give them a different name.
Yeah.
You know, so maybe every timeyou wrote a note, it was about
(32:45):
whether such a change wouldcause a charge.
Yeah.
Note if you use this app, itwill you have to be on the
silver plan, not the bronzeplan.
Let's and if you did that everytime, then I'd be like, first
of all, what an orderly brainshe has.
And then second of all, let'sjust call it something
(33:05):
different.
It then we're only talkingabout the word now.
Priscilla (33:08):
Yeah, like pricing consideration or something more
specific.
Right.
It's funny, I feel like I canfeel the people listening to
this who manage the helpdocuments.
I mean, I think of the personon my team who manages them, but
then I also think of people whoare listening to this going,
okay, I got some projects infront of me.
I got some things I gottaupdate because it's just so it's
(33:28):
just so practical.
And it's it'll take time, butit's an easy thing to go in and
update and change, and it'sgonna make a big difference for
the way people access yourinformation.
Leslie (33:39):
Yes, it is going to make a big difference.
Yeah.
Absolutely.
And then the next section youwant, and most knowledge
articles have, is uh access toother resources, or the the very
system sets up links to otherresources.
So all we'd want to be sure isthat it's named properly, right?
So read more is not a goodname, and this is definitely not
(34:03):
a good name for AI.
Right.
They are not smart enough andthey're not inferring, but they
have seen read more often enoughto know what it leads to.
Yeah.
Right.
But for human users, we want tosay something like more
information on adding users orrelated information to adding an
(34:23):
admin user.
So that's the best practiceright there.
And that's helpful for humansand for AI tools also.
Priscilla (34:30):
Yeah, I was gonna say one of the first things.
So when I started, I was thefirst support person here at
Buzzsprout, our head ofmarketing was the one who was
really training me because hehad been doing support and
marketing.
And so he was giving me sometraining.
And one of the things that hetold me on my first week, which
I remember it has likesolidified in my head, was when
you're in an email and you'regonna link to a resource, make
(34:53):
sure that the words that you'relinking are describing the
resource.
So it's not just saying clickhere or learn more, but it's
saying check out this videoabout adding users, and then you
link to the video.
And it's one of those thingsthat's stuck with me for the
last decade is that theimportance of being descriptive
(35:14):
when you're linking out, becausepeople are much more, I mean,
humans are much more likely toclick on a link if they know
what they're going to.
Leslie (35:22):
Right.
And a link is like a littledoor.
Yeah.
It's like a door in thecontent.
And if you're gonna walk out adoor, you should know where
you're going.
Yeah.
You know.
Priscilla (35:30):
So what else?
Are there any other strategies?
We've gone through resources.
So that's the end.
When you've got the resources,you link out to those other
places, the other ways to learn.
One thing we do at the end ofall of our help articles, and
you can tell me if this is agood strategy or not, is that we
always put our email at thebottom.
We say, if you still havequestions, here's our email.
It's in every single articlebecause we never want people to
(35:52):
finish an article and not knowhow to get in touch with us.
Yeah.
Do you think that's overkill?
Leslie (35:56):
That's a good thing.
No, no, that's wonderful.
If I had to lay out acontinuum, how are we going to
serve human users who did notget the help they need?
Over at the far right side ishere's our email, contact us,
what you're doing.
And over at the far left sideis those stupid thumbs up,
thumbs down.
(36:17):
Did this help you?
I hate those.
Yeah.
They're so stupid.
They don't help anybody, theydon't help the team get better.
Yeah.
What does it mean, thumbs down?
Priscilla (36:26):
And and when you're in the middle of trying to find
an answer to a question, you'rereally not concerned with
letting someone know that theirhelp article was not helpful.
You're like, well, no, thiswasn't helpful, but I'm I'm
still looking for my answer.
So I'm gonna keep looking.
I'm not gonna answer yoursurvey right now.
Like that's how I feel when Isee those.
So yeah, I'm right there withyou.
Yeah.
We always put there, if youstill have questions, or if this
(36:48):
didn't answer your initialquestion, here's an email
directly to us.
And so yes, I think that's agood strategy for making sure
things stay human.
Leslie (36:56):
I want to add one more point that's I think really
interesting.
It's very interesting to me asa lifelong writing teacher.
The tone of the articles andthe tone of the content that we
write for humans and for AIshould be a conversational tone.
So what's because the searchqueries are coming in in a
conversational way.
Yeah.
(37:16):
You know, nobody's querying aspursuant to my lease.
You know, what addenda should Ireview?
You know, nope.
So nobody's doing that.
So all the plain languagepractitioners who are in my
cohort of life, we have beenurging people to use a more
conversational tone in workplacewriting for many, many years.
(37:37):
And by conversational, I simplymean business writing that is
more similar to the way we speakthan it is to some of the kind
of vaunted academic and legaltones for writing that are
dissimilar from the way wespeak.
I'm not saying write the wayyou talk to your friends, I'm
saying move closer to the wayyou talk.
(38:00):
So use contractions, throw in afragment every once in a while,
because when we're talking toeach other, we sometimes say no
way.
Yeah.
Yeah.
Or we say, what now?
You know, and those kinds ofphrasings can go into our
business writing without harmingour authority or our
authenticity.
So the tools are trained ondiction, conversational diction,
(38:24):
because so much of theinformation they imbibe is
conversational.
So our writing in the knowledgearticle should also be
conversational.
Priscilla (38:33):
Yeah.
I think every single personI've trained on my team over the
last decade, I have had to talkto them about not talking in a
way that they think they'resupposed to talk, but leaning
way more into the easyconversation kind of talking.
And I feel like I see it themost in people who are a little
insecure about their writingbecause they're trying to write
(38:55):
in a way that they think they'resupposed to write to sound
smart.
And I have to go, hey, I'drather hear how you talk than
hear you use words that peopledon't use in their regular life.
Leslie (39:09):
This is a little different challenge in knowledge
articles than it is in emailresponses because the incoming
email may be from somebody who'squarrelsome or upset.
So sometimes the agent, thesupport person, kind of draws
themselves up and like a pufferfish, you know, I'm gonna make
myself big, I'm gonna makemyself sound stern and
(39:30):
authoritative.
So you'll stop asking me thesame question I told you no
already.
But in knowledge articles,that's not the relationship we
have with the reader.
And so conversational writing,even warm writing is appropriate
there.
Priscilla (39:46):
Are there any strategies that you used to
teach people that now in the AIera you don't teach anymore?
You leave out.
Leslie (39:54):
I thought about this question a lot, and I'm gonna
say something's gonna get mekicked out of the English
teacher world.
Like, I would be like, son, ifyou don't know how to use an
apostrophe now, you have graystreaks in your beard and you
don't know how to use anapostrophe now, it's over.
Don't learn it.
Yeah, because AI can do thatfor you.
AI is so much better thanMicrosoft Editor, you know.
(40:17):
It's a good copy editor, yeah.
Yeah, you it really is a goodcopy editor.
And how did you get to be 36and not know how to use an
apostrophe?
There's lots of reasons.
I don't know your particularreason, but life is short.
Give up.
Give up.
Give up.
Stop worrying about it.
But make sure that you know,you can tell the tool.
(40:39):
You know, you couldn't tellMicrosoft Editor, I'm terrible
at apostrophes, but you can tellthe tool, I'm terrible at
apostrophes.
So true.
So check specially forapostrophes.
Or you could tell the tool, I'ma terrible proofreader.
For example, I have a clientwho was telling me that the
proofreading they're doing,contact enter is slipping
(41:02):
through when they mean contactcenter, because of course enter
is a word.
Yeah.
Oh.
Right?
Yeah.
So you can tell the tool, youknow, don't let me make a
mistake where both words arespelled correctly but don't make
sense.
Yeah.
Jordan (41:17):
I feel like this is in the same vein as like math
teachers saying that you arenever, you're not gonna have a
calculator with you at alltimes, and so you have to learn
math.
And like it feels like we're inthat same vein here where it's
like you have to learnpunctuation because you're not
gonna have like an editor withyou all the time.
It's like, well, yeah, what canI do?
(41:38):
Yeah, yeah, you've gotcalculators on your phone.
Leslie (41:40):
Now I got sweaty, Jordan.
I didn't, I said if you havegray in your beard and you don't
know apostrophes, stop worryingabout it.
I did not see it.
Priscilla (41:49):
You're advocating for some take it out.
Leslie (41:52):
Yeah, no, I am not.
I absolutely am not.
Priscilla (41:55):
Yeah.
Leslie (41:56):
You need to know them, but for whatever reason, sir,
you slip through.
Yeah.
Priscilla (42:00):
So well, and I think the same goes for their, there,
and there, and your and your,you know, these same ideas of
like, if you've gotten to bethis age and you don't know it,
then AI can help you.
Yeah.
Leslie (42:14):
Yes.
And and the reason you don'tknow it, if you've gotten to be
in your 30s and you don't knowit, the reason you don't know it
is very likely because you havesome language-related learning
disability and you're you'reabsolutely coping a hundred
percent of the time.
And the fact that you're uhkind of a fluent, even charming
(42:37):
writer, and your spelling andpunctuation uh is atrocious,
right?
Props to you.
Yeah.
Use the tool.
Yeah, you know, use the tool.
Or you had an inadequateeducation for whatever reason.
Right.
But that I'm very happy about.
That I'm really happy about.
Priscilla (42:54):
Because it can take writers that are good at the
communication, but maybe don'thave all the technical details,
and it can make them so muchbetter.
And that really is a veryencouraging thing for all of us.
Leslie (43:06):
Especially in our industry.
Especially in our industry,because our industry on the pro
side is an industry where peoplewith less than conventional
academic credentials can getpromoted.
That's one thing I love aboutour industry.
It's like this person was agreat frontline agent for three
and a half years.
(43:26):
Now what?
Manager?
Now what?
Director?
Oh my God, that's great.
But a lot of times people inour industry flatline because
they don't have, you know,business professional writing
skills.
And they they the exposurecosts them that next promotion.
And AI tools can smooth thatout.
(43:48):
And that's really beautiful.
It is.
Priscilla (43:50):
Mm-hmm.
So you talk about this a littlebit in your strategies, but I
had a question on here, and Iwant to make sure we fully flesh
it out.
Things like headers and thatstructure that you talked about,
that is all.
The very important in AI erabecause the AI tools are looking
at the headings and they'relooking at the subheads and the
notes and all of that and usingthat as they are delivering
(44:12):
these responses.
Leslie (44:13):
Yes, absolutely.
And they want the H1, H2, H3.
They want the formatting andthe structure to be built a
certain way so that they canfind and prioritize the
headings.
And they want modular contentwritten briefly in a
conversational style that'saccurate, you know.
Priscilla (44:34):
You know, you talked a little bit about keywords and
how we used to be so focused onkeywords.
And now is it not as importantto have keyword-rich articles?
Leslie (44:44):
Again, I'm gonna give my answer, but I want to give it
modestly because, you know,imagine that a marketing person
is listening to this and amarketing person who developed
their chops, a marketing person,an SEO pro who can turn to a
prospective client and say, Yes,I can get you on the first page
of Google, right?
That's not me.
Yeah.
That's not me.
(45:04):
AI tools don't search keywordsin the same way that we used to,
you know, thinking back intothe keyword stuffing era and how
bad it was for human readers.
Terrible for human reasons.
You know, where the page wouldsay, our cashmere sweaters have
more cashmere than any othercashmere sweater that a cashmere
(45:25):
loving shopper could buy, youknow, that kind of thing.
So um I I think we're inanother era where uh natural
language and uh human sequencedstrings of words are uh how
we're searching.
Jordan (45:39):
I am looking forward to that hitting the uh recipe
blogs.
Like in where it's just theinformation's at the top and I
don't have to like scrollthrough like their family
history.
Leslie (45:49):
Yeah, oh my God.
Uh and those are definitely AIwritten, and they were slop when
people created them themselves,and they're just slop right
now.
Priscilla (45:58):
It's so hard.
Leslie (45:59):
Jump to recipe.
Jump to recipe.
Priscilla (46:01):
Yes.
Leslie (46:03):
Yeah.
Priscilla (46:04):
So we're talking about creating these help
articles, but we're also kind oftalking about using AI to
create the articles.
So, what are your thoughts onthe help article managers?
So the people who are managingthose knowledge bases using AI
to create the articles that aregoing to be used by AI.
Is there danger in that?
(46:24):
Like we've talked a little bitabout them, but what kind of
strategies do you recommend forpeople who are using AI to
develop the articles that AI isgoing to then use?
Leslie (46:33):
Hall of Mirrors and So I'm an English teacher, I'm a
writing teacher, and I just wantto say on behalf of my
profession, we need a hug.
Yeah.
We need a hug.
It's a little bit like being abank teller the day they
installed the ATM.
We're worried.
(46:54):
Yeah, we're worried.
So um, but also I'm over thatbecause there's so much
opportunity for people to dogreat work as writers using AI
tools.
So, yes, of course, write thearticle using AI and then serve
it to AI.
But here's the thing um that II worry a lot about.
(47:16):
Um, when you are a person usingAI tools to write for humans or
for other AI tools, you becomesomething like the food tester
to the queen.
So the AI tool made the draftknowledge article, and there you
are, and you are responsible.
You are the quality control.
(47:38):
You're the one who stops it andsays, Don't let the queen take
a bite of this, it'll kill her,or you're the one who says, Go
ahead.
And if you were born an analogwriter like me, like both of
you, and like nearly, but noteveryone we know who's listening
to this, if you were born ananalog writer, you have
(47:58):
developed the skills to evaluatea piece of writing by producing
it.
Yeah.
But we're looking at aworkforce, the youngest amongst
us in the workforce, who haven'tand won't be called on to
produce a piece of writing, butthey're still the food taster to
the queen.
They still have to be able todecide whether it's ready for
the human reader and for the AIreader also, the AI tool.
(48:23):
I have a great way that allwriters, but we'll talk about
writers in support and service,can be helped to build skills
that will enable them tocritique and improve an
AI-generated draft.
Okay.
Okay.
And that is that they have tobe helped to know more of their
(48:43):
customers and have theircustomers' reactions to the
written products of theircompany something that they have
access to.
So what I dream of is some kindof, I don't know, customer
expression group and not surveyresults, absolutely not survey
results, videos or live contactwith customers where the
(49:07):
customers talk about what is itlike to read this knowledge base
article?
What is it like for me?
Or where they talk about howthey liked or didn't like that
email response.
Because I do believe if an AItrained human writer can imagine
the customer's reaction, itwill strengthen their abilities
(49:29):
to edit what the AI tool hasgenerated.
But there's no replacementbecause if we're not learning
it, if we're not learning ourediting skills by editing, then
what we need is to make thereader come alive to us in a way
that it's kind of expensive forcompanies, but there's no
substitute for it.
(49:49):
Yeah.
Priscilla (49:50):
Yeah, I think that's really, I think that's really
interesting.
What is one takeaway that youwant people listening to this
who are maybe those people whoare managing their knowledge
bases?
What's one takeaway you wantthem to take from this as they
go and start editing andrevising their articles?
What do you want to them tofocus on?
Leslie (50:09):
I want them to focus on the top 10 or 7% of their
articles.
So I don't even want to whelmthem, much less overwhelm them.
I want to empower them and tosupport them.
So if they're sitting there andthinking, God, I wish I hadn't
listened to that.
Now I have an insurmountableto-do list and I feel like I
(50:30):
want to quit my job, you know.
No, no, that's not what I'msaying.
Those knowledge articles, anykinds of stored knowledge, are
used frequently or not at all,right?
Yeah.
Some are used way frequentlyand some are used not at all.
So if you focus on the onesthat are used really, really
frequently and you improve thewriting of the top three most
(50:51):
frequently used, you can teachthe AI tool to write like that.
And the AI tool can probablyedit a lot of the other ones.
So I would say don't getoverwhelmed.
Nothing I've said is somethingsomeone hasn't heard before.
You know, I'm just saying itwith more enthusiasm.
(51:12):
I don't know.
And just focus on the contentthat affects the quality of the
experience the most.
And we know what that is.
We have all the reporting weneed.
Yes.
Priscilla (51:22):
You can find that information and you can tackle
those first three first.
Indeed.
Indeed.
And I love that.
There's no need to get whelmedor overwhelmed.
Right.
Yeah.
Overwhelmed.
Right.
I love it.
I love it.
Well, thank you so much fortaking the time to be here,
Leslie.
I mean, I feel like we talkedabout a lot of things in this,
but I think we have some really,really incredibly practical
(51:42):
takeaways from this conversationthat I'm really excited to
implement on my team.
And I got to go get those notesout of there.
That's right.
But I just I think there's somereally great information here.
So thank you so much for comingand sharing that with us.
For anyone listening who wantsto level up their writing and
they're like, okay, how do Ibecome the next person in your
(52:03):
in one of your courses?
How do they get in touch withyou?
How do they get access to thatteaching?
Leslie (52:08):
Well, they can always come join me on LinkedIn.
They can definitely take theLinkedIn Learning Trainings.
And I don't know if folks knowthis, but if you have a public
library online account in theUS, if you can log into your
public library online, you canget LinkedIn Learning Training
for free.
Jordan (52:23):
Oh, I didn't know that.
I did not know that either.
I love libraries.
Oh my gosh.
Leslie (52:27):
Yeah, it's called LinkedIn Learning Patron or
something like that.
LinkedIn Learning, and you canget the classes for free.
Cool.
But I would be more than happyto look at a sample of anyone's
knowledge article.
And if they want to say, justgive me some feedback, you know,
including you, Priscilla, ifyou want me to look at your
notes, I would be thrilled to dothat.
And then one other thing isevery month I do a LinkedIn Live
(52:50):
broadcast with my colleague uhKim Saido Campbell, who's a
professor at the University ofNorth Texas.
And the broadcast is called FixThis Writing.
Two experts show you how tomake bad writing better.
And I'd love it if you join.
Priscilla (53:03):
I love that.
That's so great.
Well, we'll link to that in theshow notes.
So anyone who's listening,they'll have access and links to
those things and quick accessto get in touch with you.
Well, thank you again, Leslie.
This was really wonderful.
So if you have any questions ora topic that you would like us
to cover in a future episode,you can shoot us a text or a
voicemail using the Send Us FanMail link in the description.
(53:26):
As always, if you liked thisepisode, please share it with
someone who works in customersupport or someone who's just
trying to level up their writingskills or leave us a review on
Apple Podcasts.
We would love that.
Thank you again to LeslieO'Flehaven for joining us today
on today's episode.
Thank you all for listening.
Now go and make someone's day.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy, Jess Hilarious, And Charlamagne Tha God!