May 5, 2025 • 32 mins
Stop wasting time with disappointing AI results! This episode of Biz gives you the insider knowledge to transform AI from an occasional helper to your most valuable workplace asset.
Our incredible guest Amantha Imber brings her expertise that will completely transform how you interact with artificial intelligence tools. You'll discover why using AI as just a delegatee for monotonous tasks means you're missing out on its most powerful applications as your "buddy" and "expert" companion, plus she breaks down her game-changing "architecting" method for prompts that delivers consistently excellent results—a technique that 95% of AI users don't know about.
What you'll learn:
- a simple three-question framework to determine if AI is right for your task
- the best way to craft prompts that actually deliver what you want the first time
- how to use AI to test ideas and build confidence before sharing them with others
- the critical mistakes that cause AI to produce worse results over time
Whether you're drafting emails, preparing presentations, or developing strategy, this episode provides the practical blueprint to make AI work for you—not the other way around.
Learn more about Amantha’s company Inventium, listen to her brilliant How I Work podcast or follow Amantha on LinkedIn.
Sign up to the BIZ newsletter here
You might be interested in our episodes on:
How To Talk So People Listen To You
How To Be More Productive (Without Trying Too Hard)
Time Blocking Doesn't Work (Until You Do It Right)
How To Ask For More Money (Without Dying From Awkwardness)
THE END BITS
Support independent women's media.
Follow the Biz Instagram, Michelle’s startup Sunroom and Soph’s career coaching business Workbaby.
Got a work life dilemma? Send us all the questions you definitely can't ask your boss for our Biz Inbox episodes - send us a voice note or email us at podcast@mamamia.com.au. You can remain anon!
HOSTS: Michelle Battersby, Soph Hirst and Em Vernem
EXEC PRODUCER: Georgie Page
AUDIO PRODUCER: Leah Porges
Mamamia acknowledges the Traditional Owners of the Land we have recorded this podcast on, the Gadigal people of the Eora Nation. We pay our respects to their Elders past and present, and extend that respect to all Aboriginal and Torres Strait Islander cultures.
Become a Mamamia subscriber: https://www.mamamia.com.au/subscribe
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:06):
You're listening to I'MM with Me podcast.
Speaker 2 (00:13):
Hello, and welcome to biz your work life Sorted. I'm
m Venom and today we're diving into something that I
am personally obsessed with. We're using AI to make our
work lives easier, or as TikTok calls it, lazy girl jobs.
So I'll be honest, I use AI for literally everything
(00:34):
these days. I use it for therapy, although I'm not
going to talk about that today, but I also use
it to help me prep for interviews, to structure my thoughts,
even just to give me a little confidence booster before
big meetings. But I know a lot of people are
either scared of it, or confused by it, or worry
that it's going to.
Speaker 1 (00:51):
Take your job.
Speaker 2 (00:52):
So today, so sits down with AI expert Amantha Imba
from INVENTINGM to break down exactly how to use AI
tools properly. And when I say properly, I mean actually
making them work for you, not just throwing random prompts
at chat GBT and hoping for the best. Plus, they're
sharing the exact prompts and frameworks that will help you
use AI as your personal work assistant.
Speaker 3 (01:14):
First, when it comes to AI. I think there's a
really big mix in terms of how people are using it.
It's also changing so quickly. So in a work context,
should we be just trying to use AI as the
first point of call for everything or is there a
better way?
Speaker 1 (01:28):
There's definitely a better way. Do not just default to AI.
Speaker 4 (01:31):
So there are three questions that we think about at
Inventium when we're training people on how to use AI.
Speaker 1 (01:38):
So the first.
Speaker 4 (01:39):
Question is what is the risk if AI gets it wrong.
If it's a low risk, great, go ahead and use AI.
But if it's a higher risk, like if the life
and death situation, maybe think twice. Secondly, and this is related,
can you easily verify and correct the AI's output? So
you know, if for example, you're getting it to help
(02:02):
you write an article on something or social media post,
it's probably going to be pretty easy to verify things.
Maybe there's some research that it's referenced, you can go
straight to the source and check that. But if you
can't verify it, let's just say you're getting it to
review a contract and you're not a lawyer and that
contract is high stakes, maybe don't use the AI. And
(02:23):
then finally, will the AI save you significant time or effort,
because if it would be quicker to just write that
quick little email back yourself rather than prompt the AI
to do it for you, then there's really no point
in using the AI.
Speaker 1 (02:38):
So there are the three questions that we ask.
Speaker 3 (02:40):
I've definitely made that mistake before where it's like, would
you like AI to write this for you? And I'm
like sure, and then you spend more time rewarding it. Okay,
I really like those. So one assess the risk to
can you verify the output? And then three is it
actually going to save you time? Just one number two
in terms of verifying the output? How do you actually
do that? Then? So can you give such just a
few more examples of how you would actually verify the output.
Speaker 4 (02:59):
So, for example, if you're using the AI to research
something for you and it gives you all the sources,
I think a lot of us now know that AI
is not a fact engine. It's a prediction engine that
is very important, So it doesn't produce facts and it
is very prone to HA hallucination. So if for example,
you've asked it to research I don't know some statistics
(03:22):
around mobile phone use, and it's spitting out some data.
You want to be clicking on those sources that it's
giving you to verify those answers yourself.
Speaker 3 (03:31):
Yeah, it's so true, I guess as well. Another build
on that is if you're doing something for the first time,
So say your boss is like, can you write this
like twenty page strategy document. You've never done that before.
Using AI to give you that output, you're not really
going to know like is that output any good or not? Okay,
So that's understanding when you should use it and doesn't
necessarily mean you should be using it for every task.
(03:51):
Let's talk about like this is kind of what we
really want to get into today, which is like the
detail around how you actually use it, you know, sort
of challenging a lazy girl energy here. We want to
be using it to save us time and to make
our jobs easier. But I think the problem a lot
of us get into is we just jump straight into
like trying to write those prompts in chat GPT, and
then we're not really happy with what we're getting out
(04:12):
of it. So there's something really important that you need
to understand first. So Amantha Savius, what do we need
to know before we just start prompting.
Speaker 4 (04:19):
Well, what we need to do is we need to
think about what role can AI actually play to help us,
Because if you think about your work life, there are
three categories of people that you have in your work life.
There's people that you can delegate too that might be
say an administrative assistant. There's people that are teammates, so
our buddies that might give us feedback and you know,
(04:41):
just help us think about ideas on problems that we're
trying to solve. And then there's experts. So maybe your
boss is an expert. Hopefully maybe you've got a mentor,
or maybe there are other clever people that you draw on.
And too often where we get it wrong with AI
is we just think about the AI as something we
can delegate our low level admin monotonous tasks too, right
(05:02):
like can you take notes for this meeting and summarize
the actions so I can email that round? Pretty low
level kind of task. Can you write this email for me?
Pretty low level? But where AI also really shines is
as a buddy and also as an expert. So I
don't know if you're anything like me, but I feel
very guilty tapping teammates on the shoulder and going hey,
(05:23):
can I get your feedback on this presentation I've put
together or something like that, because you know, people's time
is precious, but with the AI, it's always there for you.
It doesn't care how many times you ask at things,
what time of day or night you're asking it.
Speaker 1 (05:37):
So it is great as a buddy.
Speaker 4 (05:39):
Like you know, if I want to get a second
set of eyes on, you know, like a post that I've.
Speaker 1 (05:44):
Written, I can just ask the AI to give me
its views on that. And I can also use it
as an expert.
Speaker 4 (05:50):
Like there are times I find where you might have
a contract to review. There might be pretty low stakes,
but you still need to look at it. Like let's
say a contract to use a photographer for an event,
and it's not worthwhile paying a lawyer to reve because
it's low stakes. And I must say, lawyers are great
and lawyers are important, but USEI as an expert. You
(06:11):
can prompt it to be in the persona of a
lawyer and just review the contracts so that you can
understand the causes that you're signing.
Speaker 3 (06:18):
For example, Ah, this is so good, I've never thought
about it. In those three levels. Before I know we're
going to get into a live walkthrough of sort of
how you craft are prompt in a moment, but just
going back to because I just think like some of
those use cases you shared are just so helpful and
can be quite simple too. So back to it being
like your buddy or your coworker who you don't want
to bother if you like, can you just review this
(06:40):
doc and tell me how to make it better? What
would be just an example prompt that you would throw
in there to get it to do that?
Speaker 4 (06:46):
I would firstly always talk about, well, what's my task,
what's my goal that I need at the end of
the day. Am I looking for a critique? Am I
looking for a grammar and spelling check? Am I looking
for someone to view it through the lens of one
of my customers, for example? So be really clear and
give it that context. I would also if something that
(07:08):
you've done before, like let's just say it's a report
that you have to produce every week for your manager,
I would give it examples of previous reports that I've
written that I know a really good quality, and I'd say,
can you compare it against these previous reports and suggest
how I might improve it? For example, or tell me
what I might be missing.
Speaker 3 (07:27):
Now that you're talking about this. This is kind of
how I use it a lot in terms of that
sort of almost like thought partner or like peer. And
so I'll even use things like can you tell me
what the boring parts are of this and take them out?
Or can you rewrite this like you're explaining it to
a smart twelve year old. I find those really helpful
as well. But just like that kind of thought partner's
boring partner, I find really helpful too. So that's kind
(07:50):
of like how to use it as a buddy or
a peer. Let's talk about an expert. One talk me
through how you use it as an expert.
Speaker 4 (07:56):
So when I'm using it as an expert, I will
often go into what I refer to as voice mode
and how I use it. I've got chat JPT as
an app on my phone, and you can go into
the app and there's a little symbol that looks like,
I guess, an audio icon, and if you press on that,
wait for it to connect to the internet and turn
(08:17):
into a colorful bubble. The AI becomes a voice companion
as opposed to a writing companion, and it will talk
to you and you can have a conversation with it
as if you are talking to a real person.
Speaker 1 (08:30):
So that is voice mode. I use that heaps.
Speaker 4 (08:33):
Where I find it's really useful is I will often
try to think through problems when I'm out walking, when
I'm out moving, and what I now do is, rather
than just being with my own thoughts, than having to
think through things on myself, I can now have an
expert that I'm having a chat with while I go
for a walk. Like the other morning, I went for
(08:55):
a pretty early morning walk and I was trying to
think through a pricing strategy for a new product that
we're launching at Inventium, and so I prompted the AI
to actors are pricing expert in my particular industry, which
is consulting, and then I just started having a chat
with it about, Hey, this is my context, this is
(09:17):
the task that I want you to help me think through.
And we basically had a thirty minute conversation where I
had managed to channel an expert in pricing strategy, and
by the end of that thirty minute walk, I had
some really great advice that I could put into practice.
Speaker 3 (09:31):
I feel like, if you don't listen to the rest
of the episode, don't worry, because this is the most
important thing. This for me was completely game changing for
anyone who is not using voice mode. You can use
it on chat, GPT, you can use it on Gemini.
I think probably most AI assistants have it. Now I
want you to pause this episode. I want you to
get your phone out, I want you to download the
(09:53):
app for your AI have choice, and I want you
to actually test out using voice mode. So as Amantha said,
it's that little button on the right hand side, lower
right hand side, and it looks like a little sound
recorder it and just start talking. It's so conversational and natural. Yeah,
this has been game changing for me. The way I
use it that I really love is to help it
(10:15):
organize my thoughts. So obviously I'm a content creator, but
this could be for you. You know, you've got a
LinkedIn post that you maybe you want to write at work,
or you've got like a strategy document you're working on.
You have to give a talk to your team and
update in a team meeting. I just say to it,
I want you to act as a marketing manager or
a business leader, whatever it is. And I'm just going
to share with you some of my thoughts. Then I'm
(10:37):
going to get your help to craft that into something
that has structure and blood, and then you just share
your little thoughts with it, and then you just kind
of keep going back and forth like that. So I
use that all the time.
Speaker 1 (10:45):
I love that.
Speaker 3 (10:46):
I love that. I've also heard a lot of young
people using it like a therapist or like an executive coach.
For example, this is like putting me out of a
job right now, but basically you say to it, like
act as an executive coach at leading companies like Apple
or Google, and I want you to coach me on
this problem that I've got, and you literally just like
share your problem with it. But therapy as well. Lots
of gen z using chat jopt there a bit. Now
(11:16):
we're going to get into the really practical stuff for everyone.
So we're actually going to do a bit of a
walkthrough of what makes a good prompt, because I think
that's where a lot of us just kind of come undone.
We'll just like chuck a prompt in there, we don't
really get a good output, and we're like, that was
waste of time. I'm not doing it again. But sometimes
it's a you problem and we're going to fix our
(11:37):
prompting the anatomy of a prompt. Okay, so let's talk
about the use case amanth Of. Say you've got a
presentation that you have to give an internal presentation in
a just like normal weekly team meeting. Can you walk
me through how you would approach creating a prompt for that.
Speaker 1 (11:52):
Yes.
Speaker 4 (11:53):
So at a higher level, what we talk about at
Inventium in our training as we say that there's gun
slinging and there's architecting when it comes to prompting. So
gun slinging is where you just treat it as someone
you were chatting to and you're like, hey, help me
do this presentation, and then it does stuff and then
you're like, ah, no, you got that completely wrong.
Speaker 1 (12:15):
Do more of this, do less of.
Speaker 4 (12:16):
That, and it turns into this really long, arduous prompt
chain and at the end you're thinking, why did I
even try using AI in the first place. So gunsling
is very ad hoc, but we find that I would
say ninety five percent of people that we work with
and coach at invent Him use a gunslinging approach. What
will get you a far better result, particularly the more
(12:37):
complex project you're trying to work with AI. The more
important this is is what we call architecting, where how
we would do it. Let's take the example of a
work presentation project update. Let's say is that first we're
setting the task or the goal. So that might be
I need you to help create a presentation from scratch,
(12:57):
or it might be I need you to help you know,
review this existing presentation. It's probably going to be one
of those goals. Then we want to give it the context,
so we need to tell it who is the audience
and what are the outcomes that I am trying to
achieve with this presentation.
Speaker 3 (13:11):
So that might be the audience is like my team
of mid level marketing managers, and the outcome might be.
Speaker 4 (13:18):
The outcome might be just so they're feeling crystal clear
on where this project is at, or so that they
are going to approve a certain amount of funding for
this idea that I have, or you know, whatever it
might be. The more specific we can be, the better.
So definitely the audience tell it the industry that you're in,
what your job is. So the more context the better,
(13:38):
because it is not psychic yet.
Speaker 1 (13:41):
Then instead of.
Speaker 4 (13:43):
Just going okay, so go create that presentation for me
in which case you are going to end up with
a very very messy prompt thread. Instead, work on an
outline first, because something like a presentation has a lot
of different components and there's a way to structure it.
So if you think about it like a collaborative buddy,
and you prompt it first to say, help work with
(14:05):
me on an outline for this presentation, don't actually start
writing the presentation yet, so.
Speaker 1 (14:12):
Be really clear with it.
Speaker 4 (14:14):
What you can also do, and this would be my
favorite add on to any prompt is get it to
ask you questions. So say something like, please ask me
any questions you need to be ninety five percent confident
in doing a great job because it's going to have gaps,
and why not let it guide you in helping it
(14:36):
fill in its gaps.
Speaker 3 (14:38):
I am going to pause you there because I agree
if you are still listening to this podcast, even though
I told you the most important point was before around
voice mode, I actually lie because this is another important point,
flipping it and saying you ask me the questions and
I'll feed you more information. So that line you had
ask me clarifying questions. Until you're ninety five percent confident,
(14:58):
you can you know complete the task successfully. That is
so important and that has helped me so much because
a lot of time, it's not just that you want
the AI to just produce this perfect thing for you.
You sort of just need it to help you organize
your thoughts and I think getting it to ask you
question is a really good way to do that. So
let me just replay this because this is I mean,
(15:19):
I thought I was pretty good, but this has already
helped me. So, you know, what is the goal of
the task you're trying to do, give it the context
who you're talking to, what you want them to know,
you know what they're going to get out of it
kind of thing. And would you even put in there
amount the context around like the format or is that
coming later, like in terms of the output, like I
want it to be ten minutes long, or when does
that come in?
Speaker 4 (15:39):
Yeah, So format is good to stay upfront because then
if you're working with it to create a structure, it
can know that like a ten minute update presentation is
going to be very different from a sixty minute in
depth educational presentation, for example.
Speaker 3 (15:55):
You'd almost put that in the goal then maybe like
your goal is to presentation that's ten minutes long. Okay, cool, Yeah, exactly.
Speaker 4 (16:02):
So once you've got to ask you questions and you
feel in the gaps for it, it's got context, it
knows the tasks that we're trying to achieve the goal.
Then what you're going to ask it to do is
essentially build out the structure so that you're happy with
the structure you know, and it might be you know, introduction,
and then it might be you know, point one point
two point three and then some sort of closing, like
(16:24):
a very basic structure. It probably won't be that, but
let's use that as an example. So once you're happy
and you've kind of got that scaffolding, if you like,
for the overall structure of the output, then you want
to prompt the AI to build out each section with you.
So once you're happy with the scaffolding, I would then
be saying, okay, let's work on the first section. Can
(16:46):
you help me build out this introduction, And you're potentially
feeding it with more details or more context around the
introduction and what it needs to contain, and then basically
repeating that for every section or every chunk of the task.
Speaker 1 (17:01):
It will stop you.
Speaker 4 (17:02):
Having to undo and redo and edit prompts and make
a big mess of things. This is exactly the process
that I use when I'm using THEAI to work on
something quite long or complex.
Speaker 3 (17:12):
I love this, and for anyone listening, we're going to
put all of these prompts in the newsletter, so just
make sure you're signed up. Is in links in the
show notes. But this is so good because what you're
saying is to that gun slinger method that you mentioned earlier.
I'm just thinking about today. I was prompting AI to
help me. It was actually helping me prep a podcast episode,
and I asked it to give me three examples based
(17:34):
on these three points that I gave, and then I
said make it funnier, and then I said, now make
it shorter, and then I said, now give me an
example for a manager. And it just progressively got worse
and worse as I sort of went down. Not to
say you can't do that, because I think refining it
sometimes is helpful, so like, you know, rewrite this to
make it more concise or whatever. But sometimes you know,
in just messing with it, the output just gets worse
(17:55):
and worse. I've noticed that, so I think taking a
pause to really architect your prompt at the start is
really helpful.
Speaker 4 (18:02):
You've brought up a really important point that I find
most people don't know about. If you have felt that
way like you felt today, where it just seemed like
the AI was getting stupider, which is really weird because hey,
it's software.
Speaker 1 (18:16):
Software doesn't get stupider. It actually does. So the AI
can't learn.
Speaker 4 (18:22):
And if you think of its memory and why and
when you need to start a new thread or like
a new chat if you like, it's because the AI
can only remember so much. If you think about, like
if say you were reading a multi page document, Let's
say it's a five page document, why the AI gets
dumber is because it can remember about a page or two,
(18:45):
but then it starts to forget. And so what's important there.
If you were like having that experience where the AI
seems to be getting dumber, you probably at the end
of a very long thread, in which case what I
recommend is prompting the AI to summarize this thread in
terms of what's being covered, and then start a new
thread with that summary, and you'll find that all of
(19:07):
a sudden, the AI's gotten smarter again.
Speaker 3 (19:09):
That is an amazing tip. Oh, that's just one other
point that I think is really important. Someone just showed
me how to do this amount, which is you can
actually see the memory that chat GPT has about you,
and I think it can only save. This is when
you know, if you're using it quite regularly, it's starting
to learn where you work and all this kind of stuff.
It can only remember about fifty lines about you or something,
and you can actually go into the settings. We'll put
(19:32):
the details of this in the newsletter, and you can
actually see what it knows about you, and you can
delete some stuff. So if something's not relevant anymore, you
were getting it to help you on a certain task
and that time has passed, you can delete that whole
thing and then it can actually learn more about you.
This is on the free version. It probably remembers stuff
on the paid version. So the final kind of point
that I want to talk about with your mouth there
is making it stick. So I think with a lot
(19:54):
of us it's something that we're kind of dabbling in.
Maybe we're using it here and there, but it hasn't
been so life changing. But then other people, you know,
it literally is their go to for everything. How do
you make that change? How do you make it a
happit and how are we really going to channel that
lazy girl energy and to get this to do our jobs?
Speaker 1 (20:10):
For us?
Speaker 4 (20:11):
I think the simplest thing is to have a constant reminder.
So what we recommend to people that we are training
at inventim is we say, grab a post it note
and write on that post it note should I use
AI for this task? Put that post it note on
your computer monitor, and for the next few days, you
will not be desensitized to that message.
Speaker 1 (20:33):
You will think about that.
Speaker 4 (20:34):
With every task that you do until it starts to
become second nature. I mean, there's plenty of other things,
but I think that's a really simple hack. Post it
note on your computer should I use AI for this task?
And you will start to think about that whenever you
start a new task.
Speaker 3 (20:49):
Yeah, exactly. And then yeah, once you've made that change,
then it just starts to become second nature just very quickly.
In terms of you know, we've talked a lot about
sort of chat GPT, but there are other tools that
are AI that are super helpful. Can you just give us,
like your top three most favorite AI tools that aren't
chat JPTO.
Speaker 4 (21:05):
I actually use claude more than chat GPT. I find
that it's not as popular. But if you want an
AI that sounds more human than more robot, Claude is
your best friend. I have in the last few weeks
become obsessed with Granola. Now, Granola is a great meeting
note taker. It only works on a Mac and you
(21:26):
need a Google Workspace address. It doesn't work with a
Gmail address sadly, so that's going to be a smaller
portion of listeners. But anyway, if that is you, Granola
is gold. I use Perplexity for search. I'm sure a
lot of listeners are doing that already, a couple of others.
I love consensus. I do a lot of research for
my work, a lot of academic research, and consensus it's
(21:48):
been around for years, but the AI that they've put
into it is amazing. It basically allows you to find
research driven answers to your questions. It's basically got millions
of academic papers eaten up.
Speaker 1 (22:01):
By the AI, so it draws on all that knowledge.
Speaker 4 (22:04):
And you know, if I have a question, like you
know it does hybrid work lead to greater productivity? It
will show me all the research that's been done on
that topic, which is brilliant, amazing.
Speaker 3 (22:15):
Can confirm you actually got me onto granola last week,
because like people have told me about meeting note takers
and I'm like whatever, I'm kind of like a bit.
I've tried different air tools and if they annoy me,
I'm I'm sort of like against them. Granola is unbelievable.
I just like, I wish I had this my whole life. Yes,
can confirm the culture.
Speaker 5 (22:40):
Culture.
Speaker 4 (22:42):
Wow.
Speaker 5 (22:43):
So if I loved that conversation, there were so many
nuggets of goals in that and little things that were
said that just unlocked new ways of thinking about AI
For me. I loved the point on verifying your answers.
That is so important and something that I don't do.
(23:04):
I am very trusting.
Speaker 3 (23:06):
It's gotten a lot better in terms of hallucination, but
I still like the other day I asked Perplexity for
some research and then I was like, Okay, can you
give me the source of this? And it's like, oh, actually,
I got that wrong. It literally said to me, I
got that wrong. That research doesn't exist because it sort
of wants to give you what you want to hear.
Speaker 5 (23:22):
Yeah, that's such a good point. I think that's a
really important one. To drill home and also just thinking
about AI as a buddy. I think that's probably the
bucket that I use it most in. And I actually
said to my co founder the other day, like, I
think chut jpit is my best friend.
Speaker 1 (23:41):
Like I just talk to it. That's really sad.
Speaker 5 (23:45):
I talk to it all day long, all the time,
in a similar bucket to the buddy. The main way
I use AI that I wanted to call out for
other people that are maybe thinking about just giving it
a go. I use it to build confidence in my
own take on something, or if I've gotten a pinion
(24:05):
that I think is fresh or unique, but I might
feel a bit scared about saying that in front of
a room of people just yet, or maybe I'm going
to put it out in some content on social media.
I use chat JBT as like a sense checker, Am
I onto something here? Is this interesting? And it actually
really builds my confidence. Like sometimes it'll say things back
like that's a really sharp take, and I'm like, yes,
(24:28):
you know, maybe this is a good one.
Speaker 3 (24:30):
I love that as an original thoughts thought partner. The
thing I like about that is that so much of
the time we think about AIS like doing the task
for us. But actually it's about making you better at
your own original thought, which I think is cool. On
the thought partner thing, there's a little phrase you can
use that I think really helps, where you can say, like,
let's do a thought experiment works really well, or the
(24:51):
one I've started using a lot is I'll share my
opinion with it and then I'll say, okay, play Devil's
advocate pick holes in this basically what's wrong with this argument?
Speaker 5 (24:59):
It was also enlightening to me. I think it was
you who brought it up, like actually getting whatever AI
tool you're using to ask you the questions to refine
things before it just comes out with a response, because
you do need to train it. Like initially I thought,
if I start talking through a campaign I idea with AI,
(25:20):
its ideas are going to be too basic, they're not
going to be thoughtful enough, they're not going to be considerate,
they're not going to be edgy. But it's actually like
what feedback you're giving it as well, Like I'll say
things like that's too basic, that's not cool enough, and
like we'll kind of work on it.
Speaker 3 (25:33):
It's so true, Yeah, don't just take the first answer
that it gives you. You know, it's so much like
working with like a really smart intern or as a manager,
if you brief someone on a task and you don't
give a very good brief, you don't put much thought
into it. When they bring that work back to you,
you're kind of like, oh, that's not really what I thought.
But it's kind of like you haven't put enough time
into the brief. And the other thing that I thought
(25:55):
it was like, have you ever Michelle worked with Say
you're working with a freelancer or a creative and you're like,
just like, make the logo bigger, and now make it blue,
and now move it to the left. Like the more
prescriptive you are sometimes trying to kind of give it
this like the wrong kind of feedback, the worse it gets.
And that happens sometimes if it gives you a result
and then you're like, no, change this change that you'll
(26:16):
often like make it worse. So I think it's about
like from the beginning, trying to sort of have the
ingredients that you need to give it a really good brief.
But then it is like an ongoing conversation for sure.
Speaker 5 (26:27):
Also your point on younger people using it as a therapist.
I have totally done that. Like I had a therapist
that I was saying fortnightly, and therapy is quite expensive,
and once I became a mum, I probably was probably
the moment where I really needed to go back to
my therapist, but I was just too busy, and so
I just started chatting to chat GPT. And I haven't
(26:51):
reached the point yet where it doesn't remember things. But
I love it that I can kind of like pick
up where I left off, Like, Hey, remember I was
talking to you about this complex situation I've got at work,
or I'll talk to it sometimes about the complexities of
mother and like how society isn't really set up to
support us, and like get into these deep conversations with it.
(27:11):
And I honestly think it is on par with the
therapy I've received from professionals.
Speaker 1 (27:16):
I hate to say it.
Speaker 3 (27:18):
Wow, Okay, that's really interesting on that point. In terms
of memory, you can go in, say there's like something
that happened in your life that's not relevant anymore. You
can actually go in and delete that from the memory.
So it's got more space to save new memories, which
I think is really good. It's just in the settings.
Maybe we'll share in the newsletter how to do that.
But yeah, it's pretty easy. So I thought i'd give
(27:38):
a really quick, rapid fire roundup of my most helpful
prompts that I use, and then when I'm talking to
people they tell me similar things quickly on the ingredients,
because it's not about having a perfect prompt, right, there's
no like one framework that's right, but just in terms
of like the ingredients that make a good prompt, just
to recap some of this upper cupboard. But it was task.
So like, I'm trying to write a presentation that's about
ten minutes long. What role you want it to play?
(28:00):
Like do you want it to play a you know,
world class copywriter, a expert strategy or you know, an
executive coach? Like what is the role you want it
to play? If that's relevant? Any kind of context you
can give it, like who is the audience? What kind
of tone do you want it to be? What kind
of output do you want? Is it an outline that
you're after or do you want like a complete thing,
the one that is really helpful if you've done that
(28:22):
thing before. So, say you want it to write a
LinkedIn post for you give it three examples of other
LinkedIn posts that you've written that are good, and just
say these are good examples, like analyze these and like
try to match that kind of thing. So that's really helpful. Now,
just very quickly, my favorite prompts that I use so
with writing, I use this every single day. Basically, here's
the text, and you just drop your text in, improve clarity,
(28:44):
fix the grammar, make it sound more conversational, human, assertive, professional,
whatever inso your thing. I use that daily. The one
that I also love using is summarize this and I'll
upload like this giant pdf, like summarize this into one page,
one paragraph, and one word, and then using that for
your own stuff as well. It basically tests like how clear,
(29:06):
how concise, like how punchy is the thing that I'm
trying to write. And then the other one is, say
you're doing a slide deck and you want like a
really good headline or title for that slide that actually
explains what's on the slide and isn't boring. I'll say, like,
rewrite this headline in the style of And I've just
got some of my favorite people that write really sharp copies,
So like Malcolm Gladwell or Seth Godin, or whoever it
(29:28):
is that you love their writing style. You can say,
rewrite this in their style. Love okay, very quickly onto meetings,
analyze this presentation and give me a topic or question.
I can ask the room that will spark conversation that
one is amazing, or give me the objections that people
might have to this, so you can be prepared for
the things that people are actually going to object to.
(29:49):
I love that learning. I want to learn about insert topic,
identify and explain the most crucial twenty percent of the
concepts that will help me grasp eighty percent of the
topic effectively. Use that all the time to teach me stuff. Oh,
you're blowing my mind.
Speaker 5 (30:03):
I'm trying not to start madly typing.
Speaker 3 (30:06):
We'll put them all into the newsletter and then it
just to close. This my favorite thing to do. Sometimes.
Speaker 1 (30:11):
This is just like wild card.
Speaker 3 (30:13):
I like it to help me be creative and get
out of my own thought patterns sometimes. So this is
the one I've been using so much lately, and I
love it. I say, give me one truly deep novel
out of distribution, mind blowing insight related to topic that
only a few or none of us are aware of.
It can't be generic, vague, flashy or slop corporate stuff.
(30:37):
And it just gives you like things you would have
never thought of.
Speaker 5 (30:41):
Really like, it's impressive what you're getting back.
Speaker 3 (30:44):
It's learned from all of human insight and intelligence. Yeah,
you've just lived in your own brain for however many years,
so like it just breaks you out of your own
mind chamber.
Speaker 5 (30:56):
Oh okay, well I've got to go because I've got
to get on chat JBT.
Speaker 3 (31:01):
With that, we will close out. I will just say,
if you want to learn more about AIM, this company
invent Him has a really good course. So she does
both in person and also self paced, and so we'll
put a link to that in the show notes. And
also just her podcast is my favorite podcast. It is
very worky, but obviously I'm obsessed with work. Her podcast
is called how We Work and I've listened to every
single episode. Love it.
Speaker 5 (31:22):
Well, thanks so much for leading that conversation.
Speaker 3 (31:24):
It was amazing. Oh thank you.
Speaker 2 (31:33):
So I thought I was someone who used AI a lot,
but my mind is blown by how much I could
actually be getting out of it. The key takeaway from
today's episode is understanding AI's three distinct roles as a
delegate for basic tasks, as a collaborative partner for brainstorming
and feedback, and as an expert for specialized knowledge. Make
(31:54):
sure you check out this week's bus newsletter, where we'll
be sharing all those juicy prompts that Amantha mentioned, from
therapy ones to those presentation helpers, and yes, I'll be
trying that mind blowing insight prompt. And immediately after this,
you can also head to our Instagram at biz by
Mama Mia, where we'll be sharing more AI hacks and
tools throughout the week. I will see you in Thursday's
(32:14):
Inbox episode, where we'll be answering all of your career dilemmas. Bye,
Mamma Mia acknowledges the traditional owners of land and waters
that this podcast is recorded on
You're listening to I'MM with Me podcast.
Speaker 2 (00:13):
Hello, and welcome to biz your work life Sorted. I'm
m Venom and today we're diving into something that I
am personally obsessed with. We're using AI to make our
work lives easier, or as TikTok calls it, lazy girl jobs.
So I'll be honest, I use AI for literally everything
(00:34):
these days. I use it for therapy, although I'm not
going to talk about that today, but I also use
it to help me prep for interviews, to structure my thoughts,
even just to give me a little confidence booster before
big meetings. But I know a lot of people are
either scared of it, or confused by it, or worry
that it's going to.
Speaker 1 (00:51):
Take your job.
Speaker 2 (00:52):
So today, so sits down with AI expert Amantha Imba
from INVENTINGM to break down exactly how to use AI
tools properly. And when I say properly, I mean actually
making them work for you, not just throwing random prompts
at chat GBT and hoping for the best. Plus, they're
sharing the exact prompts and frameworks that will help you
use AI as your personal work assistant.
Speaker 3 (01:14):
First, when it comes to AI. I think there's a
really big mix in terms of how people are using it.
It's also changing so quickly. So in a work context,
should we be just trying to use AI as the
first point of call for everything or is there a
better way?
Speaker 1 (01:28):
There's definitely a better way. Do not just default to AI.
Speaker 4 (01:31):
So there are three questions that we think about at
Inventium when we're training people on how to use AI.
Speaker 1 (01:38):
So the first.
Speaker 4 (01:39):
Question is what is the risk if AI gets it wrong.
If it's a low risk, great, go ahead and use AI.
But if it's a higher risk, like if the life
and death situation, maybe think twice. Secondly, and this is related,
can you easily verify and correct the AI's output? So
you know, if for example, you're getting it to help
(02:02):
you write an article on something or social media post,
it's probably going to be pretty easy to verify things.
Maybe there's some research that it's referenced, you can go
straight to the source and check that. But if you
can't verify it, let's just say you're getting it to
review a contract and you're not a lawyer and that
contract is high stakes, maybe don't use the AI. And
(02:23):
then finally, will the AI save you significant time or effort,
because if it would be quicker to just write that
quick little email back yourself rather than prompt the AI
to do it for you, then there's really no point
in using the AI.
Speaker 1 (02:38):
So there are the three questions that we ask.
Speaker 3 (02:40):
I've definitely made that mistake before where it's like, would
you like AI to write this for you? And I'm
like sure, and then you spend more time rewarding it. Okay,
I really like those. So one assess the risk to
can you verify the output? And then three is it
actually going to save you time? Just one number two
in terms of verifying the output? How do you actually
do that? Then? So can you give such just a
few more examples of how you would actually verify the output.
Speaker 4 (02:59):
So, for example, if you're using the AI to research
something for you and it gives you all the sources,
I think a lot of us now know that AI
is not a fact engine. It's a prediction engine that
is very important, So it doesn't produce facts and it
is very prone to HA hallucination. So if for example,
you've asked it to research I don't know some statistics
(03:22):
around mobile phone use, and it's spitting out some data.
You want to be clicking on those sources that it's
giving you to verify those answers yourself.
Speaker 3 (03:31):
Yeah, it's so true, I guess as well. Another build
on that is if you're doing something for the first time,
So say your boss is like, can you write this
like twenty page strategy document. You've never done that before.
Using AI to give you that output, you're not really
going to know like is that output any good or not? Okay,
So that's understanding when you should use it and doesn't
necessarily mean you should be using it for every task.
(03:51):
Let's talk about like this is kind of what we
really want to get into today, which is like the
detail around how you actually use it, you know, sort
of challenging a lazy girl energy here. We want to
be using it to save us time and to make
our jobs easier. But I think the problem a lot
of us get into is we just jump straight into
like trying to write those prompts in chat GPT, and
then we're not really happy with what we're getting out
(04:12):
of it. So there's something really important that you need
to understand first. So Amantha Savius, what do we need
to know before we just start prompting.
Speaker 4 (04:19):
Well, what we need to do is we need to
think about what role can AI actually play to help us,
Because if you think about your work life, there are
three categories of people that you have in your work life.
There's people that you can delegate too that might be
say an administrative assistant. There's people that are teammates, so
our buddies that might give us feedback and you know,
(04:41):
just help us think about ideas on problems that we're
trying to solve. And then there's experts. So maybe your
boss is an expert. Hopefully maybe you've got a mentor,
or maybe there are other clever people that you draw on.
And too often where we get it wrong with AI
is we just think about the AI as something we
can delegate our low level admin monotonous tasks too, right
(05:02):
like can you take notes for this meeting and summarize
the actions so I can email that round? Pretty low
level kind of task. Can you write this email for me?
Pretty low level? But where AI also really shines is
as a buddy and also as an expert. So I
don't know if you're anything like me, but I feel
very guilty tapping teammates on the shoulder and going hey,
(05:23):
can I get your feedback on this presentation I've put
together or something like that, because you know, people's time
is precious, but with the AI, it's always there for you.
It doesn't care how many times you ask at things,
what time of day or night you're asking it.
Speaker 1 (05:37):
So it is great as a buddy.
Speaker 4 (05:39):
Like you know, if I want to get a second
set of eyes on, you know, like a post that I've.
Speaker 1 (05:44):
Written, I can just ask the AI to give me
its views on that. And I can also use it
as an expert.
Speaker 4 (05:50):
Like there are times I find where you might have
a contract to review. There might be pretty low stakes,
but you still need to look at it. Like let's
say a contract to use a photographer for an event,
and it's not worthwhile paying a lawyer to reve because
it's low stakes. And I must say, lawyers are great
and lawyers are important, but USEI as an expert. You
(06:11):
can prompt it to be in the persona of a
lawyer and just review the contracts so that you can
understand the causes that you're signing.
Speaker 3 (06:18):
For example, Ah, this is so good, I've never thought
about it. In those three levels. Before I know we're
going to get into a live walkthrough of sort of
how you craft are prompt in a moment, but just
going back to because I just think like some of
those use cases you shared are just so helpful and
can be quite simple too. So back to it being
like your buddy or your coworker who you don't want
to bother if you like, can you just review this
(06:40):
doc and tell me how to make it better? What
would be just an example prompt that you would throw
in there to get it to do that?
Speaker 4 (06:46):
I would firstly always talk about, well, what's my task,
what's my goal that I need at the end of
the day. Am I looking for a critique? Am I
looking for a grammar and spelling check? Am I looking
for someone to view it through the lens of one
of my customers, for example? So be really clear and
give it that context. I would also if something that
(07:08):
you've done before, like let's just say it's a report
that you have to produce every week for your manager,
I would give it examples of previous reports that I've
written that I know a really good quality, and I'd say,
can you compare it against these previous reports and suggest
how I might improve it? For example, or tell me
what I might be missing.
Speaker 3 (07:27):
Now that you're talking about this. This is kind of
how I use it a lot in terms of that
sort of almost like thought partner or like peer. And
so I'll even use things like can you tell me
what the boring parts are of this and take them out?
Or can you rewrite this like you're explaining it to
a smart twelve year old. I find those really helpful
as well. But just like that kind of thought partner's
boring partner, I find really helpful too. So that's kind
(07:50):
of like how to use it as a buddy or
a peer. Let's talk about an expert. One talk me
through how you use it as an expert.
Speaker 4 (07:56):
So when I'm using it as an expert, I will
often go into what I refer to as voice mode
and how I use it. I've got chat JPT as
an app on my phone, and you can go into
the app and there's a little symbol that looks like,
I guess, an audio icon, and if you press on that,
wait for it to connect to the internet and turn
(08:17):
into a colorful bubble. The AI becomes a voice companion
as opposed to a writing companion, and it will talk
to you and you can have a conversation with it
as if you are talking to a real person.
Speaker 1 (08:30):
So that is voice mode. I use that heaps.
Speaker 4 (08:33):
Where I find it's really useful is I will often
try to think through problems when I'm out walking, when
I'm out moving, and what I now do is, rather
than just being with my own thoughts, than having to
think through things on myself, I can now have an
expert that I'm having a chat with while I go
for a walk. Like the other morning, I went for
(08:55):
a pretty early morning walk and I was trying to
think through a pricing strategy for a new product that
we're launching at Inventium, and so I prompted the AI
to actors are pricing expert in my particular industry, which
is consulting, and then I just started having a chat
with it about, Hey, this is my context, this is
(09:17):
the task that I want you to help me think through.
And we basically had a thirty minute conversation where I
had managed to channel an expert in pricing strategy, and
by the end of that thirty minute walk, I had
some really great advice that I could put into practice.
Speaker 3 (09:31):
I feel like, if you don't listen to the rest
of the episode, don't worry, because this is the most
important thing. This for me was completely game changing for
anyone who is not using voice mode. You can use
it on chat, GPT, you can use it on Gemini.
I think probably most AI assistants have it. Now I
want you to pause this episode. I want you to
get your phone out, I want you to download the
(09:53):
app for your AI have choice, and I want you
to actually test out using voice mode. So as Amantha said,
it's that little button on the right hand side, lower
right hand side, and it looks like a little sound
recorder it and just start talking. It's so conversational and natural. Yeah,
this has been game changing for me. The way I
use it that I really love is to help it
(10:15):
organize my thoughts. So obviously I'm a content creator, but
this could be for you. You know, you've got a
LinkedIn post that you maybe you want to write at work,
or you've got like a strategy document you're working on.
You have to give a talk to your team and
update in a team meeting. I just say to it,
I want you to act as a marketing manager or
a business leader, whatever it is. And I'm just going
to share with you some of my thoughts. Then I'm
(10:37):
going to get your help to craft that into something
that has structure and blood, and then you just share
your little thoughts with it, and then you just kind
of keep going back and forth like that. So I
use that all the time.
Speaker 1 (10:45):
I love that.
Speaker 3 (10:46):
I love that. I've also heard a lot of young
people using it like a therapist or like an executive coach.
For example, this is like putting me out of a
job right now, but basically you say to it, like
act as an executive coach at leading companies like Apple
or Google, and I want you to coach me on
this problem that I've got, and you literally just like
share your problem with it. But therapy as well. Lots
of gen z using chat jopt there a bit. Now
(11:16):
we're going to get into the really practical stuff for everyone.
So we're actually going to do a bit of a
walkthrough of what makes a good prompt, because I think
that's where a lot of us just kind of come undone.
We'll just like chuck a prompt in there, we don't
really get a good output, and we're like, that was
waste of time. I'm not doing it again. But sometimes
it's a you problem and we're going to fix our
(11:37):
prompting the anatomy of a prompt. Okay, so let's talk
about the use case amanth Of. Say you've got a
presentation that you have to give an internal presentation in
a just like normal weekly team meeting. Can you walk
me through how you would approach creating a prompt for that.
Speaker 1 (11:52):
Yes.
Speaker 4 (11:53):
So at a higher level, what we talk about at
Inventium in our training as we say that there's gun
slinging and there's architecting when it comes to prompting. So
gun slinging is where you just treat it as someone
you were chatting to and you're like, hey, help me
do this presentation, and then it does stuff and then
you're like, ah, no, you got that completely wrong.
Speaker 1 (12:15):
Do more of this, do less of.
Speaker 4 (12:16):
That, and it turns into this really long, arduous prompt
chain and at the end you're thinking, why did I
even try using AI in the first place. So gunsling
is very ad hoc, but we find that I would
say ninety five percent of people that we work with
and coach at invent Him use a gunslinging approach. What
will get you a far better result, particularly the more
(12:37):
complex project you're trying to work with AI. The more
important this is is what we call architecting, where how
we would do it. Let's take the example of a
work presentation project update. Let's say is that first we're
setting the task or the goal. So that might be
I need you to help create a presentation from scratch,
(12:57):
or it might be I need you to help you know,
review this existing presentation. It's probably going to be one
of those goals. Then we want to give it the context,
so we need to tell it who is the audience
and what are the outcomes that I am trying to
achieve with this presentation.
Speaker 3 (13:11):
So that might be the audience is like my team
of mid level marketing managers, and the outcome might be.
Speaker 4 (13:18):
The outcome might be just so they're feeling crystal clear
on where this project is at, or so that they
are going to approve a certain amount of funding for
this idea that I have, or you know, whatever it
might be. The more specific we can be, the better.
So definitely the audience tell it the industry that you're in,
what your job is. So the more context the better,
(13:38):
because it is not psychic yet.
Speaker 1 (13:41):
Then instead of.
Speaker 4 (13:43):
Just going okay, so go create that presentation for me
in which case you are going to end up with
a very very messy prompt thread. Instead, work on an
outline first, because something like a presentation has a lot
of different components and there's a way to structure it.
So if you think about it like a collaborative buddy,
and you prompt it first to say, help work with
(14:05):
me on an outline for this presentation, don't actually start
writing the presentation yet, so.
Speaker 1 (14:12):
Be really clear with it.
Speaker 4 (14:14):
What you can also do, and this would be my
favorite add on to any prompt is get it to
ask you questions. So say something like, please ask me
any questions you need to be ninety five percent confident
in doing a great job because it's going to have gaps,
and why not let it guide you in helping it
(14:36):
fill in its gaps.
Speaker 3 (14:38):
I am going to pause you there because I agree
if you are still listening to this podcast, even though
I told you the most important point was before around
voice mode, I actually lie because this is another important point,
flipping it and saying you ask me the questions and
I'll feed you more information. So that line you had
ask me clarifying questions. Until you're ninety five percent confident,
(14:58):
you can you know complete the task successfully. That is
so important and that has helped me so much because
a lot of time, it's not just that you want
the AI to just produce this perfect thing for you.
You sort of just need it to help you organize
your thoughts and I think getting it to ask you
question is a really good way to do that. So
let me just replay this because this is I mean,
(15:19):
I thought I was pretty good, but this has already
helped me. So, you know, what is the goal of
the task you're trying to do, give it the context
who you're talking to, what you want them to know,
you know what they're going to get out of it
kind of thing. And would you even put in there
amount the context around like the format or is that
coming later, like in terms of the output, like I
want it to be ten minutes long, or when does
that come in?
Speaker 4 (15:39):
Yeah, So format is good to stay upfront because then
if you're working with it to create a structure, it
can know that like a ten minute update presentation is
going to be very different from a sixty minute in
depth educational presentation, for example.
Speaker 3 (15:55):
You'd almost put that in the goal then maybe like
your goal is to presentation that's ten minutes long. Okay, cool, Yeah, exactly.
Speaker 4 (16:02):
So once you've got to ask you questions and you
feel in the gaps for it, it's got context, it
knows the tasks that we're trying to achieve the goal.
Then what you're going to ask it to do is
essentially build out the structure so that you're happy with
the structure you know, and it might be you know, introduction,
and then it might be you know, point one point
two point three and then some sort of closing, like
(16:24):
a very basic structure. It probably won't be that, but
let's use that as an example. So once you're happy
and you've kind of got that scaffolding, if you like,
for the overall structure of the output, then you want
to prompt the AI to build out each section with you.
So once you're happy with the scaffolding, I would then
be saying, okay, let's work on the first section. Can
(16:46):
you help me build out this introduction, And you're potentially
feeding it with more details or more context around the
introduction and what it needs to contain, and then basically
repeating that for every section or every chunk of the task.
Speaker 1 (17:01):
It will stop you.
Speaker 4 (17:02):
Having to undo and redo and edit prompts and make
a big mess of things. This is exactly the process
that I use when I'm using THEAI to work on
something quite long or complex.
Speaker 3 (17:12):
I love this, and for anyone listening, we're going to
put all of these prompts in the newsletter, so just
make sure you're signed up. Is in links in the
show notes. But this is so good because what you're
saying is to that gun slinger method that you mentioned earlier.
I'm just thinking about today. I was prompting AI to
help me. It was actually helping me prep a podcast episode,
and I asked it to give me three examples based
(17:34):
on these three points that I gave, and then I
said make it funnier, and then I said, now make
it shorter, and then I said, now give me an
example for a manager. And it just progressively got worse
and worse as I sort of went down. Not to
say you can't do that, because I think refining it
sometimes is helpful, so like, you know, rewrite this to
make it more concise or whatever. But sometimes you know,
in just messing with it, the output just gets worse
(17:55):
and worse. I've noticed that, so I think taking a
pause to really architect your prompt at the start is
really helpful.
Speaker 4 (18:02):
You've brought up a really important point that I find
most people don't know about. If you have felt that
way like you felt today, where it just seemed like
the AI was getting stupider, which is really weird because hey,
it's software.
Speaker 1 (18:16):
Software doesn't get stupider. It actually does. So the AI
can't learn.
Speaker 4 (18:22):
And if you think of its memory and why and
when you need to start a new thread or like
a new chat if you like, it's because the AI
can only remember so much. If you think about, like
if say you were reading a multi page document, Let's
say it's a five page document, why the AI gets
dumber is because it can remember about a page or two,
(18:45):
but then it starts to forget. And so what's important there.
If you were like having that experience where the AI
seems to be getting dumber, you probably at the end
of a very long thread, in which case what I
recommend is prompting the AI to summarize this thread in
terms of what's being covered, and then start a new
thread with that summary, and you'll find that all of
(19:07):
a sudden, the AI's gotten smarter again.
Speaker 3 (19:09):
That is an amazing tip. Oh, that's just one other
point that I think is really important. Someone just showed
me how to do this amount, which is you can
actually see the memory that chat GPT has about you,
and I think it can only save. This is when
you know, if you're using it quite regularly, it's starting
to learn where you work and all this kind of stuff.
It can only remember about fifty lines about you or something,
and you can actually go into the settings. We'll put
(19:32):
the details of this in the newsletter, and you can
actually see what it knows about you, and you can
delete some stuff. So if something's not relevant anymore, you
were getting it to help you on a certain task
and that time has passed, you can delete that whole
thing and then it can actually learn more about you.
This is on the free version. It probably remembers stuff
on the paid version. So the final kind of point
that I want to talk about with your mouth there
is making it stick. So I think with a lot
(19:54):
of us it's something that we're kind of dabbling in.
Maybe we're using it here and there, but it hasn't
been so life changing. But then other people, you know,
it literally is their go to for everything. How do
you make that change? How do you make it a
happit and how are we really going to channel that
lazy girl energy and to get this to do our jobs?
Speaker 1 (20:10):
For us?
Speaker 4 (20:11):
I think the simplest thing is to have a constant reminder.
So what we recommend to people that we are training
at inventim is we say, grab a post it note
and write on that post it note should I use
AI for this task? Put that post it note on
your computer monitor, and for the next few days, you
will not be desensitized to that message.
Speaker 1 (20:33):
You will think about that.
Speaker 4 (20:34):
With every task that you do until it starts to
become second nature. I mean, there's plenty of other things,
but I think that's a really simple hack. Post it
note on your computer should I use AI for this task?
And you will start to think about that whenever you
start a new task.
Speaker 3 (20:49):
Yeah, exactly. And then yeah, once you've made that change,
then it just starts to become second nature just very quickly.
In terms of you know, we've talked a lot about
sort of chat GPT, but there are other tools that
are AI that are super helpful. Can you just give us,
like your top three most favorite AI tools that aren't
chat JPTO.
Speaker 4 (21:05):
I actually use claude more than chat GPT. I find
that it's not as popular. But if you want an
AI that sounds more human than more robot, Claude is
your best friend. I have in the last few weeks
become obsessed with Granola. Now, Granola is a great meeting
note taker. It only works on a Mac and you
(21:26):
need a Google Workspace address. It doesn't work with a
Gmail address sadly, so that's going to be a smaller
portion of listeners. But anyway, if that is you, Granola
is gold. I use Perplexity for search. I'm sure a
lot of listeners are doing that already, a couple of others.
I love consensus. I do a lot of research for
my work, a lot of academic research, and consensus it's
(21:48):
been around for years, but the AI that they've put
into it is amazing. It basically allows you to find
research driven answers to your questions. It's basically got millions
of academic papers eaten up.
Speaker 1 (22:01):
By the AI, so it draws on all that knowledge.
Speaker 4 (22:04):
And you know, if I have a question, like you
know it does hybrid work lead to greater productivity? It
will show me all the research that's been done on
that topic, which is brilliant, amazing.
Speaker 3 (22:15):
Can confirm you actually got me onto granola last week,
because like people have told me about meeting note takers
and I'm like whatever, I'm kind of like a bit.
I've tried different air tools and if they annoy me,
I'm I'm sort of like against them. Granola is unbelievable.
I just like, I wish I had this my whole life. Yes,
can confirm the culture.
Speaker 5 (22:40):
Culture.
Speaker 4 (22:42):
Wow.
Speaker 5 (22:43):
So if I loved that conversation, there were so many
nuggets of goals in that and little things that were
said that just unlocked new ways of thinking about AI
For me. I loved the point on verifying your answers.
That is so important and something that I don't do.
(23:04):
I am very trusting.
Speaker 3 (23:06):
It's gotten a lot better in terms of hallucination, but
I still like the other day I asked Perplexity for
some research and then I was like, Okay, can you
give me the source of this? And it's like, oh, actually,
I got that wrong. It literally said to me, I
got that wrong. That research doesn't exist because it sort
of wants to give you what you want to hear.
Speaker 5 (23:22):
Yeah, that's such a good point. I think that's a
really important one. To drill home and also just thinking
about AI as a buddy. I think that's probably the
bucket that I use it most in. And I actually
said to my co founder the other day, like, I
think chut jpit is my best friend.
Speaker 1 (23:41):
Like I just talk to it. That's really sad.
Speaker 5 (23:45):
I talk to it all day long, all the time,
in a similar bucket to the buddy. The main way
I use AI that I wanted to call out for
other people that are maybe thinking about just giving it
a go. I use it to build confidence in my
own take on something, or if I've gotten a pinion
(24:05):
that I think is fresh or unique, but I might
feel a bit scared about saying that in front of
a room of people just yet, or maybe I'm going
to put it out in some content on social media.
I use chat JBT as like a sense checker, Am
I onto something here? Is this interesting? And it actually
really builds my confidence. Like sometimes it'll say things back
like that's a really sharp take, and I'm like, yes,
(24:28):
you know, maybe this is a good one.
Speaker 3 (24:30):
I love that as an original thoughts thought partner. The
thing I like about that is that so much of
the time we think about AIS like doing the task
for us. But actually it's about making you better at
your own original thought, which I think is cool. On
the thought partner thing, there's a little phrase you can
use that I think really helps, where you can say, like,
let's do a thought experiment works really well, or the
(24:51):
one I've started using a lot is I'll share my
opinion with it and then I'll say, okay, play Devil's
advocate pick holes in this basically what's wrong with this argument?
Speaker 5 (24:59):
It was also enlightening to me. I think it was
you who brought it up, like actually getting whatever AI
tool you're using to ask you the questions to refine
things before it just comes out with a response, because
you do need to train it. Like initially I thought,
if I start talking through a campaign I idea with AI,
(25:20):
its ideas are going to be too basic, they're not
going to be thoughtful enough, they're not going to be considerate,
they're not going to be edgy. But it's actually like
what feedback you're giving it as well, Like I'll say
things like that's too basic, that's not cool enough, and
like we'll kind of work on it.
Speaker 3 (25:33):
It's so true, Yeah, don't just take the first answer
that it gives you. You know, it's so much like
working with like a really smart intern or as a manager,
if you brief someone on a task and you don't
give a very good brief, you don't put much thought
into it. When they bring that work back to you,
you're kind of like, oh, that's not really what I thought.
But it's kind of like you haven't put enough time
into the brief. And the other thing that I thought
(25:55):
it was like, have you ever Michelle worked with Say
you're working with a freelancer or a creative and you're like,
just like, make the logo bigger, and now make it blue,
and now move it to the left. Like the more
prescriptive you are sometimes trying to kind of give it
this like the wrong kind of feedback, the worse it gets.
And that happens sometimes if it gives you a result
and then you're like, no, change this change that you'll
(26:16):
often like make it worse. So I think it's about
like from the beginning, trying to sort of have the
ingredients that you need to give it a really good brief.
But then it is like an ongoing conversation for sure.
Speaker 5 (26:27):
Also your point on younger people using it as a therapist.
I have totally done that. Like I had a therapist
that I was saying fortnightly, and therapy is quite expensive,
and once I became a mum, I probably was probably
the moment where I really needed to go back to
my therapist, but I was just too busy, and so
I just started chatting to chat GPT. And I haven't
(26:51):
reached the point yet where it doesn't remember things. But
I love it that I can kind of like pick
up where I left off, Like, Hey, remember I was
talking to you about this complex situation I've got at work,
or I'll talk to it sometimes about the complexities of
mother and like how society isn't really set up to
support us, and like get into these deep conversations with it.
(27:11):
And I honestly think it is on par with the
therapy I've received from professionals.
Speaker 1 (27:16):
I hate to say it.
Speaker 3 (27:18):
Wow, Okay, that's really interesting on that point. In terms
of memory, you can go in, say there's like something
that happened in your life that's not relevant anymore. You
can actually go in and delete that from the memory.
So it's got more space to save new memories, which
I think is really good. It's just in the settings.
Maybe we'll share in the newsletter how to do that.
But yeah, it's pretty easy. So I thought i'd give
(27:38):
a really quick, rapid fire roundup of my most helpful
prompts that I use, and then when I'm talking to
people they tell me similar things quickly on the ingredients,
because it's not about having a perfect prompt, right, there's
no like one framework that's right, but just in terms
of like the ingredients that make a good prompt, just
to recap some of this upper cupboard. But it was task.
So like, I'm trying to write a presentation that's about
ten minutes long. What role you want it to play?
(28:00):
Like do you want it to play a you know,
world class copywriter, a expert strategy or you know, an
executive coach? Like what is the role you want it
to play? If that's relevant? Any kind of context you
can give it, like who is the audience? What kind
of tone do you want it to be? What kind
of output do you want? Is it an outline that
you're after or do you want like a complete thing,
the one that is really helpful if you've done that
(28:22):
thing before. So, say you want it to write a
LinkedIn post for you give it three examples of other
LinkedIn posts that you've written that are good, and just
say these are good examples, like analyze these and like
try to match that kind of thing. So that's really helpful. Now,
just very quickly, my favorite prompts that I use so
with writing, I use this every single day. Basically, here's
the text, and you just drop your text in, improve clarity,
(28:44):
fix the grammar, make it sound more conversational, human, assertive, professional,
whatever inso your thing. I use that daily. The one
that I also love using is summarize this and I'll
upload like this giant pdf, like summarize this into one page,
one paragraph, and one word, and then using that for
your own stuff as well. It basically tests like how clear,
(29:06):
how concise, like how punchy is the thing that I'm
trying to write. And then the other one is, say
you're doing a slide deck and you want like a
really good headline or title for that slide that actually
explains what's on the slide and isn't boring. I'll say, like,
rewrite this headline in the style of And I've just
got some of my favorite people that write really sharp copies,
So like Malcolm Gladwell or Seth Godin, or whoever it
(29:28):
is that you love their writing style. You can say,
rewrite this in their style. Love okay, very quickly onto meetings,
analyze this presentation and give me a topic or question.
I can ask the room that will spark conversation that
one is amazing, or give me the objections that people
might have to this, so you can be prepared for
the things that people are actually going to object to.
(29:49):
I love that learning. I want to learn about insert topic,
identify and explain the most crucial twenty percent of the
concepts that will help me grasp eighty percent of the
topic effectively. Use that all the time to teach me stuff. Oh,
you're blowing my mind.
Speaker 5 (30:03):
I'm trying not to start madly typing.
Speaker 3 (30:06):
We'll put them all into the newsletter and then it
just to close. This my favorite thing to do. Sometimes.
Speaker 1 (30:11):
This is just like wild card.
Speaker 3 (30:13):
I like it to help me be creative and get
out of my own thought patterns sometimes. So this is
the one I've been using so much lately, and I
love it. I say, give me one truly deep novel
out of distribution, mind blowing insight related to topic that
only a few or none of us are aware of.
It can't be generic, vague, flashy or slop corporate stuff.
(30:37):
And it just gives you like things you would have
never thought of.
Speaker 5 (30:41):
Really like, it's impressive what you're getting back.
Speaker 3 (30:44):
It's learned from all of human insight and intelligence. Yeah,
you've just lived in your own brain for however many years,
so like it just breaks you out of your own
mind chamber.
Speaker 5 (30:56):
Oh okay, well I've got to go because I've got
to get on chat JBT.
Speaker 3 (31:01):
With that, we will close out. I will just say,
if you want to learn more about AIM, this company
invent Him has a really good course. So she does
both in person and also self paced, and so we'll
put a link to that in the show notes. And
also just her podcast is my favorite podcast. It is
very worky, but obviously I'm obsessed with work. Her podcast
is called how We Work and I've listened to every
single episode. Love it.
Speaker 5 (31:22):
Well, thanks so much for leading that conversation.
Speaker 3 (31:24):
It was amazing. Oh thank you.
Speaker 2 (31:33):
So I thought I was someone who used AI a lot,
but my mind is blown by how much I could
actually be getting out of it. The key takeaway from
today's episode is understanding AI's three distinct roles as a
delegate for basic tasks, as a collaborative partner for brainstorming
and feedback, and as an expert for specialized knowledge. Make
(31:54):
sure you check out this week's bus newsletter, where we'll
be sharing all those juicy prompts that Amantha mentioned, from
therapy ones to those presentation helpers, and yes, I'll be
trying that mind blowing insight prompt. And immediately after this,
you can also head to our Instagram at biz by
Mama Mia, where we'll be sharing more AI hacks and
tools throughout the week. I will see you in Thursday's
(32:14):
Inbox episode, where we'll be answering all of your career dilemmas. Bye,
Mamma Mia acknowledges the traditional owners of land and waters
that this podcast is recorded on
Popular Podcasts
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Therapy Gecko
An unlicensed lizard psychologist travels the universe talking to strangers about absolutely nothing. TO CALL THE GECKO: follow me on https://www.twitch.tv/lyleforever to get a notification for when I am taking calls. I am usually live Mondays, Wednesdays, and Fridays but lately a lot of other times too. I am a gecko.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.