July 14, 2025 • 24 mins
Want to save hours every week on content creation, client reporting and marketing strategy? In this episode of Digital Horizons, James Walker shares how he’s using AI to automate key areas of the business, from generating social media posts to building campaign insights at scale.
Joined by Brian Hastings, they walk through three real-world AI automation projects that are already up and running inside his business.
You’ll discover:
- How to automate content creation across YouTube, LinkedIn and Instagram
- A smarter way to deliver reports using Loom videos and AI-written summaries
- How AI can analyse your past campaigns, competitor data and brand content to recommend what to do next
These are not just ideas. They are working systems built using tools like n8n, ClickUp and custom AI agents. If you’re a business owner or marketer looking to implement AI into your workflow, this is a real-world look at what’s possible right now.
The Digital Horizons Podcast is hosted by:
James Walker - Managing Director WHD
Brian Hastings - Managing Director Nous
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:01):
I guess more powerful is what I'm finding, and maybe
it's not.
Maybe there's a whole bunch ofZapier that I just don't know
about, but this is an idealsocial schedule based on
performance over the past fewmonths, based on crawling all of
your competitive social mediapages.
We're looking at what'shappening in industry, trends
and everything else, and thenit's going to schedule and
publish it to all of our socialmedia.
Speaker 2 (00:20):
Welcome back to Digital Horizons.
This week it's it's aninternational edition.
We've got James in Thailand andI'm here in James's studio.
Welcome, james.
Speaker 1 (00:29):
Thank you.
I'm currently in Bangkok.
I've been here for, I think,just over a week now.
I've sort of set out on a bitof a one-way journey.
At the moment, I'm sure we'renot going to be returning back
home, but really wanting to usethis time while I'm away to
focus on building andimplementing processes within
our business and also, I guess,just be able to spend quite a
lot of time with my family aswell.
Speaker 2 (00:51):
Yeah, nice one.
It sounds amazing.
That feeling of freedom, of nothaving a return date.
It sounds very, very cool.
Speaker 1 (00:57):
I don't think I've ever done it before, so it's.
Yeah, it is a bit strange sortof feeling where you're an
open-ended trip and I mean itkind of feels like a holiday,
but it's also there's a lot ofwork happening as well while I'm
here, which is kind of fun itwould feel like you know the
guilt you get after being onholidays for a while.
Speaker 2 (01:12):
At least you're not going to get that guilty feeling
because you're still workingthat's right.
Speaker 1 (01:16):
I mean I can't.
I can't remember last time Ieven took a holiday when I
wasn't working.
So, um, it's pretty normal forme.
But it means I've got to thinkabout, well, how can I set up
and work and I make sure I'vegot a workspace and trying to
structure days as opposed tonormally on a holiday.
You're trying to fit workingamongst the other stuff.
This is more fitting the otherstuff amongst the work.
Speaker 2 (01:36):
So a bit of a change in the situation.
That's awesome.
We know consistency is key whenit comes to making content and
building an audience, so we'renot going to let you, being
overseas, get in the way ofrecording content for Digital
Horizon.
So we thought this week you'vemade leaps and bounds in your AI
efficiencies and AI automationjourney within your business.
(02:00):
Let's jump into it.
Do you want to take us througha bit of an update?
First, about what you've beenworking on, the one or two
scenarios that you're working onand what you've experienced,
what it feels like, whatchallenges you've come up
against and how the outputactually is working for the
business 100%.
Speaker 1 (02:21):
So I have hired a couple of AI engineers to help
us do the implementation,because I started trying to
understand and learn it myselfand whilst I was getting I guess
, getting time to do this, Ifelt that it was going to take
too long to actually come upwith a strategy, understand what
I'm trying to do and then alsotry to learn every single thing
that I need to implement it.
So I've hired two softwareengineers and we're currently
(02:43):
working on three differentprojects, which I'm super
excited about.
The first one, which is what Ithink I touched on last time, is
the automation that is goingthrough dropping a folder into a
Google Drive, dropping a fileinto a Google Drive folder, and
then it is then rolling out allthe assets that we need and even
uploading and posting thecontent for our other YouTube
(03:05):
channel, and this is what we'regoing to be rolling out for
Digital Horizons as well.
Speaker 2 (03:09):
I saw a sneak preview in one of your beachside
TikToks.
Yeah, it was awesome.
Speaker 1 (03:15):
I made a decision.
I was like, because one of thegirls, one of the engineers,
she's currently posting onTikTok about her experience
going through this as well andI'm like, oh, I'm going to do
this as well.
She's currently posting onTikTok about her experience
going through this as well, andI'm like, oh, I'm going to do
this as well.
She's posted five times a day,which I can't imagine doing that
volume, but I'm trying to beconsistent with it and I'm
enjoying doing it.
It's making me think aboutcreating content.
Speaker 2 (03:37):
It does make a bit of pressure to get progress before
the next TikTok, Otherwiseyou're like hey, James, here I'm
still at the same spot.
Speaker 1 (03:47):
I was four hours ago, that's right.
So what we've done is we'reusing NAN, which is, I think, if
you imagine, zapier, but Iguess more powerful, is what I'm
finding, and maybe it's not.
Maybe there's a whole bunch ofZapier that I just don't know
about, but what we've beenfinding is NAN's been the
perfect platform for what we'redoing.
Now I don't understand everysingle step of what's going on
in here, but basically whatwe're doing is we're taking the
(04:08):
recording and the video file ofthe podcast and we're uploading
it into a Google Drive.
So anyone who's listening alongthis is going to be a bit hard
to picture, but check us out onYouTube, but basically we've got
a whole bunch of steps that istaking the video, it's
downloading it, it'stranscribing it and it's getting
an understanding as to what theepisode's about.
From there, this is then goingto generate the content, so it's
(04:32):
using our AI agent, which isplugged into Claude, and it's
then analyzing the data.
It's creating the artwork, soit's understanding who's in the
episode, and we've got adatabase of all of our team
members' understanding who's inthe episode and we've got a
database of all of our teammembers' headshots, because
we're in the artwork for thewhat's Hot in Digital podcast.
We've got different teammembers' pictures and so I need
(04:53):
to understand who is the teammember that's in the episode and
also find and pull that pictureand then it creates the artwork
based on that.
And so then this area here isthat whole situation where it's
creating Google drive foldersand it's setting all up.
It's also writing the captions,so it's actually giving us the
name of the episode.
It's right in the wholedescription, and then it is
(05:14):
sending us a click up task,which then I have to read it and
approve it.
This section here is reallycool.
So this wasn't an area that wehad put into the initial plan,
but lydia had looked at it.
She's our AI engineer.
She looked at it and said, well, look, if you have changes to
it, it has to rerun the wholestep again.
So instead, what we've now gotit set up as within ClickUp in
(05:35):
the actual task, if I make editsto the text, it will then just
change it in there and then itpushes it through and doesn't
have to rerun the whole thing,which is super cool, saving a
lot of time there.
That's awesome.
And then it goes through andthen it's going to upload the
videos to YouTube, as you cansee on here.
So it needs to upload it toYouTube because we need the link
(05:56):
to the actual episode.
And then it's going to scheduleand publish it to all of our
social media.
So then it creates a LinkedInpost, it creates an Instagram
story and all that is thenlinked to the YouTube episode.
It's also pushing it toBuzzsprout and so that way it's
going into Spotify and AppleMusic and all the different
podcast places.
And then I get an email whichis like hey, this has all been
(06:16):
done.
Here is a link to the post,here's a link to your YouTube
episode, here's a link to theSpotify episode and everything
at the end of it.
Speaker 2 (06:28):
And this is all just from dropping a file into a
folder.
I mean, how fucking cool isthat?
So that's literally the trigger.
Speaker 1 (06:31):
The trigger is the new file being saved in there
that was what I originallyplanned it, but it was
originally.
Basically I am putting into afile and then I'm just filling
out a form.
So I have a form which isbasically, I insert the url to
the mp3, I insert the, the URLto the MP4, and that's it.
And then I submit the form andthen it just does all of these
steps.
Speaker 2 (06:50):
I'm really interested about the controls.
You know the proofing stages.
You get a click-up task withthe copy and the image for the
cover.
Is it in that click-up task oris it a link to a draft?
Speaker 1 (07:03):
video.
No, the copy is actually in thetask.
So I I open the task and it'sgot exactly.
It says linkedin caption thisis what it is.
Youtube caption this is thename of the title of the episode
.
Everything is written there,and then a link to the artwork
which is it's.
It's amazing, like it's just soeasy.
So I just yeah, a task pops up,probably about three, maybe two
(07:25):
to three minutes after I'vesubmitted that form with
everything done, and then it canbasically just hit approve and
then it will then schedule itout to the platforms if we're
happy with it.
Um, if I want to make anychanges, it'll just rerun it,
that little part of the copy,and then it will push it out to
the channels.
So it's um, it's mind-blowinghow well it worked.
Speaker 2 (07:41):
My brain immediately races to other processes in my
business that are way simplerthan that.
That would be amazing toautomate.
Speaker 1 (07:50):
I love that and I mean, this was our first one and
we've really set the bar highwith this, but I guess, when I
was thinking about it, it's likethis is basically a SaaS right.
If you start thinking aboutthese automations, yeah, could
become sellable things like youbuild this ones.
And then it could be like, hey,I could set up a quick website,
charge people whatever persubscription and then give them
(08:12):
access to this and the amount oftime like it's really valuable.
Yeah, all they need is the formthat's right, and then and and
connect it into the credentials,into all their different
youtube or whatever, whichshould be able to set up pretty
easily as a streamline process.
But I guess that's why thiswhole journey of I think
eventually we will be selling AIservices but we want to
understand from our side what'sreally valuable, and this here
is super valuable for us and Ithink anybody is creating
(08:35):
content.
So that's the first one, thatone super excited, that one's
currently working and we'reusing that one now.
As I said, we and we're usingthat one now, as I said, we're
going to be doing this forDigital Horizons in the near
future.
The next project we've got twocurrent other projects that
we're currently working on atthe moment.
One is around reporting.
So this one's to save time forour account managers.
So every month, our team gointo the reports and they'll
(08:57):
manually record a video goingover and talking about what's
going on for our clients areview on their report,
basically they and talking aboutwhat's going on for our clients
a review on their report,basically They've got a live
dashboard link which is done inMottograph and the team member
will be like all right, cool,hey, this is what's going on.
This is some challenges.
This is where we've had somegood wins.
This is what we're doingdifferent this year or this
month or whatever, and so it'sbasically just an overview
(09:20):
summary.
But when a client's doingmultiple services and they've
got multiple team membersworking on the account, each one
of those do on a differentvideo, so it's been quite a bit
of a time-consuming thing forour account managers to look at
it at the end of the month.
So they've got to go throughall of our client lists and
they've got to go pull the links, they've got to do all these
kind of things to review it andthen they've got to create a
summary to send it off toclients.
So basically, what we processway quicker.
(09:40):
So by the goal will be everyonejust adds their links and
everything into a Google sheetand then the automation will
then pull all that together,draft the emails and put it into
a format where our accountmanagers can just review it.
It's going to be written in atone of voice from our account
managers and then hit send andthen that will be sent off to
our clients, as opposed to amanual process and there's
another one that I'm going tosave hours every month for our
(10:01):
team.
Speaker 2 (10:01):
So when you say written, do they still record
the video?
Speaker 1 (10:05):
Yeah, so the team will still record the videos
because the thing that'simportant.
Obviously we don't want to takeout any human aspects out which
add a lot of value and I thinkthat currently those videos, our
clients love it.
It's a part where they'regetting a good understanding as
to what's going on, because alot of our clients don't really
I mean, they're not from adigital marketing background.
So by having that Loom videowhere it's got the person's face
(10:30):
in it and they're talking overthe report and running through
everything that's been going on,I find that's really powerful.
So we don't want to remove that.
But it's the actual sending andcollating of all this
information.
So the links to the writing asummary to the client and then
all of that, that part therewhen you're looking at, say,
(10:50):
over 100 clients.
Speaker 2 (10:51):
It's a very time consuming part, so we're trying
to work out how to make thatmore streamlined.
That's awesome, and knowingthat you've managed to do that
seemingly much more complexversion with the you know
building a YouTube caption and atitle and posting it to
multiple platforms this oneseems way more achievable.
As long as the files are beingsaved in the right place and
there's some consistency aroundhow everything is uploaded, I'm
(11:13):
now of the belief thatanything's possible.
Speaker 1 (11:16):
That's what I'm finding.
Every time we start pullingsomething together, I'm trying
to think of the wildest thingthat would be like hey, like
what do we think is completelyimpossible?
And let's actually put this asa challenge and how can we make
it not impossible?
And it kind of feels like magicthe way that it comes together.
I mean this, this thing thatwe're talking about with the
reporting thing, previously withzapier, 90 of it could have
(11:37):
been done with zapier, becauseit would just be looking at
fields in a spreadsheet, pullingthat information and then
pushing it into a next step.
So you'd have your trigger as atoggle, so you'd switch, I
guess, a change of field with adropdown and say, yep, this is
ready to send, and then itshould be able to pull all that
information, but it's not goingto be able to write the email,
and so that's what we needed todo.
Yes, what we're going to do ishave the transcripts download
(12:00):
and then of the team membersrunning through the video and
then understand the context ofthat and then writing a summary
that's put into the email anddrafting that email, because
just putting the links into anemail draft isn't going to be
powerful enough.
I think that it's that nextstep.
Yeah, understanding the contextand understanding the content
that's happening, which I thinkis going to allow for, hopefully
, our account managers,especially over time, as it gets
(12:21):
to understand the way that theywrite, because it'll be
reviewing previous emails.
It's just going to add anotherlayer to the efficiency of it.
Speaker 2 (12:29):
Yeah, well, there's the comfort level there that
you're not using a standardizedcaption that says, hey, really
excited to share our video, ourLoom video, of the great results
we got this month.
What if that client is goingthrough the worst quarter of
their lives and you're allpositive, whereas when you're
looking at the captions andsummarising it from what the
(12:52):
account manager has been sayingin the video, then it'll get the
tone right on.
Look, it's been a challengingmonth.
There's more challenges thanwins, but we're going to talk
through it in these Loom videosacross the multiple channels.
That's amazing.
Speaker 1 (13:06):
It's just, and also the highlights, and then a good
dot point.
Hey look, you know, you had awin here.
You had a win here.
This is what's changed month tomonth and like having that dot
point like summary of it so it'seasy to get the access
information, because the AI willalso be reviewing the report,
will also be reviewing thereport, and so we'll have not
just the transcript but it'llalso have the PDF report that
it's going to be able to analyzeas well.
Speaker 2 (13:26):
So yeah just that.
Speaker 1 (13:28):
Again, it seems like a simple thing in terms of all
right, cool, hey, we're sendingreports.
How can we upgrade this?
But it seems like I feel likethe end result is going to be a
massive upgrade in terms of whatwe're doing and a massive time
efficiency saver.
Speaker 2 (13:40):
I was just about to say, with these, the sort of
theme of these automations arethe AI element is empowering a
new step, which is some form ofcontent creation to enrich what,
like you said, Zapier couldhave previously done, which is a
order-based automation.
Now you can enrich that with abit of relevant content or a
(14:02):
graphic.
But also there's some actionpoints that I guess you could
have posted to YouTube beforewith Zapier.
You could have posted to thesocial channels before, but it's
that step of writing the stuff.
So it is the theme of whatyou're looking for all areas
where you've been limited inmaking an automation because
there need to be some sort ofhuman content creation input
into it.
Correct, I mean, if we couldhave, you've been limited in
(14:22):
making an automation becausethere need to be some sort of
human content creation inputinto it correct.
Speaker 1 (14:27):
I mean, if we we could have you know that if we
wanted to write the captionsaround the door at all exactly,
I could have pulled all thatinformation like from it.
But it's taking the humanelements out of it and doing it
at a level which is able toaccess so much previous data as
well.
And I think this is the reallypart important part of it all is
that when it's connecting intoall the data points and has all
(14:47):
that history and has the abilityand this is the next one which
I'm really excited I'll talkabout next but once it's able to
understand and pull frommultiple data points at the same
time, that as a human operator,it is potentially going to be
able to outperform far morebecause it has the ability to
analyze the data at a higherlevel than I think we would be
(15:09):
able to do on our own in theamount of time that you'd be
able to dedicate to a task likethis.
Speaker 2 (15:18):
Yeah, absolutely different data points, but also
the history of.
Well, when we saw the spikelast month, it was due to email
traffic or you know.
Just having everything at itsfingertips to be able to draw
insights or conclusions is goingto make it amazing.
Speaker 1 (15:33):
Yeah, and I guess that's really what builds onto
this next one that we're lookingat, and what I'm trying to
build here is, I guess, if youwere to think about, all right,
well, I want to have a resourcewhich think about all the except
for each of the differentclients that you work with.
You work with multiplebusinesses, even our own
business.
There are monthly reports goneback potentially years of data.
(15:53):
There are ongoing campaigns,campaign schedules, new
campaigns being launched everymonth or whatever.
There's organic social mediaposts consistently being posted
to all different channels, andthen all this is generating data
.
There's brand guidelines,there's client meetings with
transcriptions from notes.
There is feedback that's beengiven on the graphics and on
(16:14):
content that's been provided.
Imagine if you were able to haveall that information in a
format that's being pulled intoan agent and then you're able to
then communicate with thisagent and then it's able to
provide you with ideas andconcepts for upcoming campaigns
based on all that informationand all that data that is there,
and then being able to providehey look, this upcoming month we
(16:36):
need some ideas for somecampaigns.
So this upcoming month we needsome um concepts for social
media posts, and imagine it canthen access all of that data and
then provide you with hey look,this is an ideal social
schedule based on performanceover the past few months, based
on crawling all of yourcompetitive social media pages,
also looking at what's happeningin industry trends and
(16:58):
everything else, and it's ableto look at all of these things
and then provide an outline interms of height.
This is potentially a reallygood schedule that could come up
over the next 30 days.
So not completely eliminatinghumans and then coming up with
all the different ideas forposts, and then, because it's
like all right, well, this gotreally good engagement.
Last time we did something likethis, let's try something else
like this.
We then go, yep, cool, love allthose.
(17:19):
It then creates the contentbased on tone of voice, based on
all the different previousposts and everything like that,
and then allows our social mediateam to really then spend their
time on the strategy, and sothis is what we're going to be
implement.
This is the next thing I'mtrying to build, and then we're
also going to be able to talk toit and say, hey, look, well,
you know what has been our bestperforming post over the last
six months, what has been, likeyou know, been working, what it
(17:41):
hasn't, what should we be doingmore of when we're coming up to
planning out the next quarter ofcontent or planning out the
next quarter of campaigns, andso this is something that we're
sort of working through at themoment, and it's got to pull a
lot of information, so it's beena much slower one, but I think
by the next time we catch up,that we should have had a lot of
progress in building this,because it's, I think it, that
(18:10):
we don't want to just beuploading to gpt, right, like we
want to make sure that it's ina closed network, and because
there is going to be um, I guess, confidential information
within this that we would liketo be able to use, but we want
to also make sure that it's safeand secure, and so that's
another element and layer thatwe've got to take into
consideration here.
so I can't wait to get thisrolled out, I mean even just
from a proofing and auditingcheck.
(18:30):
Like you know, we're going tolaunch a new campaign.
I want every single piece ofcontent that is running through
an auditing process.
It's like all right, cool.
Look at all the meeting notes,look at all the past campaigns,
look at everything that we'vetalked about in the past month.
Does it adhere and is itfollowing everything that we
wanted to, before it even getsreviewed or even sent to a
client to review?
Just having even that step inthe process, I think it's going
(18:52):
to be really powerful too.
Speaker 2 (18:53):
Yeah, definitely, and I think when you get consistent
at capturing or recordingmeetings and having those notes
available for this to review andread, there's going to be fewer
and fewer moments for the AItool or the agent to be lacking
the relevant information to beable to judge or audit the
campaigns.
I think it's really extending on.
(19:16):
We've probably all got a tasteof what you can do in GPT, but
you get quite generic responseswithout this training or this
deeper level of data.
This is really empowering itwith everything the person
running the campaigns should beaware of, but bringing forward
all those learnings from lastyear or six months ago that they
(19:38):
fly out of my mind in thatmoment when I need to think
about it.
So, yeah, being able to trustthat it's dynamically created
strategies or recommendations,insights, I'd say, from the
specific client data, not just aprompt or a one-off generic
prompt yeah, I think it'll behugely powerful.
When I've tried this withprojects in ChatGPT, it feels
(20:01):
like there's only so much of thedocuments I upload it really
considers.
So if you can have some successin trying this in the tools
that you're using, it will be areally powerful next step and
then you can start to apply itto just about anything.
It's not just automating someof the painful tasks, it's
actually your co-thinkingstrategic partner for each
(20:22):
client.
It's like your brain for theagency.
Speaker 1 (20:25):
That's it, and I guess what we're not trying to
take the human element out of it.
We're trying to amplify what wecan do and I think that's the
most important part of it isthat, whilst this can do a lot
of the I guess, the research anddifferent things and data
analysis we still want toempower our team to be creative
(20:45):
thinkers and really use this asa tool to elevate what they're
able to produce for our clients,and not to replace at all.
It's more about a way toelevate.
Speaker 2 (20:56):
Well, absolutely, I think the volume of good work
that can be produced and theteam then become critical of the
work it's producing and makingsure it's the best it can be.
You'll be able to get out morerelevant insights to clients and
actual data-backed campaignideas than you have before.
(21:16):
This one feels like a biggerbeast of a job, but Even as an
agency owner, having thelearnings remembered in one
place so that if a new teammember joins and has to be
brought up to speed on a clientthey can ask it questions.
They don't have to bothersomeone else with.
When did that last campaign run?
Why don't we run one forplumbing but we do for
(21:38):
electrical?
What are the typical budgets wespent on this?
All that stuff can be at theirfingertips and that that's it.
Speaker 1 (21:45):
It's, it's going to be data and I mean, and by
incorporating a voice agent intoit, they can.
They can sit and have aconversation with it.
You can have it like with yoursexy voice um, one that you
tried out there and conversate.
Yeah, what's the best campaignis that we could be running for
you know, um, for the client.
Speaker 2 (22:01):
So, um, I don't know, I'm so excited by it.
Don't call it no, that's great,but yeah, it's not my, my chat
agent.
With this one, we testedtogether, that's right.
Just to recap, you've taken usthrough three of your initial
projects.
The first one is definitelyclose to completion, the second
(22:22):
one is underway and the thirdone is your big, hairy,
audacious goal, the first onebeing an automated workflow that
grabs a video asset and theaudio asset from a podcast and
creates the relevant captionsfor YouTube and social channels
to create posts.
It also generates a cover imageusing the photography of the
(22:44):
team members that you've savedfor it.
It creates all of this readyfor you to review in ClickUp.
Upon your changes.
It makes updates or approval.
It then goes forward and postsit and then gives you
confirmation that everything'sbeen posted.
Streamlining multiple steps anda couple of hours of work for
someone.
Streamlining multiple steps anda couple of hours of work for
(23:08):
someone.
The second one was streamliningof the collation of the Loom
videos that you use forreporting and also the links to
those reports, combining thoseassets into an email and writing
a contextual summary of whatthe reports include and in the
relevant tone of the accountmanager, and then sending it on
to the client for 100 clients.
(23:29):
And then the third one is athought partner for your
strategist to understandeverything about each individual
client, what they've run intheir campaigns in the past,
their competitor performance,and make suggestions, insights
or supporting recommendationsfor posts, content, strategies
to move forward into the nextmonth and work with your
(23:51):
internal team members to provideinsights back to your clients
and recommendations of what todo next, but also even write the
stuff for them to review Prettybig stuff.
I can't wait to see whathappens and I'll probably be
buying off you some of theseworkflows.
Speaker 1 (24:09):
Yeah, and that's it.
I mean, that's stage two, isn'tit?
It's like we're going to rollit out, make sure it works, make
sure it's valuable for us, andthen we'll work out well, how
can we then make some money outof it, because obviously a lot
of investment going into makingthese things.
So then we'll look at how wecan help them monetize this as a
service.
I guess more powerful is what I'm finding, and maybe
it's not.
Maybe there's a whole bunch ofZapier that I just don't know
about, but this is an idealsocial schedule based on
performance over the past fewmonths, based on crawling all of
your competitive social mediapages.
We're looking at what'shappening in industry, trends
and everything else, and thenit's going to schedule and
publish it to all of our socialmedia.
Speaker 2 (00:20):
Welcome back to Digital Horizons.
This week it's it's aninternational edition.
We've got James in Thailand andI'm here in James's studio.
Welcome, james.
Speaker 1 (00:29):
Thank you.
I'm currently in Bangkok.
I've been here for, I think,just over a week now.
I've sort of set out on a bitof a one-way journey.
At the moment, I'm sure we'renot going to be returning back
home, but really wanting to usethis time while I'm away to
focus on building andimplementing processes within
our business and also, I guess,just be able to spend quite a
lot of time with my family aswell.
Speaker 2 (00:51):
Yeah, nice one.
It sounds amazing.
That feeling of freedom, of nothaving a return date.
It sounds very, very cool.
Speaker 1 (00:57):
I don't think I've ever done it before, so it's.
Yeah, it is a bit strange sortof feeling where you're an
open-ended trip and I mean itkind of feels like a holiday,
but it's also there's a lot ofwork happening as well while I'm
here, which is kind of fun itwould feel like you know the
guilt you get after being onholidays for a while.
Speaker 2 (01:12):
At least you're not going to get that guilty feeling
because you're still workingthat's right.
Speaker 1 (01:16):
I mean I can't.
I can't remember last time Ieven took a holiday when I
wasn't working.
So, um, it's pretty normal forme.
But it means I've got to thinkabout, well, how can I set up
and work and I make sure I'vegot a workspace and trying to
structure days as opposed tonormally on a holiday.
You're trying to fit workingamongst the other stuff.
This is more fitting the otherstuff amongst the work.
Speaker 2 (01:36):
So a bit of a change in the situation.
That's awesome.
We know consistency is key whenit comes to making content and
building an audience, so we'renot going to let you, being
overseas, get in the way ofrecording content for Digital
Horizon.
So we thought this week you'vemade leaps and bounds in your AI
efficiencies and AI automationjourney within your business.
(02:00):
Let's jump into it.
Do you want to take us througha bit of an update?
First, about what you've beenworking on, the one or two
scenarios that you're working onand what you've experienced,
what it feels like, whatchallenges you've come up
against and how the outputactually is working for the
business 100%.
Speaker 1 (02:21):
So I have hired a couple of AI engineers to help
us do the implementation,because I started trying to
understand and learn it myselfand whilst I was getting I guess
, getting time to do this, Ifelt that it was going to take
too long to actually come upwith a strategy, understand what
I'm trying to do and then alsotry to learn every single thing
that I need to implement it.
So I've hired two softwareengineers and we're currently
(02:43):
working on three differentprojects, which I'm super
excited about.
The first one, which is what Ithink I touched on last time, is
the automation that is goingthrough dropping a folder into a
Google Drive, dropping a fileinto a Google Drive folder, and
then it is then rolling out allthe assets that we need and even
uploading and posting thecontent for our other YouTube
(03:05):
channel, and this is what we'regoing to be rolling out for
Digital Horizons as well.
Speaker 2 (03:09):
I saw a sneak preview in one of your beachside
TikToks.
Yeah, it was awesome.
Speaker 1 (03:15):
I made a decision.
I was like, because one of thegirls, one of the engineers,
she's currently posting onTikTok about her experience
going through this as well andI'm like, oh, I'm going to do
this as well.
She's currently posting onTikTok about her experience
going through this as well, andI'm like, oh, I'm going to do
this as well.
She's posted five times a day,which I can't imagine doing that
volume, but I'm trying to beconsistent with it and I'm
enjoying doing it.
It's making me think aboutcreating content.
Speaker 2 (03:37):
It does make a bit of pressure to get progress before
the next TikTok, Otherwiseyou're like hey, James, here I'm
still at the same spot.
Speaker 1 (03:47):
I was four hours ago, that's right.
So what we've done is we'reusing NAN, which is, I think, if
you imagine, zapier, but Iguess more powerful, is what I'm
finding, and maybe it's not.
Maybe there's a whole bunch ofZapier that I just don't know
about, but what we've beenfinding is NAN's been the
perfect platform for what we'redoing.
Now I don't understand everysingle step of what's going on
in here, but basically whatwe're doing is we're taking the
(04:08):
recording and the video file ofthe podcast and we're uploading
it into a Google Drive.
So anyone who's listening alongthis is going to be a bit hard
to picture, but check us out onYouTube, but basically we've got
a whole bunch of steps that istaking the video, it's
downloading it, it'stranscribing it and it's getting
an understanding as to what theepisode's about.
From there, this is then goingto generate the content, so it's
(04:32):
using our AI agent, which isplugged into Claude, and it's
then analyzing the data.
It's creating the artwork, soit's understanding who's in the
episode, and we've got adatabase of all of our team
members' understanding who's inthe episode and we've got a
database of all of our teammembers' headshots, because
we're in the artwork for thewhat's Hot in Digital podcast.
We've got different teammembers' pictures and so I need
(04:53):
to understand who is the teammember that's in the episode and
also find and pull that pictureand then it creates the artwork
based on that.
And so then this area here isthat whole situation where it's
creating Google drive foldersand it's setting all up.
It's also writing the captions,so it's actually giving us the
name of the episode.
It's right in the wholedescription, and then it is
(05:14):
sending us a click up task,which then I have to read it and
approve it.
This section here is reallycool.
So this wasn't an area that wehad put into the initial plan,
but lydia had looked at it.
She's our AI engineer.
She looked at it and said, well, look, if you have changes to
it, it has to rerun the wholestep again.
So instead, what we've now gotit set up as within ClickUp in
(05:35):
the actual task, if I make editsto the text, it will then just
change it in there and then itpushes it through and doesn't
have to rerun the whole thing,which is super cool, saving a
lot of time there.
That's awesome.
And then it goes through andthen it's going to upload the
videos to YouTube, as you cansee on here.
So it needs to upload it toYouTube because we need the link
(05:56):
to the actual episode.
And then it's going to scheduleand publish it to all of our
social media.
So then it creates a LinkedInpost, it creates an Instagram
story and all that is thenlinked to the YouTube episode.
It's also pushing it toBuzzsprout and so that way it's
going into Spotify and AppleMusic and all the different
podcast places.
And then I get an email whichis like hey, this has all been
(06:16):
done.
Here is a link to the post,here's a link to your YouTube
episode, here's a link to theSpotify episode and everything
at the end of it.
Speaker 2 (06:28):
And this is all just from dropping a file into a
folder.
I mean, how fucking cool isthat?
So that's literally the trigger.
Speaker 1 (06:31):
The trigger is the new file being saved in there
that was what I originallyplanned it, but it was
originally.
Basically I am putting into afile and then I'm just filling
out a form.
So I have a form which isbasically, I insert the url to
the mp3, I insert the, the URLto the MP4, and that's it.
And then I submit the form andthen it just does all of these
steps.
Speaker 2 (06:50):
I'm really interested about the controls.
You know the proofing stages.
You get a click-up task withthe copy and the image for the
cover.
Is it in that click-up task oris it a link to a draft?
Speaker 1 (07:03):
video.
No, the copy is actually in thetask.
So I I open the task and it'sgot exactly.
It says linkedin caption thisis what it is.
Youtube caption this is thename of the title of the episode
.
Everything is written there,and then a link to the artwork
which is it's.
It's amazing, like it's just soeasy.
So I just yeah, a task pops up,probably about three, maybe two
(07:25):
to three minutes after I'vesubmitted that form with
everything done, and then it canbasically just hit approve and
then it will then schedule itout to the platforms if we're
happy with it.
Um, if I want to make anychanges, it'll just rerun it,
that little part of the copy,and then it will push it out to
the channels.
So it's um, it's mind-blowinghow well it worked.
Speaker 2 (07:41):
My brain immediately races to other processes in my
business that are way simplerthan that.
That would be amazing toautomate.
Speaker 1 (07:50):
I love that and I mean, this was our first one and
we've really set the bar highwith this, but I guess, when I
was thinking about it, it's likethis is basically a SaaS right.
If you start thinking aboutthese automations, yeah, could
become sellable things like youbuild this ones.
And then it could be like, hey,I could set up a quick website,
charge people whatever persubscription and then give them
(08:12):
access to this and the amount oftime like it's really valuable.
Yeah, all they need is the formthat's right, and then and and
connect it into the credentials,into all their different
youtube or whatever, whichshould be able to set up pretty
easily as a streamline process.
But I guess that's why thiswhole journey of I think
eventually we will be selling AIservices but we want to
understand from our side what'sreally valuable, and this here
is super valuable for us and Ithink anybody is creating
(08:35):
content.
So that's the first one, thatone super excited, that one's
currently working and we'reusing that one now.
As I said, we and we're usingthat one now, as I said, we're
going to be doing this forDigital Horizons in the near
future.
The next project we've got twocurrent other projects that
we're currently working on atthe moment.
One is around reporting.
So this one's to save time forour account managers.
So every month, our team gointo the reports and they'll
(08:57):
manually record a video goingover and talking about what's
going on for our clients areview on their report,
basically they and talking aboutwhat's going on for our clients
a review on their report,basically They've got a live
dashboard link which is done inMottograph and the team member
will be like all right, cool,hey, this is what's going on.
This is some challenges.
This is where we've had somegood wins.
This is what we're doingdifferent this year or this
month or whatever, and so it'sbasically just an overview
(09:20):
summary.
But when a client's doingmultiple services and they've
got multiple team membersworking on the account, each one
of those do on a differentvideo, so it's been quite a bit
of a time-consuming thing forour account managers to look at
it at the end of the month.
So they've got to go throughall of our client lists and
they've got to go pull the links, they've got to do all these
kind of things to review it andthen they've got to create a
summary to send it off toclients.
So basically, what we processway quicker.
(09:40):
So by the goal will be everyonejust adds their links and
everything into a Google sheetand then the automation will
then pull all that together,draft the emails and put it into
a format where our accountmanagers can just review it.
It's going to be written in atone of voice from our account
managers and then hit send andthen that will be sent off to
our clients, as opposed to amanual process and there's
another one that I'm going tosave hours every month for our
(10:01):
team.
Speaker 2 (10:01):
So when you say written, do they still record
the video?
Speaker 1 (10:05):
Yeah, so the team will still record the videos
because the thing that'simportant.
Obviously we don't want to takeout any human aspects out which
add a lot of value and I thinkthat currently those videos, our
clients love it.
It's a part where they'regetting a good understanding as
to what's going on, because alot of our clients don't really
I mean, they're not from adigital marketing background.
So by having that Loom videowhere it's got the person's face
(10:30):
in it and they're talking overthe report and running through
everything that's been going on,I find that's really powerful.
So we don't want to remove that.
But it's the actual sending andcollating of all this
information.
So the links to the writing asummary to the client and then
all of that, that part therewhen you're looking at, say,
(10:50):
over 100 clients.
Speaker 2 (10:51):
It's a very time consuming part, so we're trying
to work out how to make thatmore streamlined.
That's awesome, and knowingthat you've managed to do that
seemingly much more complexversion with the you know
building a YouTube caption and atitle and posting it to
multiple platforms this oneseems way more achievable.
As long as the files are beingsaved in the right place and
there's some consistency aroundhow everything is uploaded, I'm
(11:13):
now of the belief thatanything's possible.
Speaker 1 (11:16):
That's what I'm finding.
Every time we start pullingsomething together, I'm trying
to think of the wildest thingthat would be like hey, like
what do we think is completelyimpossible?
And let's actually put this asa challenge and how can we make
it not impossible?
And it kind of feels like magicthe way that it comes together.
I mean this, this thing thatwe're talking about with the
reporting thing, previously withzapier, 90 of it could have
(11:37):
been done with zapier, becauseit would just be looking at
fields in a spreadsheet, pullingthat information and then
pushing it into a next step.
So you'd have your trigger as atoggle, so you'd switch, I
guess, a change of field with adropdown and say, yep, this is
ready to send, and then itshould be able to pull all that
information, but it's not goingto be able to write the email,
and so that's what we needed todo.
Yes, what we're going to do ishave the transcripts download
(12:00):
and then of the team membersrunning through the video and
then understand the context ofthat and then writing a summary
that's put into the email anddrafting that email, because
just putting the links into anemail draft isn't going to be
powerful enough.
I think that it's that nextstep.
Yeah, understanding the contextand understanding the content
that's happening, which I thinkis going to allow for, hopefully
, our account managers,especially over time, as it gets
(12:21):
to understand the way that theywrite, because it'll be
reviewing previous emails.
It's just going to add anotherlayer to the efficiency of it.
Speaker 2 (12:29):
Yeah, well, there's the comfort level there that
you're not using a standardizedcaption that says, hey, really
excited to share our video, ourLoom video, of the great results
we got this month.
What if that client is goingthrough the worst quarter of
their lives and you're allpositive, whereas when you're
looking at the captions andsummarising it from what the
(12:52):
account manager has been sayingin the video, then it'll get the
tone right on.
Look, it's been a challengingmonth.
There's more challenges thanwins, but we're going to talk
through it in these Loom videosacross the multiple channels.
That's amazing.
Speaker 1 (13:06):
It's just, and also the highlights, and then a good
dot point.
Hey look, you know, you had awin here.
You had a win here.
This is what's changed month tomonth and like having that dot
point like summary of it so it'seasy to get the access
information, because the AI willalso be reviewing the report,
will also be reviewing thereport, and so we'll have not
just the transcript but it'llalso have the PDF report that
it's going to be able to analyzeas well.
Speaker 2 (13:26):
So yeah just that.
Speaker 1 (13:28):
Again, it seems like a simple thing in terms of all
right, cool, hey, we're sendingreports.
How can we upgrade this?
But it seems like I feel likethe end result is going to be a
massive upgrade in terms of whatwe're doing and a massive time
efficiency saver.
Speaker 2 (13:40):
I was just about to say, with these, the sort of
theme of these automations arethe AI element is empowering a
new step, which is some form ofcontent creation to enrich what,
like you said, Zapier couldhave previously done, which is a
order-based automation.
Now you can enrich that with abit of relevant content or a
(14:02):
graphic.
But also there's some actionpoints that I guess you could
have posted to YouTube beforewith Zapier.
You could have posted to thesocial channels before, but it's
that step of writing the stuff.
So it is the theme of whatyou're looking for all areas
where you've been limited inmaking an automation because
there need to be some sort ofhuman content creation input
into it.
Correct, I mean, if we couldhave, you've been limited in
(14:22):
making an automation becausethere need to be some sort of
human content creation inputinto it correct.
Speaker 1 (14:27):
I mean, if we we could have you know that if we
wanted to write the captionsaround the door at all exactly,
I could have pulled all thatinformation like from it.
But it's taking the humanelements out of it and doing it
at a level which is able toaccess so much previous data as
well.
And I think this is the reallypart important part of it all is
that when it's connecting intoall the data points and has all
(14:47):
that history and has the abilityand this is the next one which
I'm really excited I'll talkabout next but once it's able to
understand and pull frommultiple data points at the same
time, that as a human operator,it is potentially going to be
able to outperform far morebecause it has the ability to
analyze the data at a higherlevel than I think we would be
(15:09):
able to do on our own in theamount of time that you'd be
able to dedicate to a task likethis.
Speaker 2 (15:18):
Yeah, absolutely different data points, but also
the history of.
Well, when we saw the spikelast month, it was due to email
traffic or you know.
Just having everything at itsfingertips to be able to draw
insights or conclusions is goingto make it amazing.
Speaker 1 (15:33):
Yeah, and I guess that's really what builds onto
this next one that we're lookingat, and what I'm trying to
build here is, I guess, if youwere to think about, all right,
well, I want to have a resourcewhich think about all the except
for each of the differentclients that you work with.
You work with multiplebusinesses, even our own
business.
There are monthly reports goneback potentially years of data.
(15:53):
There are ongoing campaigns,campaign schedules, new
campaigns being launched everymonth or whatever.
There's organic social mediaposts consistently being posted
to all different channels, andthen all this is generating data
.
There's brand guidelines,there's client meetings with
transcriptions from notes.
There is feedback that's beengiven on the graphics and on
(16:14):
content that's been provided.
Imagine if you were able to haveall that information in a
format that's being pulled intoan agent and then you're able to
then communicate with thisagent and then it's able to
provide you with ideas andconcepts for upcoming campaigns
based on all that informationand all that data that is there,
and then being able to providehey look, this upcoming month we
(16:36):
need some ideas for somecampaigns.
So this upcoming month we needsome um concepts for social
media posts, and imagine it canthen access all of that data and
then provide you with hey look,this is an ideal social
schedule based on performanceover the past few months, based
on crawling all of yourcompetitive social media pages,
also looking at what's happeningin industry trends and
(16:58):
everything else, and it's ableto look at all of these things
and then provide an outline interms of height.
This is potentially a reallygood schedule that could come up
over the next 30 days.
So not completely eliminatinghumans and then coming up with
all the different ideas forposts, and then, because it's
like all right, well, this gotreally good engagement.
Last time we did something likethis, let's try something else
like this.
We then go, yep, cool, love allthose.
(17:19):
It then creates the contentbased on tone of voice, based on
all the different previousposts and everything like that,
and then allows our social mediateam to really then spend their
time on the strategy, and sothis is what we're going to be
implement.
This is the next thing I'mtrying to build, and then we're
also going to be able to talk toit and say, hey, look, well,
you know what has been our bestperforming post over the last
six months, what has been, likeyou know, been working, what it
(17:41):
hasn't, what should we be doingmore of when we're coming up to
planning out the next quarter ofcontent or planning out the
next quarter of campaigns, andso this is something that we're
sort of working through at themoment, and it's got to pull a
lot of information, so it's beena much slower one, but I think
by the next time we catch up,that we should have had a lot of
progress in building this,because it's, I think it, that
(18:10):
we don't want to just beuploading to gpt, right, like we
want to make sure that it's ina closed network, and because
there is going to be um, I guess, confidential information
within this that we would liketo be able to use, but we want
to also make sure that it's safeand secure, and so that's
another element and layer thatwe've got to take into
consideration here.
so I can't wait to get thisrolled out, I mean even just
from a proofing and auditingcheck.
(18:30):
Like you know, we're going tolaunch a new campaign.
I want every single piece ofcontent that is running through
an auditing process.
It's like all right, cool.
Look at all the meeting notes,look at all the past campaigns,
look at everything that we'vetalked about in the past month.
Does it adhere and is itfollowing everything that we
wanted to, before it even getsreviewed or even sent to a
client to review?
Just having even that step inthe process, I think it's going
(18:52):
to be really powerful too.
Speaker 2 (18:53):
Yeah, definitely, and I think when you get consistent
at capturing or recordingmeetings and having those notes
available for this to review andread, there's going to be fewer
and fewer moments for the AItool or the agent to be lacking
the relevant information to beable to judge or audit the
campaigns.
I think it's really extending on.
(19:16):
We've probably all got a tasteof what you can do in GPT, but
you get quite generic responseswithout this training or this
deeper level of data.
This is really empowering itwith everything the person
running the campaigns should beaware of, but bringing forward
all those learnings from lastyear or six months ago that they
(19:38):
fly out of my mind in thatmoment when I need to think
about it.
So, yeah, being able to trustthat it's dynamically created
strategies or recommendations,insights, I'd say, from the
specific client data, not just aprompt or a one-off generic
prompt yeah, I think it'll behugely powerful.
When I've tried this withprojects in ChatGPT, it feels
(20:01):
like there's only so much of thedocuments I upload it really
considers.
So if you can have some successin trying this in the tools
that you're using, it will be areally powerful next step and
then you can start to apply itto just about anything.
It's not just automating someof the painful tasks, it's
actually your co-thinkingstrategic partner for each
(20:22):
client.
It's like your brain for theagency.
Speaker 1 (20:25):
That's it, and I guess what we're not trying to
take the human element out of it.
We're trying to amplify what wecan do and I think that's the
most important part of it isthat, whilst this can do a lot
of the I guess, the research anddifferent things and data
analysis we still want toempower our team to be creative
(20:45):
thinkers and really use this asa tool to elevate what they're
able to produce for our clients,and not to replace at all.
It's more about a way toelevate.
Speaker 2 (20:56):
Well, absolutely, I think the volume of good work
that can be produced and theteam then become critical of the
work it's producing and makingsure it's the best it can be.
You'll be able to get out morerelevant insights to clients and
actual data-backed campaignideas than you have before.
(21:16):
This one feels like a biggerbeast of a job, but Even as an
agency owner, having thelearnings remembered in one
place so that if a new teammember joins and has to be
brought up to speed on a clientthey can ask it questions.
They don't have to bothersomeone else with.
When did that last campaign run?
Why don't we run one forplumbing but we do for
(21:38):
electrical?
What are the typical budgets wespent on this?
All that stuff can be at theirfingertips and that that's it.
Speaker 1 (21:45):
It's, it's going to be data and I mean, and by
incorporating a voice agent intoit, they can.
They can sit and have aconversation with it.
You can have it like with yoursexy voice um, one that you
tried out there and conversate.
Yeah, what's the best campaignis that we could be running for
you know, um, for the client.
Speaker 2 (22:01):
So, um, I don't know, I'm so excited by it.
Don't call it no, that's great,but yeah, it's not my, my chat
agent.
With this one, we testedtogether, that's right.
Just to recap, you've taken usthrough three of your initial
projects.
The first one is definitelyclose to completion, the second
(22:22):
one is underway and the thirdone is your big, hairy,
audacious goal, the first onebeing an automated workflow that
grabs a video asset and theaudio asset from a podcast and
creates the relevant captionsfor YouTube and social channels
to create posts.
It also generates a cover imageusing the photography of the
(22:44):
team members that you've savedfor it.
It creates all of this readyfor you to review in ClickUp.
Upon your changes.
It makes updates or approval.
It then goes forward and postsit and then gives you
confirmation that everything'sbeen posted.
Streamlining multiple steps anda couple of hours of work for
someone.
Streamlining multiple steps anda couple of hours of work for
(23:08):
someone.
The second one was streamliningof the collation of the Loom
videos that you use forreporting and also the links to
those reports, combining thoseassets into an email and writing
a contextual summary of whatthe reports include and in the
relevant tone of the accountmanager, and then sending it on
to the client for 100 clients.
(23:29):
And then the third one is athought partner for your
strategist to understandeverything about each individual
client, what they've run intheir campaigns in the past,
their competitor performance,and make suggestions, insights
or supporting recommendationsfor posts, content, strategies
to move forward into the nextmonth and work with your
(23:51):
internal team members to provideinsights back to your clients
and recommendations of what todo next, but also even write the
stuff for them to review Prettybig stuff.
I can't wait to see whathappens and I'll probably be
buying off you some of theseworkflows.
Speaker 1 (24:09):
Yeah, and that's it.
I mean, that's stage two, isn'tit?
It's like we're going to rollit out, make sure it works, make
sure it's valuable for us, andthen we'll work out well, how
can we then make some money outof it, because obviously a lot
of investment going into makingthese things.
So then we'll look at how wecan help them monetize this as a
service.
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Special Summer Offer: Exclusively on Apple Podcasts, try our Dateline Premium subscription completely free for one month! With Dateline Premium, you get every episode ad-free plus exclusive bonus content.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.