How Small Businesses Get Real Value From AI, with Eric Bye

How this transcript was made

Transcribed by machine from the published audio, then corrected by hand. It is not a verbatim record: the changes below were made, and are counted.

This episode

78 speaker turns, 121 changes.

• closing CTA identified as CTA 1 - Share (match 100%), replaced with the authored script and attributed to Max

What was changed in these transcripts, and what was not

Generated: 2026-09-20 from 2026-09-20-named/, itself transcribed from the audio that is live on the feed. Every change below is mechanical and counted. No sentence was reworded, no claim altered, nothing removed for being awkward.

The editorial decision

Owner's choice, 2026-09-20: light clean. "Um", "uh" and stammered word repeats ("and, and") are removed, because nobody intends to say them. "You know", "I mean", "kind of" and "sort of" are KEPT - they run to roughly 2,300 instances across the catalogue and stripping them would make the page read as somebody else.

Counts across the whole set

Change	Count
stammered repeat removed	667
um / uh removed	654
space before punctuation	241
split contraction	132
guest turn named (Joe Casabona)	101
organization -> organisation	71
diarisation fragment returned to the host	62
ai -> AI	57
empty turn dropped (was filler only)	36
synthetic outro restored from script	21
organizations -> organisations	18
hyphen followed by space	14
realize -> realise	14
A.I. -> AI	10
recognized -> recognised	9
gray -> grey	9
summarize -> summarise	9
color -> colour	9
behavior -> behaviour	8
synthetic intro restored from script	8
analyze -> analyse	7
behavioral -> behavioural	7
recognize -> recognise	6
percent -> %	5
personalized -> personalised	5
skeptical -> sceptical	5
analyzing -> analysing	5
center -> centre	5
optimization -> optimisation	4
realized -> realised	4
favorite -> favourite	4
realizing -> realising	4
canceled -> cancelled	4
recognizing -> recognising	4
announcer turn relabelled from Guest	4
summarizing -> summarising	4
labor -> labour	4
behaviors -> behaviours	4
colors -> colours	4
organizational -> organisational	4
customized -> customised	3
analog -> analogue	3
optimize -> optimise	3
honor -> honour	2
optimizing -> optimising	2
summarizes -> summarises	2
personalize -> personalise	1
prioritize -> prioritise	1
prioritized -> prioritised	1
organize -> organise	1
rumor -> rumour	1
organized -> organised	1
capitalize -> capitalise	1
centered -> centred	1
organizes -> organises	1
emphasize -> emphasise	1
split name repaired (Schaefer)	1
humor -> humour	1
traveling -> travelling	1
defense -> defence	1
maximize -> maximise	1
Organizations -> Organisations	1
apologize -> apologise	1
digitized -> digitised	1
monetize -> monetise	1
analyzed -> analysed	1
narrator opener re-attributed to Max, words left as transcribed	1
theater -> theatre	1
specialized -> specialised	1
split name repaired (Wadsworth)	1
aluminum -> aluminium	1
specializes -> specialises	1
skeptic -> sceptic	1
customize -> customise	1
personalization -> personalisation	1
optimized -> optimised	1
visualize -> visualise	1
spoken domain -> humanspark.ai	0

What was deliberately left alone

• "Program". A computer program keeps that spelling in UK English; only a broadcast or a scheme is a "programme", and only a reader can tell which was meant.
• "License" and "practice". Both have a noun/verb spelling split in UK English that a rule cannot decide.
• Capitalised words mid-sentence that would otherwise have been respelled: 4 of them, most often part of a name. Left as spoken.
• Retired show and brand names. Spoken by the host and guests. A transcript that removes what was said is not a transcript.
• "Co-pilot". Sometimes Microsoft's product, sometimes the metaphor. A rule would corrupt the honest ones.

The synthetic segments

The opening voiceover on the archive episodes and the closing call to action are read by Max, the show's AI narrator, from a script written in advance. Transcribing them means guessing at words already on file, and the guesses were wrong in places - one opener came back as "an AI that boned him up" where the script says "phoned him up". Those passages carry the authored text instead, and are attributed to Max rather than to the guest: the diariser groups Max with the non-host speaker, which would otherwise put a real person's name on words they never said.

Alastair McDermottIf we want our businesses to thrive in this new era, we're going to have to get on board and figure out how to use AI,

and we're going to have to figure it out at a very practical level, and this isn't just applying to, you know, funded tech companies in San Francisco and Silicon Valley,

this is every business everywhere is going to have to figure out AI, so that's what we're going to be talking about today with,

I have a super guest on to talk about some really practical strategies and approaches to this, so we're going to talk about that. And today my guest is Eric Bye. Eric, you're very welcome.

Eric ByeThanks, Alastair. Yeah, thanks for having me. I'm forward to the chat.

Alastair McDermottSo, we were talking beforehand and in our pre-chat about some really practical things that we can do to approach this,

and one of the things that you talked about, I thought was a really great way to kind of frame it and to think about it,

so you talked about roles and workflow and tasks, can you tell me a little bit about that, explain the concept and kind of, because that will help to frame the rest of our conversation.

Eric ByeYeah, definitely. So, I mean, one of the concerns about AI is replacing jobs, but one of the common ways to look at that is it's not going to replace jobs, it's going to transform them.

So, when you think about a job, you can break that job into a number of workflows and each workflow into a number of tasks.

AI can support, augment, assist, or automate aspects of tasks or workflows, but from a role perspective, you're not going to have AI coming in and just replacing that role.

So, when it comes to understanding where AI is going to fit into your business, it's a good place to start, because it helps you break things down into kind of manageable chunks.

Alastair McDermottSo, how do we identify, like, when you're talking with a business that needs to implement this, like, where do you start? How do you start to figure that out?

Eric ByeThere's two different angles that I tend to start.

One is if I'm doing an assessment with a business, I have an AI opportunity assessment, helping them understand where it might fit into their business.

That's me going into them, starting talking to their kind of leadership within the business, and then speaking with and doing surveys of the various roles within the business,

understanding what are the common workflows that that role does? What are their kind of, what are their challenges? How is that workflow broken down into tasks?

Because I've worked with teams from all different parts of most companies I've been with in the past.

I've got a pretty good understanding of how those tasks break down, and then I can, based on my understanding of where their time is going,

what their challenges are, and what kind of the key workflows and tasks are, I can start to work with them to help them understand where AI can kind of insert into those workflows to either

improve effic iencies, but it's not just about efficiencies. It's also sometimes it can help increase the value that a workflow or task generates as an output.

Alastair McDermottHow does it do that? I'm interested in increasing the value specifically, because that's not something people talk a lot about.

Eric ByeYeah, so I guess you think about, you've got any task has an input, which might be information that just sits in someone's head,

or information from a customer, or information from a data set. And then it has an output, which could be content, it could be a decision, it could be a document or report for a customer.

So for example, I've got one client that has a number of consultants that work to generate reports, and they want to make the report writing process kind of more efficient,

and just faster in general using AI.

But one of the things that we looked at as well that I suggested was, well, do your customers, like, do you ever get feedback from your customers on the quality of their reports?

Do you ever personalise them for the type of, like, technical knowledge that they might have, or the role that they might have, or their objectives, or your commercial objectives?

So instead of just using AI, from an efficiency point of view there, we're looking at all the different ways that AI can improve the quality of the report,

the readability of the report, as well as weave in some of those commercial objectives.

These consultants aren't copywriters, no marketing team has time to go through and rewrite technical reports, and they often don't have the knowledge as well.

So it's a small example, but it's one clear one on kind of different ways that AI can actually improve the value that an output is providing to an end customer in that situation.

Yeah, that's interesting.

Alastair McDermottAnd so when I hear you talking about, you know, tasks and workflows and roles, one thing I know that, particularly in the SME,

the small business, small-medium business kind of sector, there's a lot of fluidity about roles and workflows

And people don't have standard operating procedures or SOPs quite often.

And having those does make life a lot easier when you are, you know, just when you're operating in general, but when you're using AI,

because you can start to look at those for opportunities. But for people who don't have that, like, is that the first thing you do is do you try and document what they're doing?

Eric ByeHow do you approach that? Yeah, definitely. I mean, it's a lot easier if people have their SOPs documented or their processes mapped.

But like you said, it's not overly common in the SME space.

So sometimes I'll do an initial mapping just so I make sure I understand that my assumptions are correct when it comes to feasibility and quoting.

But then as part of that engagement, I will go in and help document and map out those specific workflows.

Either working with their teams internally, because they might want to improve them anyways while they're at it, or just kind of going in myself and identifying all the different ways that

it fits in. Because often as well, sometimes process optimisation can get you further than AI as a first step. So I'm careful to make sure AI is not the silver bullet.

If there's a ton of different opportunities on the table, it's worth working those in as well at the same time.

Alastair McDermottSo then do you think about this whole thing as kind of process optimisation, which includes AI, but isn't necessarily like, is that like the overall offering that you're making here?

Eric ByeProbably not made it out loud before, but yeah, it's part of it.

Really, what we try and do rather than making it either like specifically process optimisation or digital transformation or just getting AI into the business,

it's defining what are the problems that they want to solve, or what are the objectives that they're looking to achieve,

and then looking at kind of a holistic approach for what's going to get them to those goals. Because I think a lot of the time, it's really easy to have a solution in search of a problem.

So if you just go in and say, where does AI fit into this? By definition, because they're trying to create something that's generally intelligent, it can fit in anywhere.

It's identifying where it should fit in, where it's going to actually add value.

Alastair McDermottHow do you prioritise then when you're looking at this? Like, if you're a business leader, and you're trying to figure out, okay, like, where do we even start?

We don't, like, the whole thing is unknown unknowns. We don't know what the capabilities are. It's a moving target anyway. It's continuously improving.

And we don't really know where to start with this.

Eric ByeLike, how do you deal with that? On one side, I think it's good to start with a broader understanding of the business.

That's one way that I like to work with businesses is going in and giving them kind of a snapshot. All right. Based on all your most common workflows for each of your roles.

Here are all the places that you can think about where AI might fit in.

I do a bit of try to quantify a bit on the amount of efficiency that AI could add at a task or workflow level, as well as the value add that it could add.

And then the complexity that getting that value or efficiency might have.

So that tends to give them a general picture of where AI fits in, kind of where those quick wins are, where the low-hanging fruit is,

or if they have different priorities because of specific challenges internally or objectives or problems that they're facing, then kind of focus on those places.

But that initial snapshot helps. Now, sometimes businesses prefer to kind of dip their toe in the water a little bit.

So they might have some specific challenges or workflows that they want looking at.

So sometimes, even if there might be other opportunities in the business, they want that proof of value first.

So it might be a more targeted approach to kind of look at an individual team or an individual workflow or a specific problem or objective and say,

all right, well, here are the ways that I can help or AI can help you achieve those specific things. And just ignoring kind of the rest of the business for that initial stage.

Alastair McDermottIf a business is interested in specifically quantifying the improvements, I assume you have to benchmark where their existing pre-AI setup is because you won't be able to give them a

quantifiable metric on here's how your efficiency approved or here's how productivity approved if you don't know what it was before you started using these tools, right?

Eric ByeYeah, definitely.

I don't usually have time to go in depth and do time and motion on everything, but people generally have a decent idea in their head when you get down to the people actually doing the tasks

or that on the workflows, how long different elements take.

So we try and get a relatively accurate estimate so that following the implementation of any sort of changes, we can then try to track how it's improved and where the challenges are.

Because often, especially if you don't have a standard operating procedure, the challenges that you're also introducing might be introducing best practice that might take a bit more time.

Alastair McDermottYeah, it's like you could be introducing a lot of different types of changes simultaneously and I could also imagine there being a bit of resistance to that.

Eric ByeYeah, I mean, on the change management side, it's definitely tricky.

Part of my background is doing enterprise digital transformation and you do need to make sure you're working closely with them. I try and make it as collaboratively as possible.

I'm doing workshops to introduce it and getting feedback because it's all about helping the business use AI to leverage their own expertise.

And I know that I'm not an expert in what they do for the most part. I'm an expert in helping them leverage AI to get the most out of it.

So to get the most out of it really does need to be a collaborative approach, which does help break down some of that initial resistance and skepticism.

Alastair McDermottI'm interested in what you're seeing when you talk to businesses and you talk to people working in different roles in those businesses.

I'm interested, are you seeing much uptake at the moment?

Because I know that Microsoft and LinkedIn had a survey where they surveyed 30, I think it was 31,000 people and they saw super high adoption, like around 75%.

But my feeling is that on the ground in a lot of businesses, there's a lot of businesses that aren't doing any of this

Eric ByeAt all.

Alastair McDermottSo I'm just wondering what your experience is on, you know, like, are you seeing people already using AI when you go in the door?

Eric ByeI'm seeing them already using AI, but they're doing it very informally.

They tend to be using personal or free accounts rather than it being company sponsored for the most part, which is a whole separate kind of risk.

And I've frequently seen a lot of people that have given it a few tests and found that it wasn't giving them what they needed or because it's a chat.

It's just starting with a blank page, not knowing where to start.

They do a few initial tests and don't find that it's adding a lot of value or it's unable to solve a certain problem or help them with a certain workflow.

So they kind of just give up on it a little bit. So part of a lot of my conversations are helping break through that skepticism because people have tested it.

They feel like they've found that it's not working for them. But a lot of that is just education and experimentation over a longer period of time.

I've found very little formal adoption within or plans within businesses.

But that's also the types of businesses that I talk to and that come to me are those people that need help understanding where to start.

Alastair McDermottYeah, and I think that it's a hard sell.

Like my experience has been that when I'm talking to people who've had those bad experiences, like they don't want to hear that basically that's a user or like that's a lack of education or

AI literacy. Like that's the real problem there. But usually that's what it is because they haven't really, you know, they haven't really looked.

They've started to experiment without trying to learn about it. And once you start to put real examples in front of them and show them how to use it, I found that that that adoption,

it's much easier to get them across the line when they can see real real world, you know, examples.

So I guess and then the other thing that you mentioned is people using personal accounts, like just for the audience, like the big risk there is is data privacy.

Like if you're putting in transcripts of calls with clients, like your client information in there, you know, probably violating GDPR if you're in the EU or you have EU clients.

And and then the other one is losing access to IP.

I think Samsung had a very famous case last year where they they put some some proprietary confidential company information into into ChatGPT and they couldn't get it removed.

So is that is that the kind of issues that you're seeing?

Eric ByeIs there anything I've missed there? No, that's that's exactly the kind of issues.

ChatGPT has recently released a new setting so you can turn off training, but you're still under those consumer terms.

You don't get the protections of the teams or enterprise plans or data security aspects of that.

And it's not just, I guess, yeah, intellectual property, GDPR, huge concerns, not just for your own business.

But if you work with a client businesses, it'll be interesting when the next big models start to come out when you if you can look up businesses and get a whole bunch of ideas on what their

internal strategies were, because their whole teams are using ChatGPT to write them on a personal account for the last two years. So that'll be interesting.

There's also big concerns on that side when it comes to like meeting, just meeting notes, apps.

So if you're having you're joining a meeting, someone has they signed up for example, Otter AI, which once you sign up, it auto invites itself to all of your meetings.

Everyone just approves it. If you're on a free plan, I don't think there's any any privacy protections with something like that.

So it's not only what you're using, but being aware of what people on the calls that you're on are using as well.

It gets it gets messy coming up with a policy that's simple and understandable for businesses and their employees, but is going to protect you.

Alastair McDermottI guess this is the really interesting thing about AI and figuring out the kind of lines of demarcation, because it's it's only truly valuable and useful when we give it access to all of

this information.

When when we seed some level of privacy and say, OK, I'm going to let you access all my conversations that I have at work, you know, that's when it becomes truly valuable in my experience.

I'm just wondering, do you have any thoughts about this?

Eric ByeYeah. Integrating that personal context is is huge. I think one way to think about it is if you ask an employee to create a presentation for, say, a pitch, there's a lot of there's I mean,

there's hundreds of thousands of words or tokens that are unsaid. When you ask them to create that, they understand your relationship with the client. They understand the history.

They're probably CC done a whole bunch of emails. They understood what's worked and what hasn't in the past.

People put those similar requests into an AI without any thought about what they would normally expect a person to know and then get disappointed with the results.

And that integration across your kind of your your work chat and your emails is it's coming.

It's not not necessarily here yet, but it's definitely going to bring with it a whole bunch of a whole bunch of risks, I think.

And it's not only for the individual employee, but if you think about people being able to access all the company information and from transcripts to calls that they weren't on,

maybe there's a lot of banter and other types of conversations outside of the data security side.

It's one of the reasons I think a lot of people found Apple's pitch last week or the week before so appealing when it comes to how they were handling as much AI on device as possible,

only sending what they have to to the cloud.

But yeah, giving you access to that personal context is is going to be huge and unlocking more and more value as it learns more and more about about you and your personal situation and your

relationships with different people and your objectives and projects.

Alastair McDermottYeah, I think that one of the one of the issues that we're thinking about and talking about here is you talk about on device,

you know, on device models where they're processing and it doesn't, you know, no, none of your data actually leaves the leaves the device, supposedly.

And then, you know, we're talking about enterprise plans for some of the larger models like Google's Gemini or ChatG

Eric ByePT.

Alastair McDermottAnd then there's the ability to run your own private instance, your own private server, which I think like some of the enterprise,

like some of the larger organisations will be doing that, having their own install of these open source models like Lama 3 or some of these.

I've actually installed some of those locally here on my own computer just to test them and play with them.

And it's interesting to see because you can give them, you can then put more of your own information and your own data

Eric ByeIn there.

Alastair McDermottAnd that gets interesting.

It's I can't see organisations, businesses doing that unless they're really at the enterprise level, though, because it's it's it's too it's too much to it's too much infrastructure to try

and maintain. What what are you seeing because I think like the small like the way I see it, that the solopreneurs and like tiny micro businesses,

they're going to go with the commercial personal accounts and just use that. And then the enterprise folks like the Fortune 500, they're going to be using these enterprise level plans.

But there's going to be a lot of businesses in the middle somewhere who don't know which way to go.

Like, should they be using the enterprise level and have their own private instance with that's trained up on their

Eric ByeModel?

Alastair McDermottHow do how do they get access to the true value of these systems, which is to have lots and lots of of your business data in the model itself?

Eric ByeI think it's all about the business's risk profile.

Each business has a different appetite for risk, a different level of sensitivity with the types of information that their business might kind of want to input into these models.

From, I mean, open AIs, like firms and security around their teams plan versus their enterprise plan are actually quite similar. And the team's plan isn't that much more expensive.

Claude has pretty strong commercial terms conditions as well. So a lot of it also comes down to trust.

How much do you actually trust the providers that they're going to be able to keep your data secure?

So it's almost like with any software procurement, if you're putting a whole bunch of your data in it, how much do you trust their ability to keep it secure?

One of the things that I do when I'm working with businesses, though, if that they are concerned about certain types of information,

is making sure that they have approval on using AI for certain workflows, that if they deal with particularly sensitive subjects or topics,

that they avoid putting that information in, avoid putting any PII in at all.

Don't just copy and paste whole email threads with your customers while they say they're trying to remove that in training. I mean, my last name is Bai.

So it would be difficult for them to identify that in all the training data and remove it effectively. So it's a case-by-case basis, really, on the sensitivity.

I think we're going to continue to see it evolve. Right now, it's not ingesting all of your information. It's only putting in what you want to put in right now.

But I think as we see Microsoft's co-pilot mature, as we see Google's Gemini enterprise plans mature, and they start to get a bit more predictive and kind of more proactively search out

information and context behind the scenes, so you're not necessarily aware of what's going into the prompt, then you'll probably start to see a lot more difficult decisions being made.

My prediction is, though, from a local model point of view, when you're wanting to work with the most performant models, we're going to keep seeing bigger and bigger models.

So we might have a whole bunch of smaller actions working locally.

But if you're doing kind of more challenging tasks that require a lot more reasoning, you're probably going to have to continue to go to the cloud.

So it's finding those providers that you trust that are able to help kind of achieve what you need to achieve.

Alastair McDermottAnd let's talk about the different providers then for a moment, because the way I see it at the moment, most businesses are

Well, I think the ChatGPT has certainly, like, they got the jump on everybody. So they're the most well-known.

And then we have, you know, Google with Gemini and Microsoft with Copilot and whatever, you know, and the other issue with these models is they

Eric ByeKeep renaming them,

Alastair McDermottWhich is really not helping because...

Eric ByeI still can't figure out which model of Google Gemini I have access to.

Alastair McDermottYeah. I pay for Google Gemini and I pay for ChatGPT. Those are the two paid ones that I have.

I also have access to Google's AI Studio, which is their next generation model, which is pretty cool because it's got a 1 million token

Eric ByeWindow.

Alastair McDermottSo that means that you can put in, like, the equivalent of three or four different business books worth of data, which is big because I have actually put in business books,

my own and that of a friend of mine, where we got it to help us with, you know, checking the content, grammar and structure and kind of developmental editing and things like that.

But these models, what I'm seeing at the moment is that, you know, if you use Office 365, then you're probably going to go with Microsoft because it's built into all of the stuff that you do

with your, you know, if that's what you use for email and Office productivity.

If you use G Suite or Google Workspace or whatever Google call that this week, if you use that Google system for email and everything,

then you're probably going to go with Gemini because that's just the default. I'm wondering, are you seeing people use stuff outside of what, you know, what the default would be?

Are you seeing people using different models in the same organisation?

Eric ByeYeah, I think at the moment, everyone just tends to be experimenting. I've seen a lot more limited testing of Gemini or kind of Microsoft suites.

Microsoft had a big pitch, but I've not heard a lot of positive reviews come out of it.

Google Gemini, I've got AI enterprise and my unpaid personal Google account has more AI functionality than that AI enterprise at the moment. So it's, everything's quite limited on that side.

The challenge is when you're an enterprise too, changing between suites is really expensive and making a long-term commitment.

So a co-pilot requires like a 12-month purchase up front for a license, which is a little bit obscene.

So we're starting to see more enterprise partnerships coming from like with OpenAI and their ChatGPT enterprise.

And I think I, my hypothesis, and I suspect this, just mostly based on vibes, is that we're going to see a lot more businesses kind of committing to those enterprise or those teams plans for

now, continuing to test it because they know not only that it's, we'll always be running the best model and have the newest features, but there's more control.

When you're using co-pilot, it does have an integration with your SharePoint that has your kind of active directory and your security rules set up.

How many businesses, especially SMEs, do you know, that are really confident in how that's set up right now and have the time to go and rework it and check it all to make sure it's going to

do what they want it to do. So it'll be really interesting to see how the next kind of six to 12 months plays out and how they start building AI into these kind of workplace platforms.

because I've been dying for some proper AI features in G Suite that I use at the moment. And I've almost never used anything that they've had to share yet.

Alastair McDermottYeah, I want to really, like, ultimately, for me, it's all going towards voice. And I want to be able to do that voice interaction, you

Eric ByeKnow,

Alastair McDermottFind me the last email conversation with Eric and give me a summary of it and then start to draft a reply to him. You know, that's what I want to be able to do with my email

I want to be able to do the same with my Asana projects. I want to be able to say, you know, what's the status of this project? Who are we waiting for?

Can you send a, you know, an update to somebody saying such and such a task is complete? I want the system to just be able to interact with me like

Eric ByeThat.

Alastair McDermottAnd I think that's where Apple is going a bit with what

Eric ByeThey're doing.

Alastair McDermottBut I think that even though they got such a, like, they were so far behind, I think that they've actually taken a step forward with,

you know, a step beyond maybe Microsoft and Google with the way that they've approached this.

Because for me, that's the way, like, it's all going to be about voice interaction and having those conversations, like, with our data, with some reasoning behind it.

That's how I can see a lot of this working. What about you? Like, what makes you, like, what kind of features do you really want to see from this?

Eric ByeI'm looking for automation. A lot of the stuff you can do with Zapier to say, when I receive an email with this in it, I want you to go do this.

And you have your CRM starting to plug into everything.

But if I'm having a back and forth with the client, I want to continue to do that in my email, but have it update my CRM based on the context of that email,

or the transcript from a call, as well as maybe update the project plan based on those discussions, or suggest an update to a project plan.

That sort of, kind of, doing those little admin tasks, that always take a whole bunch of time. So I don't have to, kind of, pivot between all these different systems.

Just let me live in one place and help me, kind of, keep all those other things up to date so I can get the most out of them,

rather than having to constantly, kind of, copy and paste and move things around. So we're seeing, kind of, the CRM and SaaS companies take a step ahead, build their integrations.

But I think we'll start, it was, a lot of that was in the co-pilot pitch. But we're not, we're not there yet.

So kind of being able to, whether it's text or asking it, look, every time I get an email with an invoice in it, put it, put it in zero, and kind of keep track of, keep track of all that.

And I think those basic, more, kind of, those basic actions that I can take on my behalf is going to give me a lot of value, personally.

And I think looking at a lot of businesses as well, there's, that's where you're going to see a huge unlock when it comes to,

okay, prior, like, we've got an inbox that receives all our leads, for example.

Before a salesperson looks at it'd be great if the AI went to that website, went to that business's website, gave a breakdown of everything they did,

kind of qualified and prioritised that lead, and then assigned it to the right person.

It's stuff that I can do now using manual automations, but building those workflows in an environment that is a bit more user-friendly.

I think that's where we're going to see some huge gains on that kind of new user experience going beyond talking to it and asking to it.

It's kind of the traditional digital transformation journey. Currently, it's really reactive. We're asking it to do something on our behalf, and it's going off and doing it.

I'm looking forward to the kind of the proactive stage, where it's, it knows kind of what I want to do next.

It knows some of the context, or it's able to at least pro actively manage things that I tell it to once I tell it once to do it.

And then we'll be going to the predictive, hopefully in the next one to two years, where it will say, look, here's a new project going on. It's not in the project management system yet.

Let's create a project, use the template, and crack on.

Alastair McDermottYeah, so one of the things that I think is going to be huge is those true automations. So where we can create an agent to do something for us.

And I've played around a lot with agents recently, and I find like they're still at that very early stage of where you have to kind of hand create everything.

Like I found myself installing Python, which is a programming language on my computer so that I could, I could create an agent.

And then I found myself going to a website and trying to find a CSS selector, which is like the identifier for a, for, for a label, to try and, to try and get this

Eric ByeThing to work.

Alastair McDermottSo it's still super clunky and only people who maybe have like a, coding or software engineering background, I think can really truly use these things properly at the moment.

But at some point, somebody is going to fix that. And then we're going to be able to say, Hey, for example, watch me upload my podcast audio and watch me do the whole workflow.

And once that's done, then create that as a workflow so that you can then next time, do that whole workflow of uploading my podcast and tit ling the episode and doing all that for me,

because that's something that I repeat for every podcast episode. It's an established workflow that I have.

I'm looking forward to be able to offload that step, cause it's, it's not quite there yet, but it's getting there, right?

Eric ByeYeah, we're getting there when you have kind of those multi -step, you just, it's really easy to experience a cascading failure, even at like a 95% if it's a, if it's a frequent task.

But I think one of the things that I'm really looking forward to being wrapped around that agent is the user interface and the user experience,

because you'll still have failures will still happen, but the ability to kind of, the steps it's going to take to jump in and say, Oh, that's, that's the wrong step you're going to take.

You want to do this action instead, or for it to say, Oh, it failed here, rather than having to rerun it and fix whatever it failed and then have it continue on.

I think those user experiences and user interfaces are going to get really interesting with how we think about and how we interact with these things on a day-to-day basis.

Will it be a lot more like keeping an eye on a few interns, versus kind of just waiting for a black box to do whatever it's supposed to and not being sure where it failed.

Alastair McDermottSo the, that technology of agents, which is basically giving, it's, it's giving AI power to do stuff on your device,

like on your browser, you know, go for example, for me, for the podcast, it's go navigate to the podcast website, log in,

and then click on the upload button and then go find the file on my, on my desktop and upload that MP3 file

Eric ByeThat audio file.

Alastair McDermottSo it's, it's actually able to interact with your device and do things like that's the next level that we're talking about,

apart from just being able to have a chat back and forth with us, like the, the large language models, the LLMs, like, like ChatGPT and things, but those agents right now are really

Eric ByeClunky to create.

Alastair McDermottSo that's what we're talking about here is it's just, it's, it's so difficult to set them

Eric ByeUp right now.

Alastair McDermottAnd I think that will change.

but bringing it back to technologies that are currently available, we're talking about using, you know, Gemini and ChatGPT and co-pilot and the capabilities of these existing systems are

incredible, what they can do

So like there are stuff that we can implement right now and, and are using, and I like, I think that when people are trained in their use and know how to use them properly,

suddenly they realise, Oh, the capabilities of this are far greater than I realised when I did my initial testing. And like, that's something that that bought you and I have

Eric ByeExperienced.

Alastair McDermottYeah.

Eric ByeMassively. I mean, as an entrepreneur, it's given me the ability to do multiple times what I would have, what I'm able to accomplish.

I'm just using that chat interface and all of the steps that a business needs to take to take advantage of AI in the future are, those aren't going to go away.

Understanding your workflows, understanding where it's going to fit.

There's a ton of value you can get today and all of that knowledge and all those skills around using the systems about understanding where it fits into your business.

Those will just help create a, give you your business, a foundation too.

So that as we get the next improvement and the next feature and the next improvement, you'll just be able to take advantage of them even faster rather than kind of waiting and then starting

from scratch three or six months from now, because it's not perfect or hallucinates here and there's, there's a ton of value you can get today as long as you're careful about managing those

risks and educating your team and giving them access to the safe and secure products that have your business and your legal department are comfortable with.

Because it's, there's just, yeah, there's tons of opportunity today.

And I think some people would like to wait until they see it as perfect, but it's just different technology that operates differently, which makes people feel a little bit uncomfortable.

Alastair McDermottAnd there were some tips that we were, we were kind of sharing some ideas on our last call.

And today, one of the ones that you suggested was for me, it was genius because one of the issues with using these systems is sometimes when they,

when they write for you, the writing is very clearly AI generated, like there's something about the language and you suggested that I tell it not to use adjectives.

And, and that's such a simple tip, but asking it not to use adjectives in writing. Like what's that going to do?

Eric ByeAlso just been massive for me. If you look at any of your writing, I went back and started looking at things I've been doing since before I started using that.

And you notice that it peppers in adjectives everywhere it can to try and make it sound, make itself sound smarter and more complex.

When you remove those adjectives and suddenly you get a really concise, really clear and easy to understand and far shorter description or explanation.

If I want some adjectives, I might just ask it to use it sparingly. but that seems to remove a lot of those kind of AI vibes you get from a lot of AI generated text.

You still, if you're doing something that you do all the time, you might want to use writing samples still and get it to copy that voice and that style and tone,

but asking it to not use adjectives. And I think it was it, was it Bezos or was it, Steve Jobs that told everyone to stop using adjectives in their business writing,

but that was, it's, yeah, it's massively transformed the quality of output I get, I get from these tools and it's made them right in a far more concise

Alastair McDermottWay. I love it. And I'm definitely going to do that.

are there any other piece of advice that you would have for somebody who is looking at their business and they're wondering where to start with this?

like where, what's the, where's the first place to go? What should they, what should

Eric ByeThey try first? Just try experimenting, get on a couple of the different tools, like, blog and chat to BT.

So you can see the differences, make sure you've got your privacy settings on them turn.

So you're not training the data and start experimenting, whether it's for work tasks, like writing an email or writing a document or getting a recipe once to give it my,

all the ingredients I had in my house and ask for a recipe on something, seeing where it fails. So in that situation, it got some of the proportions wrong, but it did really well.

And I was able to understand far better in that, in that context kind of, oh, it fails around this type of, this type of math. so the first thing is really just

Alastair McDermottExperimenting. like cooking and recipes is probably 30 to 40% of my entire AI usage right now. I use it all the time.

you know, whether that's like, how long should I cook this for at what temperature, or I have this thing and I'm, I want to make,

you know, can you, can you suggest like five Mexican recipes using these ingredients? or, you know, just lots of different kinds of suggestions for different ways to do things.

it's, I find it so useful for that. And I love cooking.

and, I have a lot of cookbooks, but I find this is really useful for that because it just, it gives you this specific answer that you're looking for.

So that's definitely a great kind of like test case, a great place to start. If you're not sure.

The other thing that I would suggest is if you're like me and you like talking and dictating, then install the app, like both Claude and, ChatGPT have apps that you can install. I'm sure.

I'm sure the others have them coming out or have them available now, but you can actually talk to it and being able to have a conversation or just dictating your prompts.

You can make them much longer. Like sometimes I'm dictating a prompt for three, four minutes.

And I'm, what I'm doing is I'm giving it lots of background, lots of context, because the more context you give it to something,

the more information it has to work with that, the better the output's going to be. So those are some things that I do as well.

Eric ByeYeah. Likewise. I mean, when I take my dog out for a walk, I'll often have a, just a, like an hour long conversation with ChatGPT.

And I'll say, look, I want to learn about this topic and become more of an expert in it.

If it's something I'm unfamiliar with, or I want to work through this business challenge, or I want to get feedback on this, and I'll just, yeah, just have a kind of going back and forth.

And by the end of the walk, I've been able to spend an hour talking to an expert on something that I would have had to troll,

like all the internet for days to get those answers in that type of way. So it's a, it's a great way to use the tool and understand kind of how it responds.

Alastair McDermottAwesome. Love it. Eric, where can people find you if they

Eric ByeWant to learn more? I've got a website, ericatronai.com, or I'm on LinkedIn. I do most of my content sharing on LinkedIn. You can find me, Eric Bye there's not too many of us.

and yeah, just feel free to reach out.

I do introduction to AI and discovery sessions for free, for an initial kind of get people understanding what it's about, why they should care and what they

Alastair McDermottCan do next. Awesome. I love it. And I will have links to your LinkedIn and your website in the show notes for the episode. Eric, thank you so much.

MaxThanks for listening to Something About AI. If you got something out of this one, share it with someone who would get something out of it too. That is genuinely how the show finds people.

There is more like this coming, so do subscribe wherever you are listening. And there is a pile of free resources over at humanspark.ai. Go and have a look.