Why AI Pilots Never Scale, with Patrick Ward

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

59 speaker turns, 72 changes.

• opening voiceover replaced with the authored script (match 98%) and attributed to Max
• closing CTA identified as CTA 3 - Resources (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.

MaxYou're listening to Something About AI. Patrick Ward spent years watching projects die after the technology had already worked.

His explanation is that connecting a blast furnace or a scanner to the cloud quietly changes the revenue model, the sales incentives, and who needs to be able to do what.

Nobody agreed to any of that at the start, so the project gets parked behind seventeen other priorities. He also found that the scarce thing was never data scientists.

Here's the conversation.

Alastair McDermottI'm going to talk about why successful AI implementation starts with business models, rather than with the technology.

And I'm looking forward to having this guest because he's got a fascinating background, super smart guy.

Patrick WardThat's coming next.

Alastair McDermottSo today's guest comes with 30 years of commercial experience, including 12 years at Microsoft. He transformed their global approach to IoT and AI implementation.

He's a lecturer at the UCD Smurfit Business School, and he special ised in digital transformation and AI. So Patrick Ward, you're very welcome to the show. Hey, Alastair. Good to see you.

So, yeah, there's so much I want to talk to you about, and I'm really interested to talk to somebody on the show who comes from that background of big tech and talking about how they see AI

and their approach to it. I think maybe the first thing that we could talk about is the business model versus technology thing, and I think that comes from your time at Microsoft, right?

Can you tell

Patrick WardMe a bit about that?

Alastair McDermottYeah, it does.

Patrick WardLook, as you say, I was 12 years at Microsoft. Seven of those years, I was with the IoT team.

And when I joined that team, I guess IoT became quite famous for or infamous for lots of proofs of concept, lots of trials, lots of pilots, but not a lot of scaling business.

And I could see that in an awful lot of the projects that our team was involved in. Even though the concept would get proven in the proof of concept, the project still didn't take off.

And that's not good for business. It's not good for anyone. It's not good for the client.

Alastair McDermottCan I ask you just for anybody who's a business owner listening to this, who's not familiar with IoT, can you just give a quick overview of what it is?

Patrick WardSure. IoT is Internet of Things. It's a term that was coined back in 1999 to, I guess, sign ify the next stage of evolution of the Internet.

But it's basically about the connectivity of all of the physical infrastructure that surrounds us in our lives and the management of that infrastructure and getting value from it.

So, you know, at home, managing things like turning on and off our heating system, our alarm system, cameras, all sorts of different things.

But probably, I think McKinsey say that 65% of IoT value arises in B2B.

And so that was our main focus in Microsoft was industrial robots, MRI scanners in hospitals, blast furnaces in steel mills,

milking machines, remote door locking, all sorts of different scenarios.

And it's basically about connecting those devices to the Internet, to the cloud, harvesting their information, using AI technology and other technology to remotely monitor and predict what's

going to happen with those devices. And so there's huge business value on it.

Alastair McDermottSo can you then link that up with why is it important to start with the business model rather than starting with the capabilities of the technology?

Patrick WardYeah, well, as I say, so a lot of these projects didn't really take off.

And when I considered the reasons for that, what became apparent was if you're making MRI scanners or blast furn aces or whatever those devices are,

HVAC systems for buildings, you know, heating, ventilation and air conditioning systems. If that's the business you've been in for 20, 30 years, you know that business really well.

But if you connect those devices and those systems to the cloud and you start harvesting data and offering data and software based services, your business model is fundamentally changing.

Your value proposition is going to change. You're offering a new suite of software services, probably as a subscription service. So you're into subscription for the first time.

So you're not just selling this equipment for $100,000. You're selling the equipment for $100,000 and then looking for a 5K per month or whatever it is. Revenue stream on subscription.

That's a totally different business model.

How we sell that, how we incentivize our sales team to sell that, how we incentivize our channel partners to sell that, all of that changes completely.

And so I think what was happening was we get to the end of the proof of concept and the organisation would realise the scale of change that was required to make this proposition successful

at scale out there in the market. And, oh, we've got 17 other different priorities right now. Now we'll park that up for a lot of things. And so the business model analysis came in.

I had a background in helping them define their business.

So I trained our team worldwide in Microsoft, the IOT partner team, to run a business model workshop at the outset of every engagement of every customer engagement,

to help the customer understand the huge value that there was in terms of defining the value proposition and so forth,

but also weight them up to the scale of change in terms of the kinds of things that I talked about in terms of incentivization,

in terms of the channel, in terms of organisational capability.

You know, you know, you know, an organisation who makes heating systems will tell you an awful lot about how the electrons pass through the tungsten and all the rest of it and heat values

and all of that. Well, do they know about software development? Do they know about DevOps? Do they know about data?

Do they know about cybersecurity user experience in developing software and all of that good stuff? Probably not. I used to say about IOT, and I say it now about AI more broadly.

It's kind of three fundamental facts. One, this is not easy. It's just not easy to get this stuff right. It's not easy from a business point of view.

It's not easy from a technical point of view. It is possible, but it's not easy.

The second point is you kind of got to do it because if you don't, others, other competitors or even competitors you've never even heard of will fly by you because they will take on this

challenge. They will get this right and potentially completely annih ilate. You can potentially completely disrupt the industry that you're operating in.

The third fundamental truth is for those players who get this stuff right.

There's huge value in it, new revenue streams, new ways of engaging my customer, a new channel to my customer that allows me to deliver services in the future that I haven't even thought of

today by virtue of the fact that I'm connecting this physical infrastructure to the cloud and delivering software services to my client.

And then it goes back to, but it's not easy, but you've got to do it. But, you know, if we get it right, there's a lot of value in it.

And my fundamental belief is, and I think we proved this at Microsoft, if we start with a business model led conversation and we get the custodians of the business model in the room

together, and that's generally the senior leadership team, or if it's a very big enterprise organisation, it might be a divisional leadership team or a business unit leadership team.

The people who own the business model for that particular product for that particular service need to come together. Product, sales, marketing, finance.

Finance need to understand, for example, the implications of, and the attractiveness, they get it, the attractiveness of subscription revenue. So we need those players in the room together.

And it is incredible, Alastair. You know, we went through this in my seven years in Microsoft.

I worked with Alexa Siemens, ABB, Bosch, the really big guys, but loads of very small players too.

And a year and a half ago, I left Microsoft and created I theria Partners, the company that I founded and now run.

Over those years, the single biggest, I guess, benefit that gets called out after a one or two day workshop of that nature is alignment.

It is incredible when we go into these workshops and we ask different individuals to define the value proposition of the product or service that we're bringing to market.

The 10 people in the room will get 10 different answers, quite different.

And so aligning on what that value proposition is and what that new value proposition is going to be in terms of customer pain points, in terms of customer benefits, is hugely beneficial.

Aligning on how we're going to bring this to market, the organisational capability that we're going to need. I mentioned, you know, we may not have this capability in house today.

Well, and in the AI world, this is about data science. It's about cybersecurity. It's about data engineering and so on, so forth. We've got a decision to make.

If we're getting into this game, we need that capability. And if we don't have it today, we've basically got three choices.

We can either recruit it in, we can retrain our existing people, or we can partner to bring that capability in.

And sometimes we partner in the short term while we develop the capability in the longer term. But that's what this is about.

It's defining at a business level what this new capability is that we're bringing to market. I want to be clear as well in terms of AI.

If we're talking about, I don't know, the adoption of Microsoft 365 co-pilot or Google Gemini or something like that to enhance business productivity,

we don't need a business mom workshop to do that.

But we probably should be looking at our operational model in terms of the impact of our workflow if we're taking on agents to improve that workflow.

So the business model is a kind of front end model, if you like.

It's the thing that talks about value and products and services and customers and channels and pricing and roadm aps and so forth.

Or if it's in the back end of the organisation, we're thinking about how the business is set up to execute on that business model to effectively deliver the value proposition.

And now we're talking about in a profitable way. And now we're talking about organisational structure, SLAs, billing, risk management, compliance, financial operations,

our support organisation, etc. So the point is, we need to start with a business led conversation first.

Once we have that well documented and we're very well aligned across the different functions within the organisation, we're super well teed up to hand that over to the solution architect.

And so in a theory of partners, there's myself looking after the business strategy end of that client engagement.

My colleague, Richard Jones, based in the West Midlands in the UK, ex colleague of mine from Microsoft, is a solution architect.

And so where our client needs it, we can help guide on that on the kind of technical, technological aspects, if you like.

Maybe they have that capability in house themselves, maybe they don't need that guidance. Sometimes they do, sometimes they don't.

But we will not engage on a solution architecture project without there being a really well defined business model at the outset.

We've learned the hard way, if you like, that that is a necessity.

Alastair McDermottSo let me try and boil some of that down. And please correct me now if I didn't get this right.

So fundamentally, we have traditional businesses, you know, pre-tech, pre-AI, let's say, who have their business model.

Sometimes there's great alignment with everybody in the team around what that business model is and, you know, what the value proposition is. Sometimes there's not.

So there's their existing business

Patrick WardModel.

Alastair McDermottThen we have the technical capabilities of this new

Patrick WardTechnology.

Alastair McDermottSo this new technology rolls up AI, generative AI, whatever it is, IoT.

And so a lot of businesses are looking at, OK, how do we bolt the new capabilities of this technology onto our existing business model?

And you're saying we need to actually fundamentally reeval uate that because it's going to change your business model drastically.

And you need to understand that before you look at the actual technology and the implementation.

Patrick WardRight.

Alastair McDermottHave I got that right? You've got it spot on.

Patrick WardYeah.

Alastair McDermottAnd then then there's another way that we can use this new tech, which is to simply implement it at a productivity level where we can bring in these tools and simply make people faster at

doing what they're doing already without fundamentally changing the business model.

Patrick WardSo it's kind of like a different approach to it.

Alastair McDermottRight. Right.

Patrick WardYeah, that's right.

And look, what I would say there is we generally encourage clients if they're taking on that kind of capability and it's a business productivity project to do that on a fairly piecemeal

basis. Take it on the layers of the organisation. You know, maybe it's the executives, maybe it's the IT team, maybe it's the data team who take it on first. We see how it goes.

We evaluate it and then incrementally roll out. So there's a there's a there's an approach to that as well.

If it's if it's simply that set of tools, you know, we're not going to get into a kind of deep engagement on either business model or operating model.

But where we are taking on new tech, some of these technologies are pretty transformative, either to the back end of our organisation or to the front end to our products and our services.

If it's either of those, we really need to engage in a deeper business led conversation first. But yes, if it's productivity, our approach tends to be taken on a graduated basis. Yeah.

Alastair McDermottWhat about the what about where businesses get this the other way around? They immediately jump into the tech capabilities. They don't do the business model work. What happens?

Patrick WardA number of things happen. Fundamentally, we don't get business sponsorship and therefore we often don't get business investment. And we very often will.

You know, we used to be on projects where there'd be an awful lot of work done on the technological solution on the solution architecture.

Then various different pieces of information would come to light in terms of our workflow, in terms of our customer, in terms of the industry in which we're working.

So let me give you an example. If we look at smart buildings, smart buildings use sensor data from rooms, from temperature, from presence. They bring in weather data.

They can look at things like the optical analysis of the flame in a boiler to understand how efficient that it's performing and so on and so forth.

In large commercial buildings, you're typically going to see a payback of three to five years in these types of projects.

Typically, you will be getting return of between 15 and 30% on your investment. So it really isn't luck. This is no brainer stuff.

And yet, you know, I had a conversation with a global organisation at last week who are in the HVAC business.

So, you know, they're looking after these complex systems with all these sensors. They have been successfully bringing these systems to market for some time, for many years now.

But the biggest challenge that they're having is motivating the sales team to actually go and sell the digital services. Why?

Because the sales guy is saying, I've got a multi-hundred-thousand-dollar opportunity here to sell this system into this building, into this construction company.

And you, the digital team, are asking me to kind of bolt on to the side of that a suite of digital services that will earn us, what, 5K per month?

So that's going to make my sales process much more complex. It's going to bring in the CIO. It's going to bring in the chief information security officer.

Everyone's going to have an opinion on what this, you know, digital service looks like.

People will want to know how, you know, AI is being used and its adherence to regulations and so on and so forth. And also, the construction industry is very capital intensive.

It's very much about building and building. It doesn't lend itself well to OPEX.

And there's other complications like the organisation building the building and the organisation owning the building and the organisation inhabiting the building are completely different

entities. And so the whole question arises as to where the value actually exists and who's going to pay for it on an ongoing basis and so on and so forth.

Now, you know, the approach that this organisation has taken and is not being successful is by bolting on a kind of a separate digital organisation to the side,

different KPIs, different targets, and trying to influence a sales organisation to go and sell that who are pushing back. The time to get that right was at the outset.

The time to get that right was from the CEO downwards to say digital is part of our future. In fact, we may get to the point where we're not even selling HVAC systems anymore.

We're selling some sort of HVAC as a service, which is kind of the full business model extension where this kind of thing can go.

But we need to agree that upfront and not try and bolt on a digital business and strap it on to the side of our existing business and hope that it will work. It generally doesn't.

Really interesting.

Alastair McDermottI'd love to talk to you more about that, but I'm just aware of time and I have some other questions when I talk to you

Patrick WardAbout one thing.

Alastair McDermottYou did some research on multinationals in Ireland and their focus on AI. Can you tell us a bit about that?

Patrick WardYeah, over the summer we did, we interviewed the heads of AI or equivalent roles in about 15 different multinational companies with a presence in Ireland.

And we did so focusing in on three areas.

One around governance and how the Irish operation interacts with the broader global operation, particularly from the point of view of is it a federated approach to governance or is it a more

centralized approach to understand that a little bit more.

The second was to understand use cases, how AI is being used today and what types of both generative and predictive AI are being considered for the future.

And the third is to just understand some of the opportunities and some of the challenges that these organisations saw in deploying AI in their organisations. Really interesting study.

Alastair McDermottSo can you tell me more like what what did you find? What insights could you pull from that? Was there anything surprising?

Patrick WardI guess the most one of the surprising aspects of it was I guess there's a perception out there and I suppose it to a certain extent it's real that specialist AI capability is very rare.

You know, data science and that seems to be turning a corner.

The feedback that we got was that actually, you know, data science, data engineering, DevOps, MLOps capability wasn't plentiful in the marketplace,

but it wasn't wasn't the largest challenge from a talent point of view.

The bigger challenge was how do we educate our business leaders to think about AI in the context of the overall strategy of the of the business and business leaders are struggling with this.

Some are hesitant. Some are hesitant. They don't want to move forward with it because they feel that there's a certain risk in it. And they're right. There is risk in it.

And so, you know, how do we address that? As you said, I'm a part time lecturer with Smurf at Business School.

We've just introduced a new AI and business analytics post-ground diploma, huge amount for that.

And it addresses these kinds of topics around how should we think about AI in the context of our business strategy,

both at a kind of horizontal level of the, you know, the uses of AI across the various different functions within an organisation like marketing and sales and identifying customer

opportunities and so on and so forth. All the way through to what are the specifics of given verticals? How is AI being used in healthcare? How is AI being used in the physical environment?

How is AI being used in manufacturing in financial services and so forth? And so I think that was one of the that was the surprising thing.

Now, what they were saying was, even though that supply of specialist AI capability is plentiful, firstly, those graduates are tending to know quite a lot about the theory and,

you know, a lot of the practice as well. But their exposure to various tool sets can be pretty limited.

And so there's a fair amount of retraining that needs to happen when they join the organisation.

And I guess the other is because this is such a fast moving domain in terms of the development. I mean, you know, you know, as well as I do, Alastair, this is moving very rapidly.

There's a need to continue to sharpen the saw there. There's a need to continue to educate, train both specialist AI, but also business AI skills across the workforce.

Alastair McDermottYeah, I mean, I don't know how people who are non technical, but who want to keep up with this can possibly keep up because there's so much is happening.

And there's so many changes to the technology. Like every day I get an email with, you know, the latest 20 things that have happened in the world of AI.

And being somebody who's really interested in this area, I find it difficult to keep up.

I don't know how somebody who's who's not deeply involved in this area can keep up with all of the major changes. There's a lot.

It's it's kind of the Americans call it drinking from the fire hose. It's sure it feels that way. Yeah. Yeah. So it is it's because of that.

I think what we need to understand that the major kind of benchmarks and some of those are happening now with we're moving from where we interact with these systems through chatbots to

typing in our answers or talking even with voice. To where we can start to give them control of using our actual computer to to for agents to take actions and things like that.

So there's those types of capabilities are starting to come on stream. They're still very early days, but those are going to be pretty major.

And those are the kind of things I think people need to keep up with those kind of major, you know, milestones. But it's it's tough.

And it's one of the reasons why I do the podcast is because I want to talk to people about this, you know, absolutely.

Patrick WardAnd it's fantastic.

And I suppose that's what I'd say to people is and when I'm talking to organisations is to encourage learning to encourage, you know, identify champions within business units.

Get them going out there doing research in terms of how I is being used in our industry report back, be our ear on the ground.

Go and take courses, attend podcasts and, you know, webinars from Gartner and others as well. Keep learning because this stuff is changing rapidly.

I find it very interesting and kind of stimulating. I think there are plenty of people out there who do.

So identify those people in the organisation and tap into that and tap into that capability.

Alastair McDermottOne of the things that you do is you do some AI envisioning sessions. Sorry, I'm stumbling over that.

Patrick WardEnvisioning, yes.

Alastair McDermottEnvisioning sessions for SMEs. Can you tell me a little bit more about that?

Patrick WardYeah. So I suppose this is before we get into any kind of business model analysis or anything like that.

There's a lot of organisations that are coming to me, both enterprise and SME, who are saying not so much give us an introduction to AI.

I think a lot of people have, a lot of people, you know, it depends. A lot of people have had an introduction to AI. They get the kind of general concepts.

They've used ChatGPT and other LLMs, et cetera. But they want to understand how this technology is being used, as I said earlier on, at a horizontal level.

So, you know, within general business productivity, within enterprise integration management, within marketing to generate copy and images and so forth.

But also how is it being used in our industry? What we'll do at the outset of any engagement of that kind is a level of preparation where we engage customers to understand.

Tell us more about your organisation, about your business goals and your challenges. What's the level of awareness of AI in the organisation? What's your capability at the moment?

What does organisational readiness and culture look like? Is there resistance to change or is it a fairly fluid, dynamic workforce? What do data assets look like?

What's your technology landscape look like? So we'll get an assessment up front and then tailor the content accordingly. I think that's really important.

We'll also ask about key competitors and we will try to do an analysis of how your competitors are using AI today. Some of that is through just press releases and marketing materials.

But it's also if you look at some job postings from companies, you'll get a really good indication of the kinds of areas that they're focusing in.

So there are various different ways in which we can get some level of kind of competitive insight. Then we go and run a workshop, typically takes a day or two days.

We go into the classifications of AI, I guess we call it, and then how they're being used. It might be in financial services, fraud detection and anti -money laundering.

In a software company, we'll go into software development in more detail.

In a manufacturing company, we might talk about automation, the use of computer vision for quality control and a manufacturing line, and so on and so forth.

We'll talk about their competitors and try and give them some insight into that and get some insight from them. We'll talk about risks, how to mitigate them, regulations.

And then we will do a kind of a triage, because this is a workshop, it's not a kind of a one-way broadcast.

And we'll always say, if there are people within your organisation who are already using AI, get them to come into the workshop as well, give their perspective.

But we will then do a facilitated session where we will seek to kind of brainstorm different ways AI might add value to the front end of the organisation in terms of the products and

services and how it can address customer pains and customer gains.

And/or how the operating model might be improved through the use of agents, through the use of business productivity and so forth.

Yeah, it's a nice approach to raise awareness in a kind of an applied way for the organisation.

Alastair McDermottI want to ask you something that is something that we didn't plan to talk about, but I just want to get your perspective on the AI for productivity side of the equation,

because I think that's where a lot of people are starting with because they're trying to understand it.

And so they're starting to use these tools, they're seeing some really interesting and potential productivity gains of 3x, 5x, 10x even on certain tasks.

I'm just wondering, what's your thoughts on AI for productivity and how important do you think it is and how much do you think people should focus on it?

Patrick WardDo you have any thoughts on that? Well, I think, first of all, when we're implementing change, we'll typically go to do some sort of business case.

We'll try and work out, you know, what the investment is going to be, what the return on that investment is going to be. Does it meet some sort of hurdle rate for the organisation?

Don't do that with business productivity because it's extremely difficult to measure. Go to Gartner and have a look at their published literature.

Go to McKinsey, Boston Consulting Group and so forth. And they'll all tell you that what you're probably going to see is probably a 5% to 15% uplift in productivity.

That's good enough for me. Let's go. And let's not just roll it out right across the organisation.

As I said earlier on, let's pick a layer of the organisation or a department within the organisation. Let's roll it out on a piecemeal basis.

It's much more cost effective to do that, even though you don't get the kind of discount by buying licenses for your entire enterprise organisation.

You also don't roll it out to people who aren't going to use the capability. So roll it out on a piecemeal basis. Learn from that and also provide training.

Don't expect that this will just land and be picked up. Some people will pick up and run with it. Others won't. It's only fair that we don't leave anyone behind on this.

We need to provide training. It's really important to put governance in place. And so, again, we deal with this in the envisioning workshop.

It's kind of the next steps of how we're going to roll this out. Governance, policy, use cases, training and sharing are the kind of five areas that we focus in on.

But your point about business productivity, can I give you a story related to software development, actually? Oh, yeah. But it's about my business productivity.

And you'll see what I mean in a moment. I was about to run one of these envisioning workshops and I met with the CEO. It's of a, I suppose, a 30 person organisation.

They're in software, they have software product. And in my kind of information gathering, a couple of calls with the CEO, I asked, well, you're a software company.

You've got a software product. Do you use software in your software development process? And he said, not really a little bit on a piecemeal basis.

Some individuals do, but there's nothing official. Now that's a red flag because now we're talking about the kind of consumerization of AI. It's a red flag because we've no policy in place.

We've no training in place. Those individuals could be uploading customer data. They could be uploading IP.

So it's, you know, in our interviews with these companies over the summer, it was interesting to note some companies completely blocked the likes of ChatGPT. Others didn't.

But the ones who didn't and allowed access to it had strong policies in place and strong training in place to say, here's the rules of how you can use this. Here are the risks.

Don't do the following things. Don't be putting proprietary information up to, you know, that'll train a model externally. And others said, you're not using that.

We're going to use a kind of a walled garden approach. We want to pay for a service. And you can use that, which won't train the model and knock yourselves out.

You can put whatever in your prompt. But the point is, I'm doing this information gathering. I'm noting that he also said that the CTO was a self-proc laimed AI sceptic. Fair enough.

So I thought, well, rather than me rocking in with, you know, IDC say this and Gartner say that and McKinsey say that in terms of,

you know, it's typically kind of 30 to 50 % uplift when we're talking about software development. So it really is material.

You really need to be looking at this at your software development organisation. You've got a software product.

Alastair McDermottLet me let me say it this way. If you're in software development, you're not using AI.

Patrick WardYou are absolutely nuts.

Alastair McDermottYou're absolutely nuts. It's crazy.

Patrick WardExactly.

Alastair McDermottAmazon said that they saved 4,500 developer years, not a typo, 4,500 developer years of productivity by using AI in

Patrick WardTheir processes.

Alastair McDermottSo that's just now I know Amazon are massive, but yeah. No, but, you know, scale it down. It's really useful in software more than any other.

Patrick WardYeah.

Alastair McDermottSo anyway.

Patrick WardSo this story kind of ties that in, but also my own personal productivity.

So rather than be going into the workshop with kind of, you know, slides saying this is what these different organisations say, I thought, hang on a second.

I've got a friend called Trevor Jabling. He's the founder and CEO of Dovetail. Dovetail is an Irish software development organisation. They've been around quite a long time, like 20 years.

They've got customers like IKEA and Aramark and World Rugby. They're, you know, really well-established player. Trevor's personal friend of mine were out hiking in the Wicklow Hills.

And he's telling me this anecdote probably nine months ago about using, I don't know, was it ChatGPT? And asking it for some code for a particular scenario.

And rather than giving him that code, ChatGPT came back and asked him about what you might call boundary cases. It was some sort of rostering, let's say.

How many people will be involved at any given time? When do they come on shift? When do they finish shift? Do they work 24 hours? Do they work weekends, et cetera? Trevor gives the answers.

It gives them the proposed software. And his point was that he was so impressed that it had asked further questions before it gave him the code. And it really stuck in my mind.

So I thought, hang on a second. I rang Trevor and I said, I'm doing this workshop in a couple of days. Would you mind going on a Teams call with me?

And we'll record that anecdote for a couple of minutes and I'll bring that into my workshop.

It's nice in a workshop to kind of mix the different media, if you like, and it's not all me talking. No problem, Trevor said. And I'll tell you what I'll also do.

I'll survey the team and find out how they are using AI today. And maybe there's another anecdote or two to bring into your workshop. Great.

I go on the call and Trevor is clearly taken aback. He's been around the team that morning. There were 10 devs in that morning.

And he's got pages of different, you know, scenarios, different use cases of how the teams, you know, we're developing software for a French client.

We've asked the LLM to give us a thousand realistic names and address, French names and addresses.

Beautiful, because now we're not even dealing with, you know, a customer database that needs to be managed extremely carefully because it's PII. We're now dealing with synthetic data.

So, you know, it overcomes a lot of that. And he's got a developer working for him in Brazil. Suddenly his language, his written English has just taken a complete transformation. Guess what?

He's putting it in Portuguese. It's coming out in English and it's really well crafted.

He's asking, you know, another dev is making the point that you can ask at the relative merits of two different architectural approaches.

Now, each, you know, throughout the call, he also mentioned, you need to be really careful. You can't take verbatim what the LLM is giving you.

You need a senior dev to cast their eye over it and see if that's okay. You need to be very careful to see if junior devs are using this. 24 minutes later, that call completed.

I can't bring 24 minutes of content into a workshop and play it, you know, conversation between me. And I've got a workshop in two days. Here's my personal productivity.

Download the transcript, upload it in this place to Claude. I, and before I did, so I anonymized, didn't mention Dovet ail, et cetera.

And I described it as, you know, the founder of an organisation developed software that has explained all these scenarios. Can you please articulate the scenarios? Boom.

20 scenarios, not 19, not 21, 20 scenarios. And there's the French example. There's the Portuguese example. There's the comparing the relative merits of different architecture example.

They're all there. It's all there. And it summarises at the end saying, he mentioned a number of times through the call how important it is to get, you know, a human in the loop, as we say.

Review the code. Don't take verbatim what the LNM is giving you for various, you know, all the good reasons we know about, hallucination and so forth.

And so now I'm able to go into my workshop, give that story. But it's about my personal productivity. But it's also about 20 different ways that these devs are using.

Trevor also mentioned, you know, if I take your license away, would that be an issue? And they were kind of like out of my cold hand hands.

That's the other point about this stuff is it's really enjoyable to use it.

And so if we if we if we do things right, we put policies in place, if we purchase the right tools, if we offer training,

this stuff really begins to take off and it really begins to add value. But but there is a way of doing it right. And there is a way of doing it in a very risky way.

And we need to know the difference between those two. Just one other very quick anecdote about software.

I had a conversation yesterday with a with a potential client who does embedded software development. And he's like, can I share my screen for a moment? And he puts up on the screen.

He's like, do you see that those five points? He said, that's from a code review from ChatGPT. We put in our embedded software and I find those five points.

And we said the precursor to this story is we've had an intermittent bug in our software that we discovered.

It's been a nightmare to get to the bottom of us because it's intermittent and it's very difficult to reproduce and all the rest of what's happening out there in the field.

It was the number one finding of the code review done by ChatGPT. You've got a null pointer issue here. And that's exactly what the issue was. I just thought, wow, amazing.

So, you know, there's real value in this stuff. People talk a lot about hype and, you know, kind of discard the whole lot of it because there's a lot of hype. Yes, there's a lot of hype.

But, you know, there's a lot of value as well.

And the key for, I think, someone like me and someone like you, Alastair, is it's about cutting through that hype and drawing attention to the real business value that exists.

Alastair McDermottI think that's a great place to stop. That is fantastic.

One thing I want to mention, I want to give a plug to a book that I've written, which is called An Absolute Begin ner's Guide to Using AI.

So if you're looking for somewhere to start, you can visit my website at AI powered dot biz slash guide. And you can you can find that book, An Absolute Beginner's Guide to Using AI.

And I'm 100% with you about the, you know, it's so important to keep the human in the loop. It's so important to train people. It's so important to put processes in place.

And it's not that hard to do all of this stuff either. Yeah, those those things. It's it's it's easy to get started with that part.

And I think that conversations like this one will hopefully help with that as well. And Patrick, thank you so much for coming on the show.

Where can people find you if they want to find out more?

Patrick WardSo Patrick Patrick at ItheriaPartners.com is my email address. You can find me on LinkedIn. Love to hear from you. Love to have a conversation and hear your experience about the use of AI.

And yeah, every client that I meet, I'm not just helping them, but I'm also learning from them. This is it's an amazing kind of nourishing of the soul.

I mean that quite genuinely that the capability that's out there is just mind blowing. And to be right at the centre of it is a real privilege. So, yeah, love to hear from anyone.

Alastair McDermottYeah, I think that's the thing for me. I was helping helping some lab scientists actually who do who process data samples and they process blood samples and then turn that into data.

And then they process the data and then they put that on a

Patrick WardMap.

Alastair McDermottThey say, well, the positives are here.

Patrick WardThe negatives are here.

Alastair McDermottAnd we reduced their time on all the admin work by 90%.

And so instead of doing all this drudge work of literally sheets of paper and typing in numbers and things like that, now all of that is taken care of.

And they're just actually focusing on, OK, what's the implication of this? What's the let's let's analyse it from bigger picture perspective.

And there's always really smart, highly trained people and

Patrick WardExpensive people.

Alastair McDermottAnd now they're actually enjoying the job a lot more because they're not doing all the grunt work.

That's one of the things I love about AI, but it's actually taking away the stuff that we don't like doing. So that's where I think it's really great.

Patrick, it's been fantastic to talk to you today.

MaxThanks for listening to Something About AI. Everything discussed here, and a good deal more, is backed up by free resources over at humanspark.ai. No cost, no catch.

If this episode was useful, pass it on to someone who would use it, and subscribe so the next one finds you.