The Recovering Architect: Architecting Outcomes Instead of AI Pilots with Andy Haley, Sullivan and Stanley

Jim Clark: Welcome to another aibl interview. Today I am joined by Andy Haley, CEO of Sullivan & Stanley, a specialist change and transformation consultancy. Thanks very much for taking part, Andy.

Andy Haley: It is a pleasure. Thank you, Jim, for inviting me.

Jim Clark: We have spoken to a lot of mid-market leaders at past aibl events. A lot of it veers towards why a transformation programme is not delivering. There is also no shortage of why AI pilots are not scaling. I know this is very much your patch. I do not want to put words into your mouth, but you have over 20 years of advising firms on the subject. A chunk of that time was spent at Accenture. So we are all very much looking forward to hearing your take.

Andy Haley: My background is that I am a recovering enterprise architect. I mean that in the sense that I have discovered how important the discipline is. The thinking and thought process behind enterprise architecture matters more in a modern world where the architecture of the enterprise is really important.

What every tech revolution does, and AI is an example, is force a level of disruption into an organisation that has already been disrupted. That manifests itself as a huge amount of opportunity, confusion, reticence, and a lack of belief in certain ways. AI is an incredibly powerful potential technology. But it relies on an awful lot more than just an understanding of AI to drive sustainable value.

You have to really understand what you have to re-architect to drive the outcome you care about. At Sullivan & Stanley we do not necessarily talk about AI in isolation. In fact, we never talk about AI in isolation. We are passionate about helping our clients understand the outcomes they care about, and then how you architect them. That runs across AI, human intelligence, technology intelligence, and execution intelligence. We package that up into the Intelligent Enterprise.

All of these technology hype cycles come around, and all of them have the potential to drive phenomenal amounts of value. In isolation, they work in greenfield sites. No legacy technology, no legacy mindset, no legacy process, no legacy operating model. You can build it literally to take advantage of the technology and get the outcome you care about. The inconvenient truth in an awful lot of big firms is that you do not have that benefit. There are seven or eight different operating models all trying to work together. They are all legacy from previous transformations, because you never really turned off the last one.

What AI has highlighted, yet again, is the people who can genuinely shift from “the answer is AI, what is the question”. AI is going to help me deliver automation. Is it really? What is the outcome you are looking for that will manifest as value to your client? And how is AI going to be a catalyst for you to rethink the way you deliver that? Because in isolation, it will not.

That is the big learning I had from working with big organisations. Looking at one technology in isolation does not work. Looking at one domain of the architecture does not work. Without the technology, the data, and the mindset, it does not work. If you just look at a location without looking at the technology, it does not work. You have to look at it front to back, and you have to architect outcomes. AI is a fantastic catalyst to get you to think that way.

Jim Clark: We are going to dig into that. You have said that transformations rarely collapse overnight, they just do not deliver the full value they promised. You have written extensively about that. So what is actually happening in that gap, Andy? Is it one bad decision, or death by a thousand small compromises?

Andy Haley: That is a big question, Jim. It depends on your culture. If I think of some of the really big organisations, they get budget, they spend budget, they do the job until they run out of budget. Then they say, well, we got as far as we could, I wonder what budget we will get next year. And the cycle carries on that way. That is a very real thing. The motivation of the firm is, I have got budget and I would love to do stuff. Doing stuff is incredibly complicated. So the primary driver there is the budget cycle.

There are other extremes where it is, we have got really grand ideas and we are all committed to doing it. But did we really understand what was involved in getting done what we needed to get done? We did some Intelligent Enterprise research. If you really want value from a transformation, you have to convince yourself the value is there in the first place, and that your business case is accurate. Second, you have to understand what is involved in capturing that value. What do you have to influence? The bigger the organisation, the more variables, and the harder it is to understand that blast radius.

The third thing, arguably the most difficult, is whether you are configured to be able to influence them. Or do you rely on things changing that the programme has no influence over, so it just becomes a blocker? All of those reasons are valid, and there are lots of different types of transformation. Fundamentally, that is the logic for me. Do you understand that the value you are looking to capture is there, in whatever you deem to be valuable? It does not necessarily have to be a hard business case. It could be a cultural outcome you care about. Do you understand what you have to be able to influence to get that outcome? And are you able to influence it in the way you need to?

If you cannot say yes to those three things, you probably should not start. If your answer is roughly, then you should probably do a bit more work. It is a very simple view. But unless those three things are true, you are going to really struggle. And somebody has to be accountable for making sure those three things are true.

Jim Clark: I am going to get to accountability later. You touched on culture and behaviour. What is the tell that it is the problem from the early stages?

Andy Haley: Good question. There are a number of reasons. One is when everybody smiles and says this is a good idea, and nothing actually happens.

In my old days with my previous firm, you would be full of the joys of spring. You come in and say, look, we are going to do a new strategy and transformation plan, and it is going to be great. You are full of energy because it is a new client and you are enthusiastic about why they have engaged you. You have also sold some work, so you are pretty pleased with yourself. And you cannot understand why the organisation is not meeting that enthusiasm.

Then a couple of clients will pull you to one side and say, look Andy, here is a folder. It might be a head of department, or the MD of a region. It will say technology strategy, which there is no such thing as, by the way, but that is another thing. There will be a series of folders in there. Strategy 2016, strategy 2018, strategy 2020, strategy 2022. So how many of these do you think get implemented? And they will have been done by different third parties. We never implement.

If you are going into an organisation to make a change, and that organisation knows this is all just arm-waving and PowerPoint, you are probably not going to get anywhere. So that is a sign. In the very first meeting, is anybody really up for this? Or do they all think it is just another exercise that will sit on paper and wash over them? That is a pretty good sign, and the world is full of those examples.

In my experience it is often the part of the puzzle that is least understood. It is the least well funded, and the least well provisioned for from an effort perspective. We get very excited about a new capability, a new technology, or a business case we have written that means we are going to beat the competition. But we completely forget that there are these warm organic beings, in all of their variation and beauty, that have got to be accommodated for. And there is no one way to do that.

The people who are good at it are very expensive. But it matters. It is just very difficult to say what the numerical value of a good culture is. The numerical value of replacing SAP with something else could be potentially massively compelling. But if you cannot articulate culture in language the board cares about, it is very difficult to get it funded. And to say it is an enabler of a business case, well, the first time there is a cost pressure is the first time it is going to go away. That is a cultural limitation in and of itself. If you have not got a culture of valuing culture, that is going to be a problem. But it is all part of the rich tapestry of transformation.

We have a view that as technology becomes not commoditised, but more easily available to more groups of people for less money, it becomes increasingly difficult to drive differentiation from technology alone. So you have to focus now on how you get the judgement, the desire, the leadership, the values, and the commitment. All the human stuff. That is actually the thing that makes transformation successful, and it is becoming increasingly important.

Jim Clark: Good answer, thanks Andy. Staying with that gap, particularly from our audience’s perspective, mid-market leaders, where does it usually show up? Does it start early, before anything has been signed off, because there are fewer layers of governance checking the numbers than in a big enterprise? Or does it show up later, once they realise they cannot actually pull this off?

Andy Haley: In my experience, it is later. It shows up when it does not seem to be translating into the outcomes they care about.

In mid-market organisations, making decisions is typically much easier, because everybody is in the same country or occasionally the same office building. So you can get everybody together to make a decision. You can move relatively quickly through the design phase, because you have not got a global set of stakeholders and there is no governance. And are they regulated? Probably not, for the mid-market. So you can convince yourself that you are making loads of progress.

Then it hits the road and you need teams to actually execute. That is typically when you find out it is not working. There is an inflection point there. When you get teams working on things, the run rate, or the cost of execution per week, is much higher than when you are doing the definition work. And it gets difficult when you involve the people. Transformation and change is not a scale of value, it is a vector. It has got a pace and it has got a direction.

What I find with mid-market work is that there are fewer things happening in parallel. So it is more straightforward to understand and be confident everybody is pulling in the right direction. You can be confident in your leadership team, because they are all together in the same room. It is much easier for everybody to be coherent and moving the right way, because it is a much smaller environment.

Then the challenge arrives. I am confident all my outcomes are aligned, but are the team aligned to these outcomes? Do they really understand who the customer is for this outcome, and why it is important to them? To the point where they can build an emotional attachment to what they are trying to achieve as a group? Because that is where the discretionary effort comes from. It is where you get the “come on, let us get this done together”.

So there is that lens, and then it starts to get a little bit shaky. And then there is whether the team is actually optimised to do what we want. In huge organisations you can swap people out and get the team to work that way. In medium-sized organisations you have not got a choice of hundreds of people for a role. You cannot get a TCS, an HCL, or an Accenture to come in and say you need to change out five of these people. It does not work. Your people are the ones who are going to be driving a significant amount of the progress, maybe augmented by a third party.

So there is that lineage. Is everybody pulling in the same direction? Does everybody understand where we are going? And is the team optimised to make it work? The risk increases as you get down that chain. As everybody knows: forming, storming, norming, and performing. When everyone is hanging out together there is no real impact, because everyone is talking and getting to know each other. But then it is, hang on, we have missed three milestones, we really need to start doing things. So in small and mid-market firms, the pain manifests further down the execution chain than it would in a big corporate, which faces all of those challenges all the way through the process anyway.

Jim Clark: In terms of the scrutiny that is required, particularly in the early stages, mid-market firms often have no dedicated function for it. So what would the solution be? Is it getting somebody outside to help, because you mentioned they cannot necessarily hire in the Accentures of this world? What could they come back with to mitigate that?

Andy Haley: The world is awash with methodologies to make this stuff work. Sometimes they are too prescriptive, and there are too many of them. The answer is a software platform, what is the question?

So there are three things you have got to make sure are aligned. One, you need really senior sponsorship, to make sure everybody knows how important this is. It has either got to be the CEO or a board member who sponsors any big transformation. Then you have got to be confident you have got the operational capabilities, driven by and owned by the operating officer, to actually make the change happen.

And then you have got to make sure you have got this enterprise readiness. That is all the technology, tooling, process, and data, so you are confident you have the instrumentation to diagnose this stuff. A lot of the problem with the mid-market is, well, I do not really want to spend a shedload of money on Clarity, but I cannot really run it off an Excel spreadsheet either. So what is the right middle ground? How do you make sure you have got the minimum acceptable tools and processes in place? So that diagnosing this stuff and keeping track of it does not become a cottage industry. But equally it is not every person for themselves, with all the reports completely disparate and not following on from each other.

There is a really elegant transformation architecture you have got to put in place. It makes sure you have the enterprise foundations to support where you are going, and that they do not become an afterthought. They are a deliberate choice. There are loads of outfits out there that will help you. We are one of them. I will not plug us too much, but we have our own platform to help organisations understand the degree to which they are confident in that lineage. Do I understand the outcomes I care about? Is my team optimised? We are not the only show in town, there are a few others that do it.

Jim Clark: Linked to this, you have mentioned the research you published. Only 7% of large UK organisations consistently get the full value they expect from transformation, which is shocking. Is there any reason a mid-market firm would do better than that, or perhaps worse?

Andy Haley: I think you have got to double click on why only 7% get the full value of their transformation. You could argue that even that is optimistic. In my experience, if you do get the full value of a transformation, you were not ambitious enough. So there is a way to go. Maybe 7% is okay. For me, it is about understanding why you misunderstood the value of what you were trying to do.

The default is to say it must be the process, the tooling, the skills, the third party, or the complexity of our architecture. When actually, what was the value there in the first place? A lot of big firms, and the mid-market does not suffer with this quite so much, actually. If you go back to that logic, do I understand the outcome, do I understand the blast radius or the variables, and can I influence them? That is much easier to be confident on in the mid-market. It is far more likely that mid-market firms do not capture the value because it was never there in the first place, rather than because they could not execute it. Execution is much more simple.

That would be my hypothesis. Mid-market firms generally want to grow, and want to grow quickly. They are not in saturated markets, or they have got a secret sauce, which is why they are transitioning. So they are far more ambitious. In my experience, a lot of mid-market organisations will write business cases as sales pitches for funding. Funding in the mid-market is far less available than in some of the big corporates, where you can play all sorts of financial jiggery-pokery around using OpEx to drive change.

So I think that is a reason why mid-markets will probably find it easier to be confident they can execute, but are more likely to overstate a value proposition, because the investment tends to be a little bit scarcer. The degree of optimism around the benefits of that business case may well be slightly more questionable in a mid-market. I am not saying this is gospel, but in my experience that is behaviour that may well be there. It is not conducive to claiming the full value of an investment. So for me, it is about understanding whether the value was there in the first place, or whether you were being a bit over optimistic. That then manifests itself as a delivery failure. And actually it was impossible to get the outcome you were driving at, in the way you were trying to do it, because you overstated the value in the first place.

Jim Clark: It is exposing weakness indeed. You said a lot of organisations do not have someone responsible for how the business all fits together, not just the tech side. Staying with mid-market firms, does that job usually just not exist in smaller firms? Or does it end up falling to the CEO, whether they are right for it or not?

Andy Haley: It very rarely exists. The reason I said I was a recovering enterprise architect at the start is that it is a difficult role. You could say you are accountable for everything, because it is your job to make sure everything works in isolation and together. It is not really that. It is a communication role. But I prefer to say the architect of the enterprise. That person who understands how everything fits together, and who to go to for the details.

So I think every board or executive leadership team has a role missing on that forum. Right now the assumption is that the CIO will work with the COO, will work with the people officer, and it will magically all happen. I think somebody needs to steward that conversation and that set of relationships. It makes sure they are working, and that there is a coherent understanding of how the enterprise is architected. You then stand a better chance of understanding the blast radius of the change, because somebody is able to speak all those languages and knit it together. With everybody motivated for the outcome the organisation cares about, which is typically what they commit to their shareholders.

So that role does not exist. But if it did, it has to be scoped very carefully. It is not the person accountable for the entire organisation. It is the person who understands how it fits together, and can speak all the various languages to bring the stakeholders together. For a coherent conversation that is sensitive to everybody’s individual motivations and pressures. That is a very difficult role.

A lot of organisations have got a chief architect, but it never encompasses the business, the org chart, or org design. That is normally an HR thing. So even if you have got a really good enterprise architect who understands the technology, they very rarely connect into a coherent business architecture structure. So that you can say, this business capability is optimised for the things I care about, because we have got this platform running with this data model, in this location, with these APIs, and it conforms to the functional and technical adequacy test we have got.

So it is just impossible to see how you are architecting the outcomes your clients care about. Because the existing roles in leadership are not orchestrated in the way they need to be. They do not give you the insight to be confident you understand the variables you need to change, and the influence you need to be able to change them. So I do not think it is a role that owns it. It is a role that coordinates the others, and speaks their language to disarm them and get them all working together.

Jim Clark: You made a good point there. Often it is split, and the assumption is they are speaking together, but they are not. So some sort of overriding coordination is required. You have also said, in the many papers you have published, that building comfort with change takes years of gentle conditioning. So if a mid-market leader does not have the years it might take, what gives? Is it the speed of conditioning, or the size of what they are trying to change?

Andy Haley: There is the unstoppable object and the immovable force. Culture is easily the most difficult thing to change. You can force change in, but I have this notion of how resilient your culture is. The resilience of your culture is a function of how much disruption you can place upon it.

Transformation and change is basically a little controlled explosion. It does damage. It makes people uncomfortable. It unsettles them. Now, if you prepare for that little explosion by strengthening the things that need to be strengthened, that helps. Things like people’s understanding of what it means for them, and their ability to listen and be heard about their concerns. All those things that make the human condition so rich and interesting. It means it is never the same approach in any two organisations. No two organisations have the same people, with the same motivation, the same experience, the same culture, or the same battle scars.

So it is incredibly difficult. Now, you have a choice. Just do it, and you will probably find you get a peak of attrition. Or you say, I am going to work really hard at making sure these little explosions are not so disruptive that everybody capitulates. Or you say, everybody is so susceptible to change that I can make my explosions reasonably big and go over it quickly. The nuclear bomb is that you outsource something. You say, we cannot change the culture quickly enough to get the outcome we care about, so I am going to outsource it and make it somebody else’s problem. I have seen examples where that has happened.

Jim Clark: So would you advocate, in the case of the mid-market, a series of smaller controlled explosions? Breaking that ambition into smaller steps they can work through, to avoid the ultimate atomic bomb going off?

Andy Haley: It is a terrible answer, is it not? It depends. It depends on the culture you are deploying the change into. Some mid-markets have such enterprising cultures. Everybody is hungry to do things differently, and they drive the change themselves. In the really modern mid-markets, they do not have a concept of change. They just have product teams that optimise themselves perpetually over time, and everybody is comfortable with that.

Then, all the way at the other end, you have some of the charity sectors. And they have to change. But this does not feel right to me, this is not what I joined the organisation for. I joined to do good, I did not come for my day job to change every five minutes. So that is a completely different approach.

So understanding the existing culture you are trying to deploy the change into will mean a very bespoke answer to how you divide it up. And again, very rarely do I see that pre-work done in anywhere near enough detail. Enough to understand what it is you are asking of the culture that is there at the minute. Because it is there, whether you understand it or not. And it will be the reason you succeed or fail. If you get that wrong, you will find it incredibly difficult to make anything stick.

Jim Clark: You have talked about in your research, again, that 42% of organisations are stuck in AI pilot perpetuity, linked to your previous point. I contrast that with 15.5% who have scaled AI. So how many failed pilots does it take before a leader should admit that the problem is the business, not just the technology?

Andy Haley: I think it is the enterprise rather than the technology. Business implies that it is the process gang’s fault. For me, there is an evolution. There are loads of parallels with this.

Think about what happens when you want to clean up your technology estate. You say, there is this big platform we spend loads of money on, and I am going to decommission it, because there is a really compelling business case to decommission it. Now all I am left with are loads of little things where there is no business case to decommission. So not only do I still have a complicated landscape, I now cannot build a business case to simplify it. So I am in a worse position, effectively, than where I started. Now I am trapped, and I have got to go to the shareholders with a very difficult message.

It is the same thing with AI use cases. There are some quite lucrative AI use cases you can deploy, but they are fairly easy to get to. It tends to be, let us put Copilot in to make it easier for us to proofread emails and summarise things. One assumes I can do more with the same amount of time. There is a productivity gain, and I can build a business case. Too often that saving gets banked. Rather than saying, okay, how do I reinvest in the architecture foundations that will open up the next order of business case?

Those foundations are things like, how do I transform a process front to back, and engineer the data foundations to be sufficiently robust to feed that agentic architecture with what it needs to be automated at a higher level? If you have banked the savings from the first one, you cannot offset the costs of the second one against savings from the first one. So you have got to build a business case for that, and that is typically huge. You need to do an awful lot of below the line engineering to sort out your data, your governance, and your operating model, before you even think about doing the higher level stuff. Whereas if you clubbed all your AI proofs of concept together, you could probably make a case for doing it.

Then at the really mature level, you could say, okay, at what point have you got your talent strategy? How are you going to make sure that if the cost of AI goes through the roof, which it looks like it will based on the infrastructure being built, you have options? They are going to want the money back for that, by the way. So how do I start to back out agents and replace them with people, when it gets cheaper to do that for some of them? We have not thought of that. Those are things you are going to have to solve for.

You can make AI do cool looking stuff. But when it comes back to it, is the CEO happy to be accountable for something an agent did that could result in them going to jail? Absolutely not. Do they even know that is a problem? How are you going to make sure you are not doing that? Where are the lines?

How are you going to make sure you have insulated yourself from massive compute costs, by deliberately building a context layer that watches what happens and enriches itself naturally as part of the operation of the organisation? Those are real questions that an AI architecture has to answer. Because if you think about it, building digital things is now pretty much free. So what matters is the quality of that thing, and the degree to which it is aligned to the outcome I care about.

Changing that mindset of the organisation is massively important, and without it you will never get out of pilot mode. It requires an entirely different mindset. It is like when you had to pay money for your camera film. You were very cautious about what you took pictures of, and you were careful when you pressed the button, because that camera film was a finite resource. Now, with digital cameras, you just press away and take thousands of photos. What matters is the quality of photo you want, so I can take any number of photos.

It is a whole other mindset to take your photos. You are now focused on the quality of the photo, rather than the fact that it exists at all. That is a mindset shift digital development is going to have to go through, powered by AI. It is an entirely different way of thinking about how work gets done. I do not think anyone is going to scale AI until they start thinking about real outcomes, enabled through development time and digital products that are now almost free to produce. So it is now about the degree to which that is optimised for your outcome, rather than the fact that it exists at all, if that makes sense.

Jim Clark: That has gone beyond a lot of the discussion, which is that you need to get your data foundations connected. You are talking more about the cultural impacts, and thinking about the budget. Not necessarily banking it, and putting it forward to future projects. And of course, going back to your earlier point, moving away from that folder of strategies with a new strategic plan each year.

Andy Haley: And to the point about differentiation, as the barriers to using technology get lower and lower, you drive competitive advantage by the culture you build. How committed are your people to what you are trying to build, and how much discretionary effort do you get? How good is their judgement? That is really where the sustained competitive advantage comes from. And in that lens, AI cannot help you. It is a tool to help you on the way. But investing in the emotional security and safety of those people, so they can turn up to work and be their best, most productive selves, in a way that only humans can, is really the magic of sustained differentiated value going forwards.

I think the mid-market is really well placed to go after that. Really well placed, because they have not got the complexity of the tech estate that the big firms have. I think there is a huge amount to go after in that space, and I think it is fascinating. If you think about the four intelligences I spoke about earlier, with our Intelligent Enterprise, human intelligence is becoming increasingly the loudest lens on a transformation programme. It used to be technology and execution, but now that is probably enabled through AI. The source of sustained competitive advantage, I think, is in the human intelligence.

Jim Clark: Thank you very much. I have really enjoyed this conversation. We are coming up to time. I think we have probably answered this indirectly in the previous questions, but my final one is this. A mid-market CEO is about to approve a major AI or tech programme. What is the one thing they should answer before signing off? Is it about technology, or about the organisation? I think we have really answered that. It is about the organisation. But prove me wrong.

Andy Haley: It is about being confident that this AI programme is going to result in the enterprise being better optimised to deliver the outcome he or she cares about. That is the mindset. Can I articulate why, as a result of this AI programme, my enterprise is more optimised for this outcome than it was before I started? And that is not just an AI thing. It is the front to back enterprise. However you choose to articulate it, and whatever language you use internally, does not matter. But you cannot just talk about the AI.

Jim Clark: I was going to ask, what is the question that usually does not get asked that they really should be asking?

Andy Haley: What is the outcome that we care about? Are we all aligned on the outcome we care about? It is as simple as that.

Jim Clark: A great way to end it, Andy. Thank you so much for your time, I really enjoyed it.

Andy Haley: Thank you.

Why do AI pilots stall instead of scaling?

Haley argues the blockers are enterprise questions rather than technology questions. Most organisations can make a pilot work in isolation. What they have not answered is which outcome the pilot serves, which foundations need to change to support it, and who is accountable for making that happen. The easy use cases sit on top of existing processes and data, so they scale no further than the process allows. Getting past pilot stage means redesigning a process front to back, and engineering the data foundations underneath it. That is a different order of work, and it needs a different conversation at board level.

What three tests should a business case pass before a change programme starts?

Haley sets out three. First, convince yourself the value is actually there and the business case is accurate. Second, understand what you have to influence in order to capture that value, which he calls the blast radius. Third, and hardest, check whether you are configured to influence those things at all. Many programmes depend on changes they have no authority over, so those dependencies become blockers. If the answer to any of the three is no, his view is that you should not start. If the answer is roughly yes, do more work first. Someone has to be accountable for all three being true.

What are the early signs that a change programme is only PowerPoint?

The clearest signal is universal agreement followed by nothing happening. Haley describes arriving at a client full of energy about a new plan, then being shown a folder by a department head. Inside were strategies dated 2016, 2018, 2020, and 2022, each written by a different third party, none of them implemented. Organisations with that history treat the next announcement as something to wait out. Ask in the first meeting whether anyone is genuinely up for it, or whether they expect it to wash over them. He also notes that the cultural side of change is usually the least funded part of any programme.

Why do change problems show up later in mid-market firms than in large enterprises?

Because the design phase is easy. Haley points out that mid-market leadership teams are often in one country, sometimes one building, so decisions get made quickly. There is little governance and usually no regulator, so a firm can convince itself it is making real progress. The pain arrives when teams have to execute, and the weekly cost of delivery rises sharply. That is when you find out whether teams understand who the outcome is for, and whether the team you have is the right one. Large corporates hit those problems throughout the process, so nothing surprises them late on.

Which leadership role is missing from most mid-market organisations?

Haley calls it the architect of the enterprise. It is the person who understands how business, technology, and operating model decisions fit together, and who to go to for detail. The usual assumption is that the CIO, the COO, and the people officer will coordinate between themselves and it will work out. He says someone needs to steward that conversation. Scope it carefully: this role coordinates and translates, it does not own the whole organisation. Many firms have a chief architect, but the remit rarely covers business architecture or organisation design, which normally sits with HR.

Should AI productivity savings be banked or reinvested?

Haley makes the case for reinvestment. The first wave of AI use cases is easy to reach, and the productivity gain is easy to build a case around. If that saving goes straight to the bottom line, the next layer of work loses its funding. Redesigning a process front to back, and building data foundations robust enough to feed agentic architecture, costs considerably more. You can no longer offset that cost against the first saving, so the case becomes very large and very hard to approve. His practical suggestion is to club the proofs of concept together and fund the foundations once.

What should a mid-market CEO ask before signing off an AI programme?

One question: can I articulate why my enterprise will be better optimised for the outcome I care about as a result of this programme. Haley is clear that this is not an AI test, it is a front to back enterprise test. The question that usually goes unasked is simpler still. What is the outcome we care about, and are we all aligned on it. He also flags accountability. If an agent does something the CEO could be held personally responsible for, the lines need to be drawn before deployment, not after.