Frederic Queudret spent fifteen years as a CTO before moving into private equity, most recently leading AI strategy at Acolad through the disruption that generative AI brought to the localisation industry. He now works as a Tech & AI Executive Advisor to several private equity firms, supporting deal teams through technology due diligence and then works alongside portfolio CEOs and CTOs to build and deliver AI value creation plans.
We spoke to him at our recent AI Forum about what actually moves the needle on AI transformation, why so few use cases reach production, and why he believes the biggest opportunity of the next decade sits with CHROs rather than CTOs.
Your remit spans both sides of a deal. How does that work in practice?
There are two aspects. I help the deal teams during technology due diligence before the acquisition, and once we have made the investment, I become the sparring partner for the CTO in place. That involves working with recruitment partners when we need to bring new leadership in. I work across the portfolios of several funds, with a subset of those companies intensively enough to mean weekly interaction.
Most of what I do sits in the growth phase. We generally take a local champion and do a lot of M&A over the following five years to build a company that is ready for the Champions League, ready to play worldwide. The thesis is growth rather than cost reduction, but growth still requires transformation, and AI is now central to that.
You work across several funds rather than from inside one. What does that vantage point give you?
Pattern replication, above all. Portfolios run from manufacturing to vertical software, and you would be amazed how often something that works in one company transfers to another. That is especially true with AI transformation. Sitting across several portfolios rather than one makes the sample bigger and the patterns clearer, faster. When an experiment works somewhere, the question is immediately whether the same model transfers elsewhere.
It also puts the specialist versus generalist debate in its place. A pure software investor knows the metrics of one industry better than anyone, while a generalist sees patterns that a software business would never meet inside its own vertical. Neither is the better model, they are different bets with different blind spots. What a company picking a partner for the next five years responds to is evidence that you have solved its problem before, somewhere else.
How do you frame the AI question when you look at an asset?
Not as a single threat vector. We look at it across the whole equity story rather than at revenue alone. Can AI lift revenue, can it improve margin, can it compress the time to value between a customer request and the feature reaching the end user, can it improve software quality, can it strengthen or create a moat.
I keep a list of around a hundred KPIs, and I can assure you I am not using a hundred KPIs every day. Each plan uses a subset mapped to that company’s investment thesis, so we understand where the business is coming from, where it wants to go, and what the plan is to get the value out. There is no single value creation plan you can lift and reuse.
What I would say is that no company stays protected indefinitely. A software business that does not move will eventually find that its market has moved without it.
What have you learned about what makes AI transformation actually happen?
Three things stand out. The first is that it always starts with the CEO. Every time I see real momentum, there is visible sponsorship at executive level and at company level. I have seen CEOs build their own town hall material with AI and then talk the whole company through it. At a CEO forum earlier this year, a large part of the agenda was making the point that if they want to trigger an AI dynamic inside their business, it must start with them. Several of them went back and started.
The second is that the best use cases are not handed down. Leadership brings the strategic priorities, but a lot of the ideas that survive contact with reality come up from the field, from the people who are living the pain point.
The third is the community effect, which is why a forum like this one matters. Deploying the technology is quick. Any CIO can roll out AI tooling across a company in short order. Getting people to work in an AI first way is something else entirely. It is not a technical challenge, it is a change management challenge, and peer learning between teams sharing what has actually worked for them creates the conditions for adoption far more effectively than a training programme does.
You are notably unimpressed by long lists of use cases. Why?
Because we have all seen the presentation where a team says they have ten AI use cases, and then nobody can tell you the ROI. What is the cost associated with it? We do not know, ask us in six months. That is a bet, not a plan.
Realistically, out of a hundred use cases you might see five to ten reach production at company level. So we are not trying to deliver a hundred. We are trying to find the ones with the best success ratio given the effort and the investment, and we score them on risk, on the degree of autonomy the AI has, on data sources, on readiness for production. We also make sure businesses understand the unit economics before they commit. If you are going to offer a copilot experience at twenty or thirty euros per user per month, you need the pricing and packaging worked out, otherwise you are bleeding money through uncontrolled usage.
We start every plan from strategy and we finish every plan with ROI.
What is changing inside software teams?
Quite a lot, and quickly. I spend most of my time in software, and I can tell you that job specifications are no longer published without a prerequisite around agentic development. Experience with coding agents, familiarity with an agentic software development lifecycle, automated pull request review with an agent in the CI/CD pipeline. That has become standard language in the last year.
The bigger shift is spec driven development and rapid prototyping. Traditionally a product owner produces a specification after three weeks, or a month, or four months, and then the developers challenge it, the requirements are unclear, and you go back around the loop. Now you can build a working prototype in a day. It is not going into production and it is not going into any branch of any repository, but it is a living artefact you can experiment with and show to end users. That is a genuinely different way for product and engineering to work together, and it takes real time out of the cycle.
Where does AI ownership sit in the companies you work with?
This is one of the most common failure points. The CEO understands that sponsorship starts with them, but at the operational level you need someone whose day job is the AI plan and who is accountable for it progressing.
The pattern I see repeatedly is a company six months in where the tools are deployed, everybody has access, there is an assistant in production, and nothing is really moving. Execution requires IT, engineering, revenue and marketing to collaborate, and nobody owns it. Everyone can prompt, which is fine for individual productivity, but it is not transformation. So I challenge leadership teams on whether they need an AI transformation leader, in the way we needed a Chief Digital Officer ten or fifteen years ago. Or a Chief AI Officer. I wrote a piece on LinkedIn about the five flavours of Chief AI Officer, which is amusing given we spent thirty years debating the flavours of the CTO.
Finding those people is hard. There are a lot of pretenders, people who present as doing a great deal with AI until you go into the detail. And this is not the same profile as a data scientist or a head of data. I have worked with AI and data people who genuinely did not recognise iterative delivery, who would say it will take six months to train the model. Meanwhile competitors are putting AI in their roadmaps and shipping value now. People who have worked in a protected, isolated way are entering a world where they have to demonstrate and deliver value, and understanding the difference between running a project and delivering a product is not a small adjustment. The default response across Europe is to subcontract to a large professional services firm, which is expensive and slow when what you need is a small team assembled quickly with the right mindset. Accelerators, not blockers.
What keeps you awake at night?
Change management, particularly inside software teams. There is a wide gap between the early adopters and the people at the back who quietly believe AI is going to replace them. It is a daily effort. Every day you are convincing an engineering manager, every day you are re-challenging the mindset. Plenty of people are comfortable with AI without having an AI first approach, and they reach for it to speed up their existing habits. I will code with an assistant and get ten per cent more productive. That is not the destination. The point is to rethink the way the work is done, and that is much harder.
I say this as someone who has been writing software since I was ten, in the eighties. I use AI constantly, I pay for my own subscriptions on top of anything work provides because the return is obvious to me, and I have built ten pieces of software in the last three months, including rebuilding an entire CRM. Nothing I start now, I start without it. Once you have made that shift, most of your day runs through it, and you cannot go back. The day your people are taking advantage of AI in the large majority of their work, you have won the change management argument.
What does that mean for employers competing for talent?
It becomes an employer brand question very quickly. Put two offers in front of a strong candidate, one from a company leading on AI and one from a company that is lagging, and I can tell you which one they take. If someone has made that mindset shift, they are not going somewhere that takes the tools away. So you will lose your high potentials if you do nothing, and you need to transform in order to both attract and retain them. Companies need to be communicating on this, not just deploying it.
There is also a responsibility to bridge the gap between how the current workforce operates and the AI native people arriving on the market in the next three to five years. I have children who will never do the job I did in the eighties. Companies have perhaps eighteen months to three years to get ready for that. Change management is the number one priority and the most underestimated challenge, because there is technology time and there is human time, and you cannot fight human time. Bad habits are not fixed by AI. They get replicated, and they get worse, because people apply it to the way they already work rather than changing the way they work.
You have described this as an opportunity for CHROs rather than a threat. Why?
Because it may be the biggest rethink of how a company is designed that we have seen in fifty years, and it lands squarely with HR.
Take assessment. As a CTO I used to recruit sixty or seventy per cent on technical skills and thirty or forty per cent on soft skills. I always wanted a good developer who could also communicate and articulate an argument. With AI capable of writing better code than almost anyone, is this not the moment to invert that, to assess thirty per cent on technical skill and seventy per cent on judgment, communication and adaptability?
I would go further and say the companies that have stopped recruiting apprentices are making a mistake. You have senior people who can exercise judgment, but you have to train young people using AI up to that level, otherwise you have no bench. Some organisations have gone the other way and increased apprentice intake significantly, which I think is the right call.
Some of the strongest hires I have made came from outside engineering entirely, including people with humanities backgrounds who went on to lead technical functions. In every case it was engagement, mindset and involvement that made the difference rather than the technical grounding, and AI widens the range of people for whom that is now true. Product owners are building working software. Developers are writing better specifications. There is an opportunity for people to be far more versatile than their job title has historically allowed.
That is the wider point. AI is a prediction engine. I studied it in the nineties and there is no magic in it. You give it context, it predicts the next thing based on what it was trained on, and the more context you give it the more accurate that prediction becomes. What it does not have is creativity. Judgment, creativity, and human relationships are where people still hold the advantage, and they are what education and organisations should be developing much more deliberately. So the opportunity for any CHRO is to redesign the company as a hybrid one, where AI and people work together, and to decide very deliberately where you want human energy to go.
A big thank you to Frederic Queudret for sharing his time and insights with us. Heads Up is our interview series with the leaders shaping how businesses are built, backed and scaled.