“Do not wait until you have the final answers. Start now, act now, and iterate.”

Hervé Couturier has spent four decades at the intersection of enterprise software and leadership, as CTO and CPO at IBM, SAP, Amadeus, and Business Objects, and now as a board member and senior advisor to PE funds including Advent, Bridgepoint, Goldman Sachs, and KKR. We sat down with him to talk about where AI adoption actually is, what it means for technical leadership, and why the companies winning right now are the ones willing to move before they have all the answers.

You have been working across enterprise software and PE portfolios for decades. What is your headline read on where AI adoption actually is right now?

My view is very simple. Most organisations are sitting somewhere between exploration and tool selection. They have intent. They lack a roadmap. And there is a real gap between how companies are talking about AI and where they genuinely are.

The companies making real progress are the ones that have moved AI from an initiative owned by engineering to a company-wide agenda tracked at board level. The ones that are stalling tend to be managing it as a project rather than a transformation.

I would describe four stages of AI maturity: exploration, prototyping, implementation, and scaling. Most companies today are between the first two. My one piece of advice to any leadership team is this – do not wait until you have the final answers, because it is changing too fast. Start now, act now, and iterate. Otherwise, you will fall behind.

There has been a lot of talk about AI reducing headcount, particularly in R&D. Is that what you are seeing?

No, and I think people fundamentally misread this. AI is not killing software. First of all, because AI is software. It creates software, and it is built on software.

What I have seen consistently is that productivity gains in R&D – and we are talking three to ten times improvement – do not lead to reduction in force. They lead to increased demand. I work with a retail technology company in Italy, with around 100 million euros in revenue, whose customers include Nestlé and Coca-Cola. They do trade promotion management – helping clients decide where to allocate advertising spend to maximise revenue. In the past, they could run one simulation per day. Now they run one per minute.

You might think: great, let us cut the team. What actually happened is that customers immediately asked for more. More complex inputs, more data, more variables. The demand simply expanded to fill the capacity. That pattern is repeating everywhere.

The initial wave of companies thinking they would fire ninety percent of their R&D? It is not happening. Reductions I do see are linked to broader enterprise performance, not to R&D efficiency.

“Customers immediately asked for more. The demand expanded to fill the capacity.”

Is software actually at risk?

There will be a SaaSpocalypse – but only for a specific group of companies. Tony Leng, Managing Partner,  H.I.E.C  put it well: the ones at risk are the superficial, high-level businesses that simply grab information and make it look pretty. The ones that will survive and thrive are those with deep contextual understanding, businesses that own their processes and data, and those converting from a SaaS model to an AI-enabled platform where agents can take action.

If you stay within the application you have been developing for the past twenty years and nothing more, then you are finished. But if you move to offering your customers a platform, pre-trained agents, orchestration, context servers – you are offering something fundamentally different. The question is whether you answer that challenge before your customers ask it.

What does that mean for how PE funds are thinking about their portfolios right now?

There has definitely been a pause. Some valuations have been massively corrected. Funds bought high and do not want to sell low, so there is a degree of waiting – watching how their portfolio companies respond to the shift before committing to new investments or exits.

What has not slowed down is hiring. I hear it constantly across the funds I advise: we have a CTO who cannot drive this change, we have a CPO who is not capable. There is real urgency around finding the right technical leadership. That, I think, is where the real activity is right now.

Longer term, I am seeing PE increasingly active in energy management, energy optimisation, and the intersection of hardware and software. AI is a massive consumer of energy – a single prompt consumes between ten and one hundred times more energy than a standard search query. Everything related to energy infrastructure is becoming increasingly important as a result.

Where do we find strong AI talent, and how do we assess it?

Honestly, I do not have a simple answer to where you find them. There is not enough seniority in the market yet to point to obvious sources the way you might have done historically.

What I can say is that the startup world is probably your best hunting ground. People who have had to move fast, make decisions with incomplete information, and build governance instincts under real pressure rather than in a structured enterprise environment.

On assessment, the question is not whether someone can recite an AI strategy. It is whether they raise the right issues spontaneously. The person you want as CTO or CPO should tell you, without prompting, that their main challenge will be orchestrating autonomous agents – that without proper governance you end up with cowboys, very powerful cowboys, doing unpredictable things in production. They should raise IP ownership questions when agents create code. They should bring up security, compliance, and transparency without being asked.

That instinct – to identify those challenges spontaneously – is the mark of someone who genuinely understands what AI deployment involves at scale. And crucially, the right person is not looking for the right answers. They are asking the right questions.

“The right person is not looking for the right answers. They are asking the right questions.”

What about the “ten years of AI experience” requirement that keeps appearing in briefs?

It simply does not exist. Nobody has ten years of relevant AI experience. The maximum you will find is two years. And wanting a senior developer without being willing to consider someone junior – well, how do you become senior without first being junior? That is the mystery.

Rather than looking for specific experience, look for creativity, innovation, and a genuine appetite to embrace AI. Look for people who can articulate what a solid AI strategy looks like, who understand the shift from static applications to autonomous, event-driven, action-oriented systems, and who can speak to orchestration and governance as naturally as they speak to product.

Where does AI maturity show up at board level? What should boards actually be tracking?

Most CEOs I work with are not yet thinking in KPIs. They are still at the stage of: I absolutely must do something here. The level of maturity at CEO level is not that great yet.

At the engineering level, we are starting to see more rigour – traditional DORA metrics, for instance, where deployment time has moved from weeks to days or hours. But the new metrics boards should be asking for are different: what percentage of your R&D population is using AI-assisted coding? What percentage of product management is using AI in specification and design? Which functions are falling behind?

And there is a maturity ladder beyond that. Are people using AI as a better user interface – stage one? Or are they using it for simulation and scenario modelling? Or to automate existing workflows? Or, the most advanced stage, to create genuinely dynamic, autonomous workflows? As a board, that is what I would want to know. Where are we on that journey, honestly?

“The one constraint I would remove is the idea that AI is an engineering thing. It touches everyone. Every function. The entire chain.”

If you had to remove one constraint across the organisations you advise, what would it be?

The appetite across the entire organisation to use AI. Getting it away from the idea that it is an engineering thing or a coding thing. It is an enterprise-wide challenge.

If you do not approach it that way, you simply move bottlenecks. You end up with a highly efficient coding machine on one end and everything else is unable to absorb the output. The question of where to apply that efficiency, how to monetise it, how to invoice customers, how to stay compliant – none of that gets answered if you treat AI as a technical project rather than a business transformation.

Hervé Couturier is a board member and senior PE advisor with over four decades of experience in enterprise software and technology leadership. He has served as CTO and CPO at IBM, SAP, Amadeus, and Business Objects, and currently advises funds including Advent, Bridgepoint, Goldman Sachs, and KKR across companies ranging from early-stage startups to €4bn in revenue.

Tim Chamberlain is a Managing Partner at H.I. Executive Consulting (H.I.E.C) and leads the Technology & Software practice. This conversation was part of an internal H.I.E.C Q&A session.