Why AI Is Forcing a Reinvention of Product and Technology Leadership
By Pierre Bretin
Head of Executive Search, Ozone Executive Search
With contributions from Maria Chilikov and Gunes Italiaander
The SaaS model of selling seats, licences for humans to use tools, is undergoing a structural shift.
When the end user is no longer a person but an autonomous AI agent that can draft a contract, route it for approval, negotiate the redlines, and file the executed version without anyone touching a keyboard, the buyer stops asking what software do I need? and starts asking what outcome do I need?
As the whole category is being redrawn, it puts a question in front of every board and every tech leader that isn’t going away: not whether to adopt AI, but how to lead through a structural shift that will reshape teams, commercial models, and what software is worth.
At Ozone Executive Search, we sit inside this shift every day. Our mandates take us into product, technology, and commercial leadership conversations across industries from legal to healthcare, information services to financial infrastructure. The same themes keep surfacing. The leaders who will define the next era are not those with the most sophisticated models. They are those who understand how autonomous systems change what it means to build, sell, and lead.
To test and sharpen that view, I brought in two product leaders: Maria Chilikov has led AI-native product transformations across high-growth companies and thinks hard about where commercial strategy meets organisational design. Gunes Italiaander is a fractional CPO and product strategist who founded Product of Foresight to help companies figure out what AI-native actually looks like in their products, their teams, and their operating models.
This is Part 1 of a two-part series. Here we look at the structural shift: what is changing, who is accountable, and what it means to sell outcomes instead of software. In Part 2, we turn to the leadership response: governance as a weapon, the new product leader profile, and where value collapses and thrives.
The Agentic Shift
Start with the numbers. Most companies spend roughly 10% of operating expenditure on software. That is the addressable market SaaS has been fighting over for two decades.
Maria’s argument is that AI-native products blow open the other 90%:
AI-native products can now reach into the other 90% of OPEX: workflows that were too messy or too contextual to digitise before. Customers already expect products to act, not just inform. If your product surfaces data and waits for a human to decide, you’re behind.
— Maria Chilikov
Gunes sees the same shift from a different angle. She works across sectors where the incumbent’s moat has historically been curation and presentation. She calls what is happening the middle-layer collapse:
Enterprise stacks have always been built in layers – collect data, process it, display it, act on it. AI agents are swallowing those middle layers whole. If your product’s job is making information legible to humans – dashboards, reports, workflow tools – you’ve got a problem. AI agents don’t need legibility. They need access.
— Gunes Italiaander
The tools that survive, she argues, will own proprietary data or sit at the point of decision. Not the point of display.
We are already seeing this reshape the briefs we receive. Clients increasingly describe leaders who can think across product, engineering, and commercial strategy at the same time because the walls between those functions are coming down.
Who Is Accountable When AI Decides?
More autonomy means harder questions about where humans stay in the loop. Gunes, whose background is in neuroscience, brings a useful framework:
I use three variables with clients: reversibility, stakes, and interpretability. If a decision is easily undone, low-stakes, and you can see why the AI made it, automate fully. If it’s irreversible, high-stakes, or the reasoning is a black box, keep a human in the loop.
— Gunes Italiaander
Simple enough in principle. Harder in practice, because oversight itself has a cost most leaders underestimate:
If you’re asking a compliance officer to review five hundred AI decisions a day, you haven’t created oversight. You’ve created a rubber stamp.
— Gunes Italiaander
Maria’s take is more incremental. Start with humans everywhere. Step them back as confidence grows based on performance data, not assumption. But be clear-eyed about where they cannot step out:
The model is the same as line management. If a direct report makes a mistake, their manager takes responsibility. With AI it’s the same, just more extrapolated. For decisions that carry real consequences, a human needs to own the outcome regardless of whether AI generated the recommendation.
— Maria Chilikov
She also flags a measurement problem that is more serious than most organisations realise. She references Sol Rashidi’s Human Amplification Index:
Most organisations measure AI with the wrong metrics: hours saved, tasks automated, cost reduced. These measure productivity, not effectiveness. The better question is whether AI made the human better at their job. If your team is producing more of the wrong things faster, you’ve gained nothing. Maria Chilikov
The companies getting this right keep humans central in customer-facing communications, core IP, and people decisions. Where the cost of getting it wrong is relationship damage or regulatory exposure, humans stay. Where the cost is a wasted iteration, let the agent run.
Selling Outcomes, Not Software
Every strategy deck now talks about selling outcomes instead of tools. Actually doing it is a different thing.
Gunes calls the core difficulty the value attribution problem:
With a tool, the value exchange is clean – you provide the software, they do the work, they get the result. With an outcome, the value is co-created. Your product team needs to understand the customer’s business deeply enough to actually deliver on the promise.
— Gunes Italiaander
That means different metrics. Not DAUs or NPS. Time saved, decisions improved, revenue generated, risk reduced. It requires product managers who understand their customers’ businesses properly not just their UX flows.
Maria frames it even more tightly:
When code is cheap and fast, the bottleneck moves upstream. The hard part is knowing what to build and why. Every squad needs a clear commercial mandate: the result they’re driving and the number attached to it. When AI compresses build cycles, unclear priorities cost even more because you’re shipping the wrong things faster.
— Maria Chilikov
In our search practice, this shows up as a rising premium on candidates with operational or commercial backgrounds not just product craft. Leaders who cannot put a number on the value they create have, as Gunes puts it, a shelf life.
What AI-Native Actually Means
Most companies claiming to be AI-native are not. Gunes draws the line sharply:
It’s in the bones, not the skin. Bolting a chatbot onto your legacy platform doesn’t make you AI-native. It makes you AI-adjacent.
— Gunes Italiaander
She identifies three structural markers. Data treated as a product in its own right, not a byproduct of operations. Composable architecture that agents can orchestrate, rather than monolithic apps that need a human to click through. And learning loops every interaction, every outcome, every failure feeding back into the system.
Maria offers a complementary lens that is especially useful for assessing leadership:
First, AI for product engineering velocity - code velocity is up across the board, but feature velocity hasn’t kept pace because review and architecture are now the constraint. The value is in the thinking, not the typing. Second, AI for the broader organisation. Third, and this one gets lost most often, AI as core product strategy, how is AI used at the heart of the product to make it better for customers? Not a feature for the board deck.
— Maria Chilikov
■ Signal to watch: Maria notes that forward looking companies now track token usage per department as a proxy for AI adoption across the broader organisation. AI-native companies have all three baked in from day one. Everyone else tackles the first, dabbles in the second, and struggles with the third because it means rethinking the product. If the AI strategy is a separate workstream managed for optics, it does not matter how much you spend on models. You are AI-adjacent.
In Part 2: we turn to the leadership response, governance as a competitive weapon, what a future-ready product leader looks like, and where value collapses and thrives as AI reshapes SaaS.
About the Contributors
Pierre Bretin is Head of Executive Search at Ozone Executive Search, where he advises technology and product leaders on senior hiring across AI, SaaS, and digital transformation.
Maria Chilikov is a product and technology executive who has led AI-native product transformations across high-growth companies, with deep expertise in commercial product strategy and organisational design for the AI era.
Gunes Italiaander is CPTO at Helio Intelligence, a former fractional CPO and product strategist who founded Product of Foresight to help companies build in the agentic AI space. Her background spans product leadership, venture building, consulting, and innovation.