Interview with Anaïs Ghelfi
Agentic enterprises are changing how companies work by turning employees into builders, unlocking knowledge across teams, and scaling smarter collaboration and innovation.
The rise of agentic systems is redefining how organizations structure work, manage knowledge, and scale decision-making across teams. Anaïs Ghelfi, VP of Platform and Agentic Systems at Malt, is helping lead Malt’s transformation into a fully agentic enterprise through the design of scalable infrastructure that codifies knowledge and makes it widely accessible. This interview discusses her perspective on building agentic enterprises, the challenges of operationalizing organizational knowledge, and the future of work in AI-driven organizations.
In your current role as VP of Platform and Agentic Systems, what does your work involve day to day, and what excites you most about building in this Space?
What excites me most? Everything is so new. There are so many things to try, to learn, to figure out for the first time. Changes of this magnitude happen once in a career. Not just watching, but actually shaping it from the inside, is something I find genuinely thrilling. We don’t have all the answers yet, and honestly, that’s exactly what makes it exciting.
My mission is to build the agentic infrastructure that turns the company into one where data, know-how, and playbooks are codified and accessible to every employee and every agent.
In this new role, my mission is to build the agentic infrastructure that turns the company into one where data, know-how, and playbooks are codified and accessible to every employee and every agent. The role brings together what used to sit in separate worlds: backend platform, frontend platform, developer experience, and data platform because with AI, developer experience can’t just be about engineers anymore. It has to extend to everyone building applications and automations.
Day-to-day, that means a mix of time with the team to create space to experiment, time with business counterparts because this is as much a company transformation as a tech one, and time looking outward at what others are building. In a space moving this fast, a vision you set six months ago might already be outdated, and that’s part of what makes the role so different from any I’ve had before.
Before this role, you led global data work at Pernod Ricard. What key lessons from that experience still shape how you approach building systems today?
Pernod Ricard and Malt are very different environments, but both present equally exciting challenges. And the biggest lesson I took from Pernod Ricard applies just as much here: building the best system with the best technology is completely useless if nobody uses it, or if it doesn’t bring real value to the people it was built for. That principle is true whether you’re building inside a global multinational or a fast-moving scale-up.
What it means in practice is that technology is never the point. What matters is whether the people you’re building for actually use it, and whether it genuinely helps them do their work better. That sounds obvious, but it’s where most projects fail. Teams get excited by the system they’re building, and lose sight of the humans on the other side.
So the questions I keep coming back to are simple. Who is going to use this? What problem does it actually solve for them? What friction might it create? Where will it sit in their day? These questions sound basic, but they should be asked by anyone building anything: not just product teams.
That’s also why iteration matters so much. You start with a first version covering one use case, put it in front of real users, talk to them, learn, and improve. You build step by step, always keeping the people who’ll use it in mind.
The idea of an “agentic enterprise” is still quite new. In simple terms, how would you explain what this looks like inside a company in practical terms?
It sounds complex, but it’s actually simple. In an agentic enterprise, everyone becomes a builder. Expertise that used to stay locked inside specific teams becomes leverage for the entire company. It’s a structural shift in how organisations operate and create value.
Think about how a company operates today. Hundreds of people across different divisions, all working toward the same goal, but the knowledge of how things actually get done lives in people’s heads, and nowhere else. New joiners spend months catching up. Teams work in silos. And too much of the working day is spent on tasks that pull people away from where they create the most value.
The agentic enterprise changes that. The knowledge, the how-to, the processes, the team missions are codified and made accessible to everyone, including agents. You then build AI agents to take on entire workflows on your behalf. You give them the context, the tools, the goals, and they operate. They execute, they progress, and they come back to you only when they need a decision or your validation. Your time shifts to what actually requires you: strategy, expertise, judgment, redesigning the way things work. A sales person spends more time building long-term client relationships. A marketing team focuses on narrative, brand, and strategy. Agents take care of the rest.
But the deeper shift goes well beyond this. When knowledge is codified and accessible, the barriers between teams disappear: someone with an idea for a new revenue line no longer needs to pitch it and hope; they can use agents and the company’s codified know-how to prototype it themselves, respecting every design, product, and brand guideline along the way. That’s how expertise stops being locked inside teams and becomes leverage for the entire company.
When work moves beyond tools that simply support people to systems that actively shape how work gets done, what changes most in how teams collaborate and make decisions?
The way I see it, agents become collaborators, not tools you use, but teammates you work with. Just like you’d onboard a new colleague, brief them, give them context and check in on their work, you’ll do the same with agents. Some people will naturally take on the role of managing these agents: ensuring consistency, keeping the knowledge they operate on accurate, and making sure they’re aligned with how the team evolves.
This also means everyone needs to think differently about how they work. You can’t just delegate to an agent and walk away. You have to be intentional about what you codify, because the agent will follow it precisely, including your blind spots. Documentation becomes a critical, and core part of the job.
Decision-making shifts too. Strategy, priorities, changing direction: those stay human. But for everything that’s been codified, the agent operates autonomously and escalates only when something falls outside its scope. The result: people spend more of their time where their expertise and judgment actually matter.
Turning knowledge and ways of working into something that can be shared and reused across anorganisationsounds powerful, but also challenging.
Before you can build anything, you have to surface and structure knowledge that has never been written down. Doing that at scale, across every department, is genuinely hard.
The knowledge, the company’s context, is the real foundation of the agentic enterprise. It’s what every agent will rely on to navigate the organisation. Without it, nothing else works.
The hardest part is that most of this knowledge is scattered, or doesn’t exist anywhere at all: it lives in people’s heads. To get there, we’ll test different approaches. One idea we’re exploring is AI interview agents that ask the right questions and turn the answers into structured, usable knowledge. Asking people to document everything has never worked, and never will. Letting conversation do the work, with the system capturing and organising it, removes that friction and gives us a real foundation to build on.
But capturing the knowledge is only half the problem. There’s also an organisational challenge: you need dedicated owners in each team, responsible for keeping that knowledge accurate and up to date. These are new roles. At Malt, we already have AI Builders across every team, responsible for building AI agents for their functions. Tomorrow, those same people will be the ones ensuring the expertise of their function is codified and I’m convinced this role will become a standard across every company. Because that knowledge, the company’s context, is the real foundation of the agentic enterprise. It’s what every agent will rely on to navigate the organisation. Without it, nothing else works.
For leaders trying to build more connected and adaptable organisations, what mindset shifts are needed to move away from traditional ways of working?
The first shift is structural. At Malt, we’ve created a network of AI Builders & AI Champions embedded in every team; finance, sales, marketing, support. They’re not engineers. They’re the people closest to the work, building the agents that change how their team operates. That’s the shift: AI capability stops being centralised, and becomes distributed across every function. Most companies are still treating AI as a tech project owned by a tech team. That model won’t scale.
The second shift is to stop running AI like a traditional transformation program. Too many leaders are launching AI initiatives the way they launched digital transformation ten years ago: with steering committees, roadmaps, and long planning cycles. That approach will fail. The companies that will transform successfully won’t be the ones with the best plan. They’ll be the ones where everyone is empowered to experiment. The starting question shouldn’t be “how do we do this?”, it should be “what can AI do here, and what can it amplify?”
The third shift is getting comfortable with uncertainty. Nobody knows exactly what tomorrow’s ways of working will look like, we’re figuring it out as we go. The only way forward is to put AI in the hands of your teams, get use cases into production, and create the space to experiment. Clarity comes from doing, not from planning.
How do you see the relationship between people, knowledge, and intelligent systems evolving insideorganisationsover the next few years?
The real shift is this: as agents handle more of the execution, people get to focus on where their expertise and judgment actually matter, and on the strategy that only they can shape.
Expertise will matter more than ever. Experts are the ones who codify the knowledge that everything else runs on. The company narrative, the tone of voice, the go-to-market strategy, those remain human decisions, designed by the teams who live them every day. Builders then create agents to scale those teams on top of that codified expertise.
And progressively, everyone becomes a builder. If you have an idea, you’ll have the tools, the context, and the know-how to prototype it yourself. That removes barriers, spreads capability across functions, and gives real leverage to people who never had it before. A marketer can build. A finance analyst can build. A marketing manager can build. The question stops being “who can do this for me?” and becomes “what do I want to build?”
The companies that start operating this way early will build a structural advantage that’s very hard to catch up with. Not because they’ll have better technology, everyone will have access to the same models. But because they’ll have a different organisation. One where knowledge is centralized and accessible, expertise scales, and every person is amplified. That’s the real differentiator.
