Why every AI deployment needs a Forward Deployed AI Strategist, not just an engineer
We have never talked so much about AI in business. We have never invested so much. And yet, we may have never wasted so much either.
This is the paradox of the particular moment we are living through: an unprecedented technological frenzy, driven by a very real pressure (not to miss the AI train), but one that produces, in the overwhelming majority of cases, disappointing results. Pilots that fail to scale. Budgets swallowed up by secondary use cases. Teams mobilized for months to end up with… very little.
The problem is not the technology. The technology is there, powerful and accessible. The problem is what happens, or rather what doesn’t happen, before we deploy it.
What the numbers say, and what we’d rather not hear
The data is unsparing. According to the report The GenAI Divide: State of AI in Business 2025, published by MIT’s NANDA initiative, 95% of enterprise generative AI pilots stall at the early stages — only 5% genuinely make it through the pilot phase to be integrated into real operations and produce a measurable financial impact. Gartner, for its part, estimated as early as July 2024 that at least 30% of GenAI projects would be abandoned after the proof-of-concept phase by end of 2025, due to poor data quality, uncontrolled costs, or insufficiently defined business value upfront. A figure its analysts have since revised upward for agentic AI projects.
The RAND Corporation, in a deep-dive analysis conducted with 65 experienced data scientists and engineers (The Root Causes of Failure for Artificial Intelligence Projects, 2024), identifies one dominant cause, present in virtually all failures: stakeholders don’t know, or can’t clearly articulate, what problem AI is supposed to solve. This diagnosis breaks down into five recurring factors, all non-technological: poorly defined business needs, overly ambitious or miscalibrated use cases, confusion between real organizational challenges and trendy offerings, insufficient data, and absent governance.
Five causes. Zero connection to model quality, infrastructure robustness, or engineer skill.
The conclusion is inescapable, however uncomfortable: AI failure in enterprise is rarely technological. It is almost always strategic.
FDEs: a remarkable role, built to answer a different question
To address this deployment challenge, the major AI players (Palantir as the pioneer, followed by OpenAI, Anthropic, Cohere, and many others) have popularized a model that has become indispensable: the Forward Deployed Engineer (FDE).
We ourselves, at Adobe, have moved in this direction and invested in our ability to deploy such experts at our clients’ sites.
And the principle is sound. Rather than delivering a tool and hoping the client figures it out, you send your best engineers directly to them. They immerse themselves in the client’s environment, understand its technical constraints, build custom workflows, and get solutions running in production. Palantir describes the role as similar to that of a startup CTO: full autonomy, high stakes, end-to-end execution.
The model works. Job postings for FDEs have exploded by more than 800% between 2024 and 2025. Compensation packages reach $350,000 to $550,000 annually at the major AI labs. It is, quite literally, the most sought-after role in the tech ecosystem right now.
But here’s what gets said less often: the FDE is designed to answer the how, not the why or the what first.
Their excellence is technical. Their value lies in implementation. They arrive at a client’s site with powerful tools and the ability to deploy them quickly, but they often arrive in an environment where no one has yet clearly established which problems deserve to be solved first, with what expected impact, and for which internal stakeholders.
In a context of frenzy where companies want, under pressure from boards, competitors, and the media, to “do something with AI,” the FDE ends up building fast, on strategic foundations that haven’t yet been laid. They do their best, with an incomplete brief. And often, the results suffer for it.
The real problem: the missing strategic jump
What I observe, and what many in the industry observe without yet naming clearly, is a gap between the promise of AI and its translation into real business value.
Most companies engaging in AI projects today have not done, or have not had the time to do, the prerequisite work of prioritization. They know they want, once again, to “do AI.” They don’t know precisely on what, in what order, with what expected return, or how to align their internal stakeholders around a shared vision.
This gap produces well-documented effects:
• POCs that don’t scale, because they were designed to impress, not to be operationalized;
• Misallocated budgets on secondary use cases, while the real value levers remain untapped;
• Disconnected business teams, who are subjected to a technology deployment they didn’t co-build;
• And ultimately, mutual frustration: clients who feel nothing is sticking, and vendors who can’t understand why their tools aren’t delivering the expected impact.
The question isn’t “is the tool good?” It’s “did we deploy the right tool, on the right problem, for the right people, at the right time?”
And to that question, the FDE, however brilliant, was never mandated to respond.
The Forward Deployed AI Strategist: defining a role that doesn’t yet exist
This is where I introduce an idea I believe is both necessary and urgent: we need to invent the Forward Deployed AI Strategist.
Not to replace the FDE, but to do what the FDE cannot, or should not, do alone.
The Forward Deployed AI Strategist is a hybrid profile, at the intersection of strategy consulting and product management. Not an engineer. Not a traditional consultant who delivers a report and disappears. Someone who immerses themselves in the client organization, speaks the language of the business, and has the ability to translate business challenges into AI investment decisions, before the first engineer is engaged.
Their typical profile: a business manager who has developed a genuine AI culture, or a hybrid tech + business profile able to navigate with ease between the executive boardroom and a working session with operational teams. Neither purely tech nor purely strategy, but both at once.
Their core responsibilities are four in number:
1. Identify and prioritize high-impact use cases
Not every possible use case, but the two or three that, if solved, create demonstrable business value and pave the way for lasting adoption. This is a work of listening, analysis, and rigorous prioritization.
2. Calculate potential ROI before deploying
Before a single line of code is written, the Strategist must be able to establish an impact estimate: how much time saved, how much cost reduced, how much revenue potentially generated? This economic modeling work is what transforms an intuition into a defensible investment decision.
3. Align business stakeholders
AI adoption cannot be decreed. It must be built by bringing in the right sponsors, managing resistance, and creating the conditions for genuine ownership by teams. The Strategist is the orchestrator of this alignment.
4. Drive change and adoption
Deploying a tool is 20% of the work. Getting it adopted is the remaining 80%. The Strategist doesn’t disappear after the project kickoff: they stay to ensure that the promised value materializes in real-world usage.
The framework: the four phases of a mission
How does this translate concretely into a client engagement? I propose a four-phase framework I call DPAH: Discover, Prioritize, Align, Hand over… and stay.
DISCOVER
The Strategist starts first. But not alone: ideally alongside the FDE, so that they too hear the needs, the blockers, the ambitions from the outset. The Strategist maps business challenges, conducts interviews with key stakeholders, understands existing processes, identifies unresolved pain points and latent opportunities. The goal: to emerge with a complete, documented picture of the business context. Not a list of desired features, but an understanding of the problems worth solving.
PRIORITIZE
From this diagnosis, the Strategist builds an impact/feasibility matrix. Each potential use case is evaluated along two dimensions: the business value it can generate, and the complexity of its implementation. This phase produces a prioritized backlog, ranked and annotated with an ROI estimate for each retained case. This is the roadmap the FDE will receive to work from. Their role will also be to assess the technical complexity of the different use cases.
ALIGN
Before deploying anything, decision-makers must be aligned. The Strategist, still alongside the FDE, organizes validation sessions, manages trade-offs, and brings business and IT sponsors together around a shared vision and a set of common KPIs. Without this step, even the best technical deployment risks running into insurmountable organizational resistance.
HAND OVER… AND STAY
This is where the dynamic becomes truly distinctive. The Strategist doesn’t step back and leave the FDE to run the project alone. They stay, and the two work together, as a continuous duo, throughout the deployment. There is an inflection point here, however: this is where the FDE takes the lead. But the Strategist doesn’t disappear, and I’ll come back to that in a moment.
The Duo: 1 + 1 = 3
As I began to outline above, the Forward Deployed AI Strategist and the Forward Deployed Engineer are not in competition. They don’t substitute for one another. Nor do they work in silos.
They form a complementary duo, whose strength lies precisely in the creative tension between their two logics.
The Strategist opens the way. They set the context, qualify the problems, prioritize what matters, and prepare the ground so that the FDE arrives with a clear target, not lost in a fog of vague intentions. But once the FDE is on-site and deployment begins, the Strategist doesn’t fade away.
They remain present for one essential reason: to prevent the project from drifting toward pure technical performance, losing sight of the initial business objective.
This is a real and frequent risk. A skilled engineer, immersed in the complexity of a deployment, will naturally be drawn toward technically interesting problems. A more elegant architecture. A model optimization. A more sophisticated integration. These are good things, but not if they come at the expense of delivering the concrete business value the client was expecting.
The Strategist is the guardian of that value. At every stage of deployment, they ask the uncomfortable question: “Is what we’re building still addressing the problem we identified at the start? Is it going to create the impact we promised?”
This duo resembles, for those who lived through that revolution (the word may be a touch strong ;-)), what the Product Manager / Tech Lead tandem produced in product development ten years ago. Alone, the Tech Lead builds brilliant things that don’t always match the need. Alone, the PM defines needs that no one knows how to implement with judgment. Together, they create products that work and deliver value. Enterprise AI needs the same balance.
Who can play this role?
The profile question is critical, because this role doesn’t yet formally exist and it can’t be filled by just anyone.
Two archetypes seem most promising to me.
The first is the business manager who has converted to AI: someone who has spent years in an operational function (finance, supply chain, HR, sales) and who has developed a solid understanding of AI tools and capabilities. They naturally speak the language of the business, they know what “creating value” means in a real operational context, and they have learned to assess what AI can do and what it cannot.
The second is the hybrid tech + business profile: someone with a technical background or experience, who has chosen to develop expertise in strategy, transformation, or product management. They can hold a technical conversation with an engineer, and a business conversation with a sales director, and above all, they can bridge the two.
What this role is not: a traditional strategy consultant who delivers a PowerPoint deck and moves on. What fundamentally differentiates it from traditional consulting is the grounding in execution, the sustained presence at the client’s site, and accountability for impact, not just for the quality of the recommendations. It’s also this mixed dimension that allows them, alongside the FDE, to engage with all business functions and decision-makers, to listen and truly understand them.
Why now?
One might ask: why hasn’t this role emerged naturally? Why haven’t companies created it themselves?
The answer lies largely in the dynamics of this moment. The frenzy, the excitement, and the impatience around AI creates pressure to move fast, to show results, to not appear behind. In this context, decision-makers call on FDEs not because they have a clear plan, but precisely because they don’t, and the FDE gives them the feeling of doing something. It’s reassuring. It’s visible. And it’s often insufficient. It’s also because this is the profile most commonly put forward by the majority of AI labs: they sell technology, they want to ensure it gets deployed and used. And they often believe the technology speaks for itself.
It’s also worth noting that even Palantir, the pioneer of the FDE model, has felt this need. The company works with profiles it calls “Deployment Strategists,” specialists tasked with bridging technology and clients’ operational priorities. This isn’t exactly the Forward Deployed AI Strategist as I describe it here, but it is an acknowledgment by the very creator of the model that a purely technical profile is not enough.
The market will catch up with this reality. The question isn’t whether this role will emerge: it’s when, and who will be the first to formalize it.
Conclusion: create the role before the market does it for you
If you are a business leader investing in AI, the question to ask yourself is therefore not: “do we have the right tools?” or even “do we have the right engineers?” The question is: “Do we have someone whose role is to ensure we are deploying the right things, in the right order, with the right impact?”
If you are a consultant, product manager, or business manager with an appetite for AI, this role may be yours to invent or to claim.
If you are an FDE, what I’m describing is not a threat. It’s the profile that will finally allow you to work with clear briefs, defined priorities, and a business context solid enough for your technical talent to produce the impact it deserves.
The AI revolution is not short of engineers. It is short of strategists capable of giving it direction.
It is time to create this role.
This article was written in French and English. If this topic resonates with you, I’ll love to discuss it, whether you’re a business leader, a consultant, or an AI professional in the process of repositioning yourself.
Olivier Binisti — Western Europe Manager, Digital Strategy Group, Adobe
The Missing Link in Enterprise AI was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.