The scale of the bet
Microsoft is committing 2.5 billion dollars to hire roughly 6,000 forward deployed engineers, staff embedded directly with enterprise customers to connect AI rollouts to actual business processes. AWS has put in a billion dollars of its own, and Google, OpenAI, Anthropic, Accenture, and Deloitte are all building comparable programs. This is not a marginal experiment. It is a coordinated, multi-billion dollar wager by the largest AI vendors and services firms that the binding constraint on enterprise AI adoption is integration talent, not model capability.
By the end of 2026, more than 85 percent of tech providers are expected to have launched some version of a forward deployed engineer program. That level of convergence across competing vendors is itself informative. When every major player in a market independently arrives at the same staffing model within the same year, it usually means they are all responding to the same signal from customers, in this case a demand pattern strong enough to justify billions in hiring even amid an otherwise cautious spending environment.
Why the gap opened up in the first place
Gartner senior director analyst Alex Coqueiro frames the underlying problem starkly: AI is not just one more technology, it will actually change the way people work, and Gartner's own research suggests as many as seven in ten enterprises may need to abandon agentic AI projects because of internal skill gaps. That is a far higher failure rate than typical enterprise software rollouts, and it points to a different category of problem than a missing feature or an unclear roadmap. The skills required to connect an agentic AI system to a real business process, spanning data engineering, workflow design, and change management, simply do not exist in most internal IT organizations at the depth needed.
IDC's Jennifer Hamel, research VP for enterprise data and AI services, captures the shift in framing: the industry is moving beyond the period of experimentation. That transition from experimentation to execution is exactly where forward deployed engineers are positioned to operate. Pilots can survive with a small, dedicated team improvising as they go. Production rollouts at enterprise scale cannot, and that gap between pilot-stage and production-stage skill requirements is what vendors are now racing to fill with embedded staff.
What this looks like from inside a large enterprise
Mojgan Lefebvre, EVP and chief technology officer at Travelers Insurance, described the internal version of this same problem: for the cross-functional agile teams solving specific business problems, the company wants to make sure they all have AI expertise, which is a harder staffing target than it sounds because that expertise is scarce and every team wants a piece of it. Andrew Palmer, EVP and CIO at Liberty Mutual, was even more direct about the resulting internal competition: it is very hard to compete for that attention, describing the scramble across business units for a limited pool of AI-literate staff.
Both executives are describing the same dynamic from the buyer side that vendors are responding to on the supply side. Internal teams cannot hire or train fast enough to meet demand for AI-literate staff across every business unit that wants an agentic AI initiative. Forward deployed engineers are, in effect, a rental model for exactly that scarce capability, letting enterprises borrow integration expertise from a vendor rather than compete internally for a headcount allocation that will never stretch far enough.
The build versus rent decision this forces
For a CIO, the rise of forward deployed engineer programs changes the calculus on whether to build internal AI integration capability or rent it from a vendor. Building internally is the more durable long-term answer, but Coqueiro's seven-in-ten failure rate suggests most organizations do not have the runway to build that capability before their current agentic AI initiatives stall out. Renting forward deployed talent buys time, but it also means the institutional knowledge of how a given AI system was actually wired into the business process lives with the vendor's staff, not the enterprise's own team.
The organizations getting this right appear to be treating forward deployed engineers as a bridge rather than a permanent arrangement, using the embedded expertise to ship a first wave of production use cases while simultaneously building internal capability to take over maintenance and the next wave of initiatives. That is a harder discipline to maintain than it sounds, because once a forward deployed team is delivering results, the organizational incentive to invest in the slower, more expensive internal build tends to fade.
What CIOs should watch next
The billions committed by Microsoft and AWS alone suggest this market is not a temporary staffing fix but a durable new category of enterprise services, one that will show up as a recurring line item in AI budgets going forward. CIOs negotiating these arrangements should press vendors on knowledge transfer terms up front, not after the engagement, since the value of a forward deployed engineer decays sharply once the enterprise cannot operate what was built without ongoing vendor support.
The deeper signal in this story is what it says about where the enterprise AI market actually is in its maturity curve. If the biggest constraint required a multi-billion dollar, cross-vendor staffing response, the technology itself has clearly outpaced the organizational capacity to deploy it. That gap is the real story behind every pilot that never ships, and forward deployed engineers are the industry's most expensive acknowledgment yet that closing it will not happen through better software alone.
How to negotiate this without losing leverage
CIOs approaching a forward deployed engineer engagement should treat it as a staffing negotiation with the same rigor applied to any systems integrator contract, not as a free bundled service that happens to come with a software license. Ask which specific deliverables the embedded team owns, what documentation they are contractually required to leave behind, and how quickly internal staff can be paired alongside them to absorb the integration knowledge as it is built rather than after the engagement ends and the vendor's engineers move to the next account.
The enterprises likely to get the most durable value from this wave of vendor investment are the ones that use forward deployed engineers to compress the time to a first production win while simultaneously running their own hiring and training pipeline in parallel. Waiting for the skills gap to close organically before starting an agentic AI initiative risks falling behind competitors willing to rent the capability now. Treating the rented capability as a permanent substitute for internal skill building risks a much longer-term dependency that is far more expensive to unwind.
