In May 2026, the competitive landscape of the AI industry shifted from raw model development to enterprise-level on-site deployment. OpenAI, Anthropic, and Google Cloud all adjusted their commercial strategies to prioritize the deployment phase by focusing on a specific role: the Forward Deployed Engineer (FDE).
OpenAI announced the establishment of OpenAI Deployment Company (internally codenamed DeployCo) on May 11, 2026, with a $4 billion initial investment. As part of this rollout, it acquired Tomoro, an AI consulting firm bringing in around 150 deployment specialists. Anthropic partnered with financial technology provider FIS to embed its applied AI team and FDEs on-site, developing a Financial Crimes AI Agent to compress anti-money laundering (AML) investigations for BMO (Bank of Montreal) and Amalgamated Bank from hours to minutes. Simultaneously, Google Cloud initiated a hiring campaign for hundreds of FDEs to support its practical engineering teams.
Contrasting FDEs with Traditional Technical Roles
The FDE concept originated at Palantir in the early 2000s. While serving defense and industrial clients with volatile workflows and locked-down data, Palantir bypassed standard software delivery cycles by placing engineers directly within customer offices. These developers wrote production code on-site, identified common workflow patterns across projects, and routed these insights back to the core platform team for standardized feature development.
This model differs from traditional enterprise tech roles:
- Sales engineers focus on pre-sales demos and exit the project once the contract is finalized.
- Solutions architects provide high-level technical advice and system design but do not write production code.
- Traditional consultants offer business methodology without participating in the vendor's core product feedback loop.
In contrast, FDEs act as temporary tech leaders for high-risk projects. They achieve end-to-end delivery under complex customer conditions and funnel real-world product feedback back to engineering, bridging the gap between bespoke service and standard software platforms.
Why Enterprise AI Agent Deployment Mandates On-Site Engineering
The structural demand for FDEs in the Agent era is driven by a fundamental shift in software behavior. While traditional SaaS products act as tools with clear parameters, AI agents are designed to execute complex, multi-step operations autonomously on behalf of users.
This transition introduces three deployment challenges:
First, agents must be deeply integrated into the client's actual operations. For example, a financial compliance agent must fetch evidence across legacy databases, match transactions against money laundering patterns, and compile suspicious activity reports (SARs). These compliance parameters reside in institutional memory rather than standard documentation.
Second, agent failures lead to operational failures. An overlooked transaction can result in severe regulatory penalties for a bank, raising the accuracy requirements for localized domain knowledge.
Third, because the enterprise agent market lacks established standards, customers struggle to define their requirements beforehand. FDEs must work on-site to iteratively build, test, and refine system interfaces based on real-time feedback.
Operational Cost Bottlenecks and Vendor Lock-in Risks
Despite its early traction, the FDE model faces commercial scaling challenges.
First, human-centered deployment remains expensive. FDEs use manual engineering to align products with customer systems. Gartner reports that by 2028, 70% of FDE-led agent deployments may be abandoned due to high vendor costs and a lack of customer capability to maintain the systems. Without converting field code into reusable product platforms, the FDE model risks degenerating into low-margin IT outsourcing.
Second, the reliance on field engineers impacts margins and company valuation models. Expanding FDE teams pushes AI developers toward a hybrid model of consulting and software, threatening the high gross margins typical of SaaS. Furthermore, proprietary deployment arms, such as OpenAI Deployment Company, focus exclusively on their own models, which exposes enterprise buyers to vendor lock-in risks.
Finally, rapid advancements in automation are squeezing basic integration work. As automatic API configuration, prompt optimization, and data mapping tools improve, the routine tasks of FDEs will be automated. The future value of FDEs will shift to system-level orchestration, such as multi-agent architectures and Model Context Protocol (MCP) design.
Assessing the Long-Term Value of the FDE Strategy
The FDE model serves as an intermediary stage for enterprise AI to mature from proof-of-concept to production. For AI vendors, FDEs should function as a product discovery mechanism that feeds field insights back into standard platform features to generate recurring value. For enterprise clients, establishing knowledge transfer mechanisms during deployment is vital to ensure internal teams can manage and scale these agents independently.
For engineering professionals, hybrid FDEs who combine model expertise with deep business domain knowledge remain highly sought-after. For investors, evaluating these AI companies requires measuring the speed at which field engineering insights are converted into standardized, scalable software platforms.