Are we paying for software, or are we paying for someone to make it actually work?
Forward deployed engineering (FDE), an approach closely associated with Palantir, is gaining traction across the contact centre as a service (CCaaS) and customer experience (CX) market. It places engineers alongside customers to solve the difficult problems that arise between a promising AI demonstration and a working production deployment.
On 5 November 2026, Genesys announced an expansion of its FDE capabilities, with engineering teams working directly with customers to help move agentic AI use cases into production. Salesforce, AWS and specialist AI providers are also investing in versions of this model. That raises an important question: are customers buying implementation capacity, or accountable help delivering business outcomes?
What Genesys Announced
Genesys describes FDE as part of a broader customer success approach. Its Agentic Experience Labs and Experience Orchestration Plans are intended to establish early value and guide longer-term transformation. Expert Apps provide pre-packaged agents and tools through AppFoundry. FDE teams work with customers on the integration, governance and operational details needed to make these capabilities useful in practice.
Genesys has identified Banco de Crédito del Perú, eir and Riachuelo among customers using FDE support for agentic AI deployments. The model also creates a feedback loop: lessons from customer projects can inform future product capabilities and Expert Apps.
The timing matters. At Xperience 2026, Genesys outlined an agentic orchestration direction that includes the AI Control Plane, Contextual Intelligence, Cloud Navigator and Cloud Orchestrator, with elements planned for phased availability. Embedding engineers with customers as those capabilities arrive is not just a services decision. It is also a way to accelerate platform adoption.
This Is Bigger Than Genesys
Salesforce has outlined ambitious plans to expand its forward-deployed engineering capacity around Agentforce, including a partner network. AWS has announced a significant investment in FDE. Sierra uses the term “agent engineers”, while NICE has recruited for AI-focused FDE roles. The terminology differs, but the underlying objective is similar: shorten the distance between technology deployment and operational results.
There is also a useful counterargument. Decagon CEO Jesse Zhang has questioned whether customers should need embedded engineers indefinitely. Decagon uses engineers to accelerate initial deployments but argues that customers should ultimately be able to manage their own agents. That distinction matters. An FDE engagement should build customer capability, not permanent dependence.
Accelerator or Dependency?
I can see both sides. Agentic AI projects frequently stall not because the demonstration failed, but because production introduces integration challenges, data quality issues, security requirements, governance decisions and operational ownership questions. Traditional professional services can address much of this, but an embedded engineering model may close the gap faster, particularly when the product itself is still evolving.
The risk is that the customer becomes dependent on the vendor to maintain what was built. Custom solutions can increase switching costs, weaken knowledge transfer and leave existing implementation partners with an uncertain role. There is also a commercial question when work developed for one customer influences products later offered to the wider market.
Four Questions I Would Ask Before Signing
1. What does success look like, and when does the FDE team leave? Agree on measurable outcomes and exit criteria before the work begins.
2. Will our people work alongside the engineers from day one? Knowledge transfer should happen throughout the engagement, not in a final handover.
3. Who owns the resulting intellectual property and custom work? Clarify rights, reuse and commercial terms if a customer-specific solution becomes a packaged capability.
4. How will our existing partners be involved? Systems integrators and consulting teams need clear responsibilities so the solution remains supportable after the embedded team leaves.
I would also watch for warning signs: no defined exit criteria, reliance on custom code where standard platform capabilities would suffice, minimal customer participation and time-and-materials billing without meaningful outcome measures.
My Perspective
I do not see FDE replacing solutions architects, professional services or implementation partners. I see it as another delivery model, particularly valuable where AI capabilities are advancing faster than organizations can operationalize them. The real test is not how many engineers a vendor embeds. It is whether the customer can operate, govern and improve the solution after those engineers have moved on.
A successful FDE engagement should eventually make itself unnecessary. Otherwise, have we solved the customer’s problem or simply created another layer of dependency?