The Enterprise AI Problem Nobody Wants to Talk About

According to Accenture’s own Pulse of Change research, only 32% of business leaders report sustained, enterprise-wide AI impact. Other data is harsher: MIT’s NANDA initiative found that 95% of generative AI pilots fail to deliver measurable P&L impact. The RAND Corporation puts the overall AI project failure rate at 80.3% — roughly twice the failure rate of non-AI IT projects. Whatever the exact number, the pattern is clear: enterprises are good at starting AI projects and bad at finishing them.

On May 6, 2026, at ServiceNow’s Knowledge 2026 conference in Las Vegas, ServiceNow and Accenture announced a direct structural response to this problem: a joint Forward Deployed Engineering (FDE) program designed to move enterprise agentic AI from isolated pilot to production at scale.

The FDE model isn’t new — Palantir pioneered the approach for data infrastructure, and AI labs including OpenAI and Anthropic have used it to land large enterprise deals. What’s different here is the intent to productize it: a repeatable, two-vendor program specifically built around the pilot-to-production gap.

What Forward Deployed Engineering Actually Means

The mechanics are straightforward. ServiceNow deploys its own AI-native engineers inside mutual customers’ environments. Accenture contributes FDEs with industry domain depth — people who understand how an automotive supplier’s supply chain works, or how a hospital’s clinical workflow is constrained by regulatory requirements. The two teams work together inside the client’s environment to build agentic AI workflows natively on the ServiceNow AI Platform.

“We’re not simply handing over instructions,” said John Aisien, SVP and General Manager of Central Product Management at ServiceNow. “Our teams are in the customers’ environments, implementing ServiceNow, customer and third-party building blocks, and demonstrating the resulting value metrics in the ServiceNow AI Control Tower.”

This distinction — physically building inside the environment rather than building externally and handing over — is the whole point. The delivery gap isn’t primarily a technology problem. It’s a last-mile problem: getting AI agents connected to real enterprise data, running inside real enterprise security constraints, integrated with real legacy systems, and accepted by the people who have to use them. None of that can be solved remotely from a vendor’s office.

ServiceNow’s Platform and the AI Control Tower

The FDE program is built on the ServiceNow AI Platform, which gives clients access to more than 300 pre-built AI agent skills and agentic workflows. These aren’t demos — they’re production-ready building blocks for tasks like IT service management, HR case resolution, procurement approvals, and customer service routing. The idea is that FDEs configure and connect these skills to the customer’s specific data and systems, rather than building from scratch.

At the center of the governance layer is ServiceNow’s AI Control Tower: a centralized dashboard that gives organizations visibility into every AI agent running in their environment — what each agent is doing, its performance metrics, and where it’s consuming resources. This addresses one of the core governance gaps that Gartner flagged as a leading cause of agentic AI project cancellations: organizations deploying agents without visibility into what those agents are actually doing.

What ServiceNow calls the “single continuous motion” from first build to enterprise-wide deployment is the architectural bet here. The FDE team builds the initial workflow in production — not in a sandbox — and the same platform and governance layer scales it to the rest of the organization. There’s no migration step between pilot environment and production environment, because they’re the same environment from day one.

The Honest Questions This Model Raises

The FDE model is expensive. Senior FDEs at leading AI companies command $300,000–$600,000+ in total compensation. A joint program that deploys ServiceNow and Accenture engineers inside a customer’s environment for weeks or months isn’t a product you can buy for a few thousand dollars a month. This is enterprise consulting dressed in engineering clothing — valuable, but accessible mainly to organizations with large IT budgets.

There’s also a platform lock-in question. The 300+ pre-built agent skills run on the ServiceNow AI Platform. If a customer’s workflow is built natively on that platform, migrating away later becomes substantially harder. That’s not unique to this program — it’s a general property of platform-native AI — but it’s worth naming explicitly, especially for organizations still evaluating their long-term AI architecture.

The delivery gap problem also has causes beyond the last mile. As previously noted on vortx.ch, the 95% pilot failure rate often traces back to poor data readiness, absence of clear success metrics before the project starts, and loss of executive sponsorship during execution. Embedding excellent engineers doesn’t fix bad data or absent governance. Organizations that haven’t addressed those prerequisites will get expensive FDE engagements that still stall.

Why This Model Is Becoming a Template

Despite the caveats, the FDE program reflects a broader shift in how enterprise AI is being sold and delivered in 2026. Vendors are increasingly accepting that the product alone doesn’t ship — delivery and integration are part of the value, and they require people on the ground, not documentation.

Accenture’s Ram Ramalingam framed it clearly: “The question our clients ask is not whether to invest in AI — it’s how to make it work at enterprise scale. This program brings together Accenture’s industry depth and implementation reach with ServiceNow’s AI Platform to deliver real results, not roadmaps.”

The “not roadmaps” framing is pointed. After two years of enterprise AI announcements that produced plenty of slide decks and very little production software, customers are demanding proof of delivery before they sign. The FDE model is a structural answer: we build it, in your environment, before the broader rollout begins. You see it working in production before you commit to scale.

For the broader agentic AI market, where 54% of enterprises already run AI agents but governance consistently lags, the addition of an embedded engineer who builds governance and visibility in from day one — rather than retrofitting it after deployment — addresses a failure mode that has already cost a lot of organizations a lot of money. Whether the ServiceNow + Accenture program proves replicable at scale will be worth watching through the rest of 2026.

Further Reading

• Official Accenture announcement — the primary source, including details on the AI Control Tower and the 300+ pre-built agent skills
• Why AI Projects Fail and How Forward Deployed Engineers Fix It — a practitioner breakdown of what FDE roles actually do on the ground
• Accenture Pulse of Change research — the underlying data on why only 32% of enterprises report sustained AI impact