The Pilot-to-Production Gap
In an exclusive interview, UiPath CMO Michael Atalla laid out a reality that most enterprise AI vendors won't say out loud: "Most companies have AI pilots. Very few have AI in production."
According to UiPath's internal data, 73% of Fortune 500 companies have at least one AI pilot running, but only 12% have deployed AI agents in production workflows. The gap between experimentation and execution remains the industry's biggest challenge.
Atalla identifies three patterns that separate successful deployments from stalled ones:
- Start with the process, not the model — companies that map workflows first, then apply AI, succeed 4x more often
- Human-in-the-loop is not a weakness — the best deployments augment workers, not replace them
- Measure business outcomes, not model accuracy — a 90% accurate model that saves $2M beats a 99% accurate model that nobody uses
The RPA + AI Convergence
UiPath's strategy in 2026 centers on what Atalla calls "agentic automation" — combining traditional RPA (robotic process automation) with AI agents that can handle unstructured data, make decisions, and adapt to changing inputs.
The convergence makes sense: RPA handles the deterministic, rule-based tasks (data entry, form filling, system integration), while AI handles the judgment calls (document understanding, exception handling, customer communication). Together, they create workflows that are both reliable and intelligent.
Key products driving this vision include UiPath Agent (autonomous task execution), Document Understanding+ (AI-powered document processing), and Process Mining AI (automatically discovering automation opportunities).
What Enterprise AI Looks Like in 2026
Atalla shared concrete examples of production AI deployments that are delivering measurable ROI:
- Insurance claims processing — AI agents handle initial assessment, route complex cases to humans, reducing processing time from 14 days to 2 days
- Financial reconciliation — Agentic automation matches transactions across systems, catching errors humans miss
- Supply chain optimization — AI predicts disruptions and automatically adjusts procurement, saving one manufacturer $15M annually
- Customer onboarding — Document AI extracts data from KYC documents, reducing onboarding from 45 minutes to 8 minutes
Atalla's framing of "agentic automation" is smart positioning from UiPath. The company was written off by many as a legacy RPA vendor, but their AI pivot is genuine. The key insight — process first, model second — is something I've seen work in my own projects. Too many teams start with "we need to use GPT" instead of "we need to solve this workflow problem."