Enterprise Agents Are Past the Pilot Phase — The Governance Infrastructure Arriving Now Proves It
textak holds enterprise agents at 77% for widespread deployment in enterprise workflows, up from 76%. Today's news doesn't just confirm that number — it adds precision to the picture. What's arriving in production isn't autonomous agents operating in isolation; it's agents embedded inside ERP, CRM, procurement, and project management stacks, with identity governance infrastructure now being built to manage them. That's a qualitatively different signal than pilot announcements.
We weight this forecast heavily because the evidence today is unusually direct. Asana acquiring StackAI for multi-system agent orchestration, Coupa acquiring Rossum for intelligent document processing inside source-to-pay workflows, Salesforce acquiring Contentful to own the AI execution layer across commerce and CRM — these aren't R&D bets. These are enterprise software vendors paying real acquisition premiums to embed agent capabilities into live operational workflows. The InfoTech LIVE signal reinforces this: when 200+ CIO conference sessions pivot from 'should we adopt AI' to 'how do we govern and scale AI we're already running,' you're not looking at experimentation anymore.
The AppViewX Agent Identity Security launch is the signal we'd point to most directly. Security vendors don't build PKI-based lifecycle management products for agents running in sandboxed pilots. They build them when agents are touching production systems with real credentials, real access scope, and real audit requirements. The market for governing agents presupposes agents worth governing. That's a structural confirmation of deployment momentum, not an aspiration.
The strongest counterargument to our 77% is the one we've consistently named: hallucination rates remain too high for regulated industries — finance, healthcare, legal — and those sectors represent enormous shares of Fortune 500 workflow surface area. The Asana-StackAI and Coupa-Rossum acquisitions are happening in relatively forgiving domains: project coordination, procurement document processing. The jump to autonomous agents in compliance-sensitive workflows is a different risk tolerance entirely. We're also watching whether the 'human maintains project context' framing in the Asana deal is doing more work than it appears — that's a disclosure, not a feature. It may signal the human-in-the-loop requirement is stickier than the deployment narrative implies.
What would move us above 85%: A major regulated-industry firm — a top-10 bank or insurer — publicly documenting agent deployment in a compliance-sensitive workflow with measurable ROI and no mandatory human review gate. What would drop us below 65%: A high-profile enterprise agent failure that triggers liability litigation and causes the CIO community to reintroduce mandatory human-in-the-loop requirements as governance default. Neither has happened yet. The 77% reflects robust deployment in operationally forgiving domains with the regulated sector still holding back.