Half of all enterprise agentic AI projects never make it out of the pilot stage. That figure, cited by Dialpad ahead of Enterprise Connect 2026, which opens in Las Vegas next week, is, if anything, conservative.
A July 2025 study from MIT’s Project NANDA, based on analysis of 300 public AI deployments, found that only 5% of integrated AI pilots delivered measurable P&L impact. Gartner, meanwhile, predicts over 40% of agentic AI projects will be cancelled by 2027, citing escalating costs, unclear business value, and inadequate risk controls.
The gap between adoption and results is what Dialpad is targeting with new platform capabilities announced this week.
CEO Craig Walker said: “Enterprises aren’t struggling with AI ambition, they’re struggling with AI execution. Billions have been spent on agentic AI, but too many projects stall before delivering real, measurable results.”
Skill Mining and Proving Ground: Validating Enterprise AI Agent ROI Before Deployment
The two headline additions are Skill Mining and Proving Ground. Skill Mining analyses historical conversation data to surface customer experience friction points and rank potential AI use cases by expected impact, removing the guesswork from deciding where to deploy agents first. Proving Ground then lets teams model expected outcomes, including resolution rates, average handle time, and CSAT scores, before any agent goes live.
Hayley Sutherland, Research Manager for Conversational AI at IDC, said:
“The real value for customers right now is moving beyond retrospective analytics to understanding the specific impact and resolution rates ahead of time. By showing them the quantifiable ROI before deployment, you help them reduce failed AI pilots and maximise ideal business outcomes.”
No-Code AI Agent Builder Targets Contact Centre Automation at Scale
Agent Studio is Dialpad’s no-code environment for building AI agents across voice and digital channels. Using a conversational interface and a library of pre-built connectors, operations teams can configure enterprise-grade agents without developer involvement, tied directly to existing workflows and security policies.
This is territory that several larger vendors are also contesting: Microsoft has expanded its Copilot agent capabilities to Teams users, while Genesys, NICE, and Five9 have all made agentic announcements with no-code tooling attached.
Dialpad is also introducing Guardian, a real-time safety supervisor that monitors agentic interactions continuously to limit data exposure and flag compliance issues throughout the agent lifecycle. For regulated industries this matters at the procurement stage, and Dialpad has a live example to point to. Chris Martinez, Global CIO at Healthcare Outcomes Performance Company (HOPCo), said:
“Dialpad’s agentic AI capabilities helped us move from testing to enterprise-wide deployment with confidence by identifying where AI would have the greatest impact. As a result, we reduced resolution times, improved patient satisfaction, and maintained the strict governance our industry requires.”
Dialpad AI Adoption: Key Contact Centre Metrics
Dialpad reports that 97% of its contact centre customers currently use its AI capabilities, with more than 775 million AI Recaps and 450 million AI CSAT scores generated to date. A Forrester Total Economic Impact study published last year found customers reported a 20% reduction in average handle time, a 50% drop in post-call work, and a 75% reduction in manual management tasks — alongside a 10% improvement in annual agent retention.
The platform runs on a multi-LLM architecture combining proprietary speech, intent, and task models with third-party models, selecting dynamically based on task complexity. As Martinez added:
“Healthcare organizations don’t have the luxury of trial and error when it comes to patient communications. Dialpad helped us identify where AI would have the greatest impact — and gave us the confidence to scale it across the enterprise.”
