SDSignal Desk

Disrupt 2026 roundtables keep returning to one theme: AI pilots that stall before real work

· Updated · Signal Desk newsroom

Analysis of reporting by TechCrunch. How we work

Image: TechCrunch

TechCrunch’s list of interactive roundtables for Disrupt 2026 reads as a snapshot of startup priorities: stalled enterprise pilots, physical AI reliability, the cost of competing in frontier AI and open-weight models, with less attention on affected workers and customers.

Key takeaways

  • Disrupt 2026 runs October 13–15 at Moscone West; TechCrunch expects more than 10,000 attendees.
  • Several sessions focus on why enterprise AI pilots stall.
  • Physical AI sessions emphasize reliability and trust over demos.
  • PitchBook’s Harrison Rolfes discusses what it calls a $100 billion entry fee for AI.
  • Some sessions are sponsored; this is TechCrunch promoting its own event.

Reading the program

This is TechCrunch promoting its own conference, so it is best read as a program listing. Disrupt 2026 runs October 13 to 15 at Moscone West in San Francisco, and TechCrunch expects more than 10,000 founders, investors and operators. The roundtables are small, discussion-style sessions across two dedicated spaces. Agendas like this offer a snapshot of where startup money and attention sit and which problems people will admit in front of peers. Earlier coverage looked at Disrupt’s safety sessions (safety sessions) and its Real World AI Stage (Real World AI Stage).

Pilots, not breakthroughs

The recurring word is pilot. Kodamai’s Maha Achour and 2468 Ventures’ Pankaj Kedia ask why so many enterprise AI projects stay stuck there. Databricks’ Ankita Mehta covers why pilots stall and why not every workflow needs an agent. Balerion’s Naren Krishna and Pruven Capital’s Travis Skelly argue AI-native approaches will not work for large companies that cannot rip out core systems, and Smartsheet’s Rajeev Singh frames AI as exposing what enterprise software got wrong rather than replacing it. A circuit that admits the gap between demo and deployment is healthier than one that pretends it is closed, and it gives enterprise buyers cover to press vendors for production results.

In healthcare, Questa Capital’s Brian Butler contrasts decades-long electronic health record adoption with generative AI uptake in about three years, which raises the stakes on oversight.

Physical AI and market structure

Physical AI is the other drumbeat: iMerit’s Radha Basu on data problems behind robots and autonomy, MetaProp’s Jackson Feder on construction, property and energy, Rugged Robotics’ Logan Farrell and Brick & Mortar Ventures’ Austin Yount on what a robot must deliver beyond a demo, and SandboxAQ’s Jack Hidary with Nvidia’s Stacie Calad-Thomson on AI grounded in physics, chemistry and biology. The emphasis on reliability is apt: a failed robot on a job site can hurt someone.

Two sessions address market structure. PitchBook’s Harrison Rolfes asks whether anyone can still compete given what it calls a $100 billion entry fee, while Premise’s Mercedes Bent and Vanessa Larco cover open-weight models as an alternative startup stack, one of the few forces against concentration at the top. Nvidia speakers have debated open versus closed AI elsewhere at Disrupt (open or closed). An Anthropic forward-deployed engineering session signals how labs plan to sell into enterprises.

What is thin

There is a CISO career session and much on reliability, but little centers the workers agents are meant to absorb or customers on the receiving end of agentic shopping. Walmart’s Srishti Chaudhary asks who owns the customer relationship in that world. Some sessions are sponsored, including by the New York Stock Exchange, Twin Peaks Wealth Advisors and UC Berkeley’s Sutardja Center, and a quantum computing session had no speaker announced at publication. More in our enterprise coverage, startups coverage, Anthropic coverage and NVIDIA coverage.

Who feels it

Enterprise buyers
Stalled pilots treated as normal justify hard questions on production results.
Physical AI founders
Investors prioritize reliability, workflow fit and ROI over demos.
Startups weighing open models
Open weights framed as a cost, control and security trade.
Healthcare organizations
Rapid adoption raises oversight stakes.

What to watch

  1. Concrete pilot-to-production numbers from sessions.
  2. Investor views on capital concentration and open-weight builders.
  3. Physical AI startups with real deployment and safety records.

FAQ

When is Disrupt 2026?

October 13–15 at Moscone West in San Francisco.

What theme dominates the roundtables?

Why enterprise AI pilots stall before production.

Is physical AI covered?

Yes, with an emphasis on reliability, data and deployment beyond demos.

Are sessions sponsored?

Some are, including by the NYSE, Twin Peaks Wealth Advisors and UC Berkeley’s Sutardja Center.

Read the original

Continue at the source.

TechCrunch

Companies: Anthropic, NVIDIA

More on Anthropic

All Anthropic coverage →