Enact GTM brain

category

market profile

• text: Robotics teams must close the gap between polished demonstrations and reliable performance when the physical world changes.
• text: Physical AI model builders can source task-specific evaluation and training data externally rather than building every collection operation themselves.

personas

• title: Robotics foundation model lead
  pains:

  • Policies fail in rare recovery states that generic demonstrations miss
  • Need evidence that post-training improves real task performance
• title: Head of robot learning
  pains:

  • Repeatedly reproducing physical failures to collect useful training data
  • Coordinating real-world evaluations and retraining loops
• title: Robotics deployment engineering lead
  pains:

  • A policy works in a demo but fails when objects slip or the scene changes
  • Need task-specific measurements before deployment

icp

• text: Robotics model labs and robot developers with existing policies that can be rolled out on physical tasks.
• text: Physical AI teams working on high-value manipulation tasks and seeking targeted recovery data; company size is not specified.

message pillars

• text: Train recovery behavior in actual physical environments.
• text: Find failing states and create demonstrations focused on those failures.
• text: Measure whether the resulting data actually addresses observed failures.

objections

• objection: We already collect robot demonstrations internally.
  response: Evaluate a focused failure-recovery dataset on one task against your existing collection and evaluation baseline.
• objection: Will the data improve our particular robot and task?
  response: Agree on failure states and a held-out real-world evaluation before commissioning collection.
• objection: Can this work with our policy and hardware?
  response: Discuss policy access, robot compatibility and evaluation conditions directly; public material does not specify supported platforms.

competitors

• name: LightWrk
  why: Offers expert failure-mode evaluation and reasoning data for robotics models.
• name: Cortex AI
  why: Supplies real-world workplace robotics training data.
• name: Linewise
  why: Robotics model training and physical data provider identified as a nearby alternative.

keywords

• robotics recovery training data
• robot policy evaluation
• physical AI post training
• robot manipulation failure data
• real world robotics benchmark
• robot reinforcement learning environments

voice

• technical
• concise
• evidence-oriented

signals

• text: A robotics lab is shipping a model that performs well in demos but struggles with recovery from physical disruptions.
• text: A robot developer is scaling task-specific evaluation or collecting demonstrations for a new manipulation task.

pricing

motion

channels

• X
• LinkedIn
• YC launch

confidence

Founders

• name: James Stevens
  title: Founder/CEO
  role: CEO
  bio: Founder & CEO @ Enact
• name: Govind Chada
  title: Founder/CTO
  role: CTO
  bio: Founder at Enact. Previously: Robotics research in Chelsea Finn’s IRIS Lab at Stanford AI Lab, Meta Reality Labs.