Process discovery

Surface

Map the work, then deploy agents.

01

Agent-led interviews

  • People do their normal work while Surface records their screen through a Chrome or Edge extension.
  • Private data on screen is masked on their device before the recording is saved.
  • A HappyRobot voice agent leads the interview, reacting live to the work on screen, asking why each step is taken and following up where something is unclear.
  • Each session is saved and analyzed for the process.
it asks about the screen, and never sees what is masked on it.
02

End-to-end discovery

  • The first recording becomes the base process graph.
  • Every interview after that is compared with it. Matching steps gain support, and new behavior becomes a candidate step.
  • Different ways of doing the same work stay on the graph as variations.
  • Surface shows when new interviews stop turning up new patterns.
every interview feeds one graph. a different way of doing a step stays on it.
03

Process redesigns

  • Surface proposes a redesigned process from the recorded work.
  • Each stage of it goes to an agent, a person or both.
  • It drafts the HappyRobot build behind it, with workflows, app access and Context tables. Every piece cites the recorded work, and your team reviews it before anything is built.
the one marked stays with a person.
04

Built for enterprise scale

  • Every study adds to one map of your operation, across divisions and teams.
  • Open any process to see its proposed design and where agents take over.
  • Where processes share a step, a tool or an input, Surface flags it, so one agent can serve more than one team.

FAQs

Process mining reads system event logs, so it only sees steps that leave a record. Process discovery follows the work and asks the people doing it, so it also captures the steps between systems, the workarounds and the reasons behind them.

A browser tab, an application window or the whole screen, through a Chrome or Edge extension. The worker picks the study and starts the session, and a HappyRobot voice agent can lead the interview alongside it.

Private data on screen is masked on the worker’s device before the recording is saved. Admins choose what’s masked, from personal information to all text, and passwords and access keys always are. If detection fails, the text stays hidden. Capture is consented, and saved evidence is encrypted.

Surface tracks when new recordings stop adding steps and what is still unsupported or contradicted, so you see when a process is ready for review.

Surface drafts the HappyRobot build: workflows, app access and Context tables, each tied to its evidence. Saving it deploys nothing.

Put agents to work in your enterprise.