Skip to content
PATHSUM

PathLAB Studio / Enterprise builders

Intelligent Automation.For consequential workflows.

PathLAB Studio is Pathsum’s forthcoming workspace for enterprise builders: create, operate, investigate and improve custom AI automations—and turn their work into reusable training data.

Local-first. Your models, tools and data.

Coming soon · Expected January 2027.

Hi, I’m Grep.I’ll show you around.
The Pathsum principle
Confidence = ƒ(Control)

Control creates confidence.

All in PathLAB StudioExplore the workspace

One connected workspace

The whole life of your automation.

Carry actual operating experience into experiments, datasets and the next implementation.

  • Create & trial

    Author an automation using code, an agent or guided discovery. Try it before putting it to work.

  • Run & observe

    Run under your chosen rules, permissions and limits. Inspect inputs, outputs, human-gate decisions and observed outcomes.

  • Experiment & improve

    Compare commercial, open-weight or your own models in the same runtime. Adjust each candidate’s constraints to find the lowest operating cost that meets your requirements, including retries and human review. Test prompts, decisions and structure too, with comparisons and configured Ratchet loops.

  • Research & investigate

    Build automations that generate and assess candidate solutions. Keep the question, attempts and evidence connected.

  • Prepare reusable datasets

    Curate run inputs, outputs, human-gate decisions, corrections and outcomes—for evaluation, machine learning or another authorized purpose.

Machine learning / Data from your operations

Training data from the work your apps already do.

Your Pathsum automations create examples as they run. Combine reviewed examples across your company into training datasets for models built around your tasks and procedures.

Let’s take a closer look.

What you can build with the data

  • Train models for your automation

    Use reviewed data from your automation to train a model for the work it already does. Test whether a smaller, specialized model can take over that work at lower cost.

  • Build evaluation datasets

    Reserve reviewed examples as repeatable test cases, separate from training data. Assess models and automation versions against your requirements before adopting changes.

  • Research behavior and outcomes

    Use the interactions and outcomes your automations capture to investigate customer needs, user behavior, recurring problems and process patterns. Prepare datasets for product research, marketing analysis or other authorized purposes.

You choose what to include in each dataset and how it may be reused.

Test trained candidates in PathLAB before adopting them into your automation.

The Pathsum principle

Confidence comes from control.

Define what the automation may do, where it must stop and what needs your approval. Keep a verifiable record of the path it took.

Human-gate decisions and observed outcomes also become evidence for experiments and reusable datasets.

Human gateSome steps need your say-so.Grep holds a stop sign, then turns it over: I need a human.
Confidence = ƒ(Control)

Control =PATHSUM

  1. PathPLANsDefine the workflow, checks and budgets.
  2. Sealed ToolsRestrict what each step may use.
  3. Human GatesRequire approval at the actions you choose.
  4. PathPROOFKeep an independently verifiable record.

Confidence is a function of control. Control is the PATHSUM of PathPLANs, Sealed Tools, Human Gates and PathPROOF along the execution path.

Build with PathLAB Studio

Make the work useful.Keep what it teaches you.

Create, operate, investigate and improve—with the evidence and data connected.

Coming soon · Expected January 2027. Opens your mail app.

View pricing and licence terms