A question companies ask

The question: Our staff want to build their own tools with AI. How do we allow it without losing control?

How to let staff build tools with AI, safely

Give every tool your staff build one governed path to production. On Substrait, each tool has a named owner, passes a security check before it reaches production, and reads company data only after the data owner approves it.

Last updated

Why this is hard

Building a working tool with AI now takes an afternoon. Putting that tool in front of colleagues safely is a different job. The tool needs a place to run, a sign-in, a security check, the right data, an owner and a way to switch it off. Most companies have none of that ready for tools built outside IT.

So one of two things happens. IT blocks the tools, and the work goes back to spreadsheets. Or staff run the tools anyway, on personal accounts and other people’s servers, and nobody can see them. That second path is the shadow IT most companies already have.

What Substrait does

Substrait gives every tool your staff build the same governed path.

  1. The builder keeps their AI model. A plugin connects Claude Code, the Claude desktop app or Cursor to Substrait.
  2. The tool starts in dev, behind your sign-in. Only your organisation can open it while it is being built.
  3. The tool passes a security check before production. Substrait scans the release for secrets, weak code, vulnerable dependencies and container issues. A release with too many issues stops for a human approver.
  4. The tool reads only the data its owner approved. Each grant is approved by the owner of that data and lasts for a set time.
  5. The tool has an owner and a record. Every change is versioned, every tool can be rolled back, and every tool can be switched off.

What changes for IT

IT stops picking which projects get built. IT sets the rules once, and Substrait applies them to every tool. A tool that stays inside its approved limits ships without a new review. Only a tool that asks for more data, starts writing instead of reading, or faces customers goes back for approval.

What changes for staff

Staff build the tools they need, with the AI model they already use. Substrait tells each builder what to do next, in plain words, from their first sign-in to their first tool in production.

Questions and answers

Should a company let non-engineers build software with AI?

Yes, with a governed path. The person closest to a problem often knows best what the tool should do. Substrait makes it safe to say yes: each tool has an owner, a security check before production, and data access its owner approved.

Do our staff need to learn a new AI model to use Substrait?

No. Staff keep the AI model they already use. A Substrait plugin works inside Claude Code, the Claude desktop app and Cursor, and one command takes a tool to Substrait.

What stops a staff-built tool from leaking company data?

On Substrait, a tool reaches company data only through a gateway, under a grant the data owner approved for a set time. Outbound traffic from a tool is denied unless a grant allows it. Substrait records every read.

What if our staff do not know how to build yet?

Substrait coaches each builder inside the platform, step by step. For teams that want people to teach people, the Builder Academy is an optional service, and Substrait's team can build the first systems alongside your staff.

Related

Start with a result, not a rollout.

Book a demo and bring one real use case. We will show you how a tool your people build goes live on Substrait, with its data approved and its security check passed.