Agents that run under their own identity, and can be checked

We can, and the ERP and CRM half is work we already do. TechFabric implements Dynamics 365 Business Central and Dynamics 365 CRM as a Microsoft Solutions Partner, and lands ERP and CRM data in Databricks under Unity Catalog.

An agent built on that reads only what its grants allow, writes back through the same interfaces a person would use, and waits for approval before it posts anything to the ledger. Funko's warehouse system and AmTab's store-to-ERP integration both run on Business Central, so we know the records an agent would be working with.

Ask what happens when it fails, because a demo never shows that.

A firm that has shipped to production can tell you how the system survives a model timeout halfway through a run, how it's kept from reading data its user can't see, who approves the steps that can't be undone, and how you'll know next month's version is better than this month's.

Then ask for one system they built that's still running, and talk to the engineers who'd do the work. For us that's Canvass, which we built on Cloudflare and Temporal.

Four things, and none of them is the prompt. It survives a restart without losing its place. It cannot read data its caller could not. A person can approve or stop it, and the run waits without falling over.

And somebody can measure whether a change made it better. A pilot that has none of these is a demo, and the gap between the two is most of the work.

Its tools resolve through Unity Catalog and it runs under its own service principal, so the grants that bound a human analyst bound the agent too. That is a platform property rather than a prompt instruction, which matters because a prompt instruction is a request and a grant is not.

Model access goes through a provider registry, so a fallback is configuration rather than a rewrite. On Databricks that registry is Unity Gateway. Runs that are mid-flight are durable, so they resume rather than restart.

No. Harness is Apache-2.0 and it exists because we needed it, so we are happy when a team takes it and never talks to us again. We also work in whatever you already have. The framework is how we go faster, not a condition of the engagement.

A fixed set of cases with known answers, run on every change, scored the same way each time. Without that, quality is whoever spoke last in the review meeting. TechFabric Experiments is the accelerator we use for it and the approach works the same way without it.