AI Engineered For Production

Can we keep our data on our own infrastructure?

Yes. Sirma.AI Enterprise deploys cloud, on-premise, hybrid, or fully air-gapped. Sensitive data is anonymized before any external model call and locally de-anonymized on return - your data stays in your jurisdiction by default.

The six solutions don't quite match our use case - can you build something else?

Yes. The six are productized starting points, not the whole menu. Underneath all of them is the same set of building blocks - agents, multi-agent teams, workflows, RAG, and tool/system integrations - composed differently per project. If your workflow isn't one of the six, we scope it the same way: one painful workflow first, a number we'll defend, then build.

What if we don't have AI/ML talent on the team?

That's the most common starting point. Engagements are designed for teams without dedicated AI engineers - one senior team designs, ships, and operates the system, trains your people to run it, and can stay embedded longer-term if you want senior hands on your side of the table.

How do you handle confidential or regulated data?

Three layers: deployment (on-prem or air-gapped if needed), routing (sensitive data goes only to local models), and audit (every agent decision logged for compliance review). The platform is pre-configured for GDPR, HIPAA, SOC 2, ISO 27001, NIS2, and DORA.

How does an engagement actually work - what are we committing to?

We start narrow: one workflow, with a written ROI we'll defend before we take the work. We build it the way you'd build any production system - versioned prompts, evaluation harness, observability, guardrails, rollback - and you only commit to the next step once the previous one has paid back. Fixed scope, no time-and-materials creep, no long-term lock-in. If the work surfaces that AI doesn't pay back yet, we'll tell you and stop there.

How is this different from a typical AI agency, or from hiring in-house?

Most agencies sell you a slice and hand off a demo; we design, ship, and operate end-to-end - and we're part of the same group that builds the Sirma.AI platform, so it's one accountability line with no third-party version drift. Versus hiring: recruiting takes 6+ months and produces a person - we produce a shipped, governed system, with the option to stay embedded until you've hired or transitioned ownership internally.

How does the EU AI Act affect our roadmap?

Most AI use cases fall in the "limited" or "minimal" risk tiers - manageable without major compliance work. High-risk uses (recruitment, credit scoring, medical decisions) need risk-tier mapping, audit trails, and human oversight. We classify every use case under the Act's tiers before the project starts, not after deployment, and bake the controls into the build.

Is Sirma.AI Enterprise a separate product we have to license?

Yes - Sirma.AI Enterprise is Sirma Group's commercial GenAI platform, and it can be used two ways. Most clients have Roweb implement on top of it, with licensing folded into the engagement (pricing depends on deployment mode - cloud, on-prem, or air-gapped - and scale). But if your team wants to build, extend, and operate on the platform directly, it can also be licensed standalone - as the base layer for your own AI infrastructure roadmap. Your engineers own the platform; ours stay on call for the hard parts. Either way, because Roweb is part of Sirma Group, there's no third-party license overhead or version-drift risk.