AI Assurance
For: a project before acceptance or production.
A golden set, quality and security evaluations, guardrails, risk register, Evidence Pack and go/no-go. Co-branding is preferred; white-label keeps agreed evidence of delivery.
A specialist AI workstream without taking over the client relationship.
We support grant advisers, transformation agencies, software houses and integrators. We close the technical workstream: AI Assurance, a pre-check for a funded implementation, and Rescue PoC delivery with a path to production.
Three engagement modes
Dig.IT/SMART technical pre-check
For: a partner’s client considering grant funding.
We review AI scope, data, integrations, architecture, acceptance criteria and budget assumptions. The partner’s client is the grant beneficiary; Semitora is the technical subcontractor — not legal counsel and not a guarantor of funding.
Rescue PoC → production
For: a prototype blocked before deployment.
Gap diagnosis, remediation plan, security, observability, integrations, regression testing, handoff and production-release preparation.
Production proof
302,516
drug-interaction records in the production knowledge base
drug-interaction records in the production knowledge base
USD 0.0006
rounded inference-layer cost of one AI scan
rounded inference-layer cost of one AI scan
Source: the mojApteczka production system, June 2026 operating snapshot. These are parameters of this system and point in time, not a benchmark for another deployment. The cost covers the inference layer only; it excludes application development, data preparation, storage, monitoring and human work. This is GenAI/RAG proof, not a claim of an autonomous-agent deployment.
Engagement models
- open co-delivery with documented responsibilities
- white-label specialist back office
- independent assurance before acceptance or production
- joint discovery or bid support
What we hand over
- architecture and acceptance criteria
- golden set and evaluation results
- risk register and remediation plan
- guardrails, logs and agent-action boundaries
- technical evidence for governance
Protecting the partner relationship
The partner keeps the client. We do not upsell directly outside the agreed model. Before accessing data, we agree roles, communication, IP, acceptance criteria and escalation.
Suggested first step
A 25-minute qualification call: current partner AI scope → one joint workstream → responsibility boundaries → decision on a small pilot or assurance sprint.