Software house
You build the application or integrations and need support with RAG, agents, evaluations and AI quality control.
Partner delivery
We close AI risk gaps in partner-led projects.
We support software houses, systems integrators and consultancies with evaluations, guardrails, AI Act work and AWS architecture. We can work openly as the specialist or white-label under the partner’s brand.
Who it is for
When you own delivery but need a specialist AI workstream.
We do not replace the partner’s client relationship. We complement delivery where independent quality evidence, governance or secure GenAI architecture is required.
Systems integrator
You connect ERP, CRM, DMS or client data and need a secure GenAI layer with acceptance criteria.
Consulting firm
You lead strategy, process or compliance and need a team to validate the solution technically.
Internal client team
You have developers but need an independent architecture review, testing and production plan.
Practical starting packages
Three scopes that fit cleanly into a partner-led project.
Each package has a concrete outcome and a clear responsibility boundary. After the first call, we adapt it to the architecture, risk and commercial model.
Data-to-Agent Accelerator
For a data, cloud or BI partner that owns the data layer and client relationship.
A secure RAG knowledge base or AI agent connected to the existing data platform, with acceptance criteria and a production plan.
The partner owns data and platform delivery. Semitora owns the GenAI layer, evaluations, guardrails and governance.
AI Agent Test & Risk Sprint
For a QA firm, software house or integrator before acceptance or production launch of an AI system.
A golden set, quality and security tests, a risk register and an evidence-based go/no-go decision with a remediation plan.
The partner keeps delivery and the client. Semitora provides an independent, measurable assurance workstream.
Secure Agent Launch
For a cybersecurity partner or team launching an agent with access to tools and sensitive data.
Permission architecture, controlled actions, human-in-the-loop, audit logs and a technical evidence pack before production.
Semitora designs and hardens the agent layer. An independent security partner can lead red teaming and final assurance.
Engagement models
Visibility and responsibility matched to the partner contract.
We agree scope, communications and ownership of materials before work starts. We do not contact the end client outside the agreed model.
Co-delivery
One joint team with a transparent division of responsibility. Semitora leads the agreed AI workstream.
White-label
We work as the partner’s specialist back office, adapting materials and communication to their process.
Independent assurance
A separate review before acceptance, production release or a decision to scale.
What we can own
Modules that fit into a larger delivery programme.
AI Assurance Sprint
Golden set, evaluations, security tests, a risk report and go/no-go criteria.
AI Act and governance
System inventory, risk classification, roles, human oversight, logs and technical evidence.
RAG and AI agents
Architecture, prototype or production component with control of sources, permissions and actions.
AWS and architecture review
Solution design, security and cost review, and hands-on support for the delivery team.
Bid support
Technical scope, acceptance criteria, responses to AI requirements and risk review before pricing.
Post-launch care
Regression evaluations, quality monitoring, incident handling and recurring system-owner reports.
Production proof
We do not learn AI on the partner’s project.
mojApteczka is Semitora’s own production product on AWS, with ETL, RAG, mobile apps, evaluations and AI support. The June 2026 snapshot describes this system and point in time, not a benchmark for another deployment. Cost covers the inference layer only; it excludes application development, data preparation, storage, observability and human review.
RAG
versioned knowledge, source citations and controlled refusal
$0.0006
rounded inference-layer cost per AI scan
302,516
drug-interaction records in the production knowledge base
For internal discussion
Partner Capability Brief — one page to share with your team.
Capabilities, three entry packages, engagement models, client-relationship safeguards and Semitora contact details.
Working principles
The partner keeps the client. Responsibility stays clear.
Before starting, we document workstream boundaries, communication channels, acceptance criteria and escalation. The client receives one coherent delivery model and the partner remains in control of the relationship.
- ✓No direct upselling to the partner’s client outside the agreed model
- ✓Clear ownership of data, integrations, model, security and acceptance
- ✓Evidence and decisions recorded in the project repository or system
- ✓Confidentiality, access and ownership agreed before data is shared
Starting together
A small initial scope to validate the fit.
01
Context and boundaries
The partner describes the client, project stage, architecture, risks and preferred visibility model.
02
Workstream and criteria
We agree the outcome, dependencies, access, responsibility and objective acceptance criteria.
03
Delivery and evidence
We work in the project tooling, report risks and hand the artifacts to the partner team.
04
Acceptance or ongoing care
We close with a decision and documentation or move into an agreed operational control loop.
Common questions
Collaboration without channel conflict.
Yes, when we agree the white-label model, communication scope and responsibility. We can also work openly as a specialist subcontractor.
Yes. Independent assurance before acceptance or production is a standalone scope and does not require us to take over implementation.
We can run evaluations and monitoring on retainer, but support boundaries and SLA are agreed with the partner for the specific system.
Have a project that is missing an AI workstream?
Send a short description of the scope, stage and engagement model. We start by aligning responsibility, not with a sales presentation.