The Semitora Path

Three levels. One path — from diagnosis to a working system.

We start with a paid audit, because free analyses end in a sales offer, not in knowledge. Every level has a clear scope, outcome and pricing model.

Level 01

AI readiness audit + AI Act compliance

An AI readiness audit is a paid, 2–4-week diagnosis of your processes, data and AI Act risk. We map the processes where AI can add value, inventory uncontrolled tool use (“shadow AI”), classify your systems by AI Act risk level, and build a prioritised implementation roadmap with estimated costs and ROI, plus architecture and funding (grant) recommendations. Optionally we close it with a proof it works (PoC) on your real data. The output is a working document for your board and IT — a basis for fact-based decisions, not a sales deck.

What you get

✓A process map with AI potential and a full “shadow AI” inventory
✓AI Act-compliant risk classification of your systems
✓An implementation roadmap with priorities, costs and estimated ROI
✓Architecture recommendations and a funding model (grants)
✓An optional PoC on your company's real data

Level 02

Production implementation

A production AI implementation is the building and launch of a system that runs in production — with architecture, code, integrations, security and knowledge transfer, not a demo on a slide. It can include RAG on your knowledge bases (answers from your documents, with cited sources and hallucination control), customer-service automation (chat, e-mail, internal helpdesk) and connecting AI to the systems you already run — ERP, CRM, DMS. We host the backend on AWS in a production-grade architecture, and add web and mobile apps where you need them. Scope and price are set after the audit and depend on the number of systems, integrations and data volume.

What you get

✓RAG on your knowledge bases — answers from your documents, with sources
✓Customer service automation (chat, e-mail, internal helpdesk)
✓Connecting AI to the systems you already run (ERP, CRM, DMS)
✓Cloud backend on AWS — production-grade architecture
✓Web and mobile applications where you need them

Level 03

Retainer — ongoing care

An AI retainer is ongoing care for a live system on a predictable monthly subscription — because an unattended system loses quality as data, models and regulations change. It covers AI governance (policies, roles, human oversight), RAG quality maintenance and monitoring (evaluations, knowledge-base updates), continuous AI Act compliance as systems and rules change, team training for the AI literacy the AI Act requires, and priority support and further development. This keeps answer quality, compliance and team skills up over time. Pricing depends on the number of systems maintained and the SLA level.

What you get

✓AI governance — policies, roles, human oversight
✓RAG quality maintenance and monitoring (evaluations, knowledge base updates)
✓Continuous AI Act compliance as systems and regulations change
✓Team training — AI literacy required by the AI Act
✓Priority support and system development

Entry offer

A safe AI PoC in 30 days

Instead of an “agent off the shelf in a week”, we offer the cheapest test: a proof it works, on one process and a real slice of your data, with limited budget exposure. We start from data and risk, not a demo — and 30 days is a PoC window, not a promise of production.

Stage 01

Discovery & shadow AI

Pick one process with a measurable result and inventory uncontrolled AI tools.

Stage 02

Data with permissions

A small, representative slice of data with mirrored permissions and a lawful basis.

Stage 03

Prototype + golden set

A proof on the slice and a reference set — real questions, expected answers, sources.

Stage 04

Evaluations, guardrails, logs

Quality measurement, hallucination control, action limits and a log of every answer with its source.

Stage 05

Unit cost & go/no-go

The cost of one operation on real data and an informed decision — with rejection criteria written down up front.

How it differs from “fast agents”

A fast off-the-shelf agent starts from a demo; we start from data, AI Act risk and measurable quality. We don’t promise “deployment in X days” — cost and schedule depend on the state of your data and integrations, and we set the risk class during the audit. An honest “no-go” is a cheap success: it costs a fraction of a failed deployment.

Before you start

When a PoC makes sense — and what it doesn’t promise

When does a PoC make sense?

When you have one repeatable process with a measurable result and data you are allowed to use. If it isn’t clear which process to pick, we start with an AI readiness audit.

What don’t we promise?

We don’t promise a fixed date or “production-ready after 30 days”. A PoC shows whether it’s worth deploying and at what cost — it does not replace AI Act compliance or production itself.

What do you keep after the PoC?

A working prototype on the slice, a reference set and evaluation results, a unit cost on real data, and a go/no-go decision with criteria — a basis to deploy, or to say an informed “no”.

Specialist services

Independent assurance and partner-led delivery.

Two focused ways to add AI quality, governance and delivery capacity without replacing the existing team.

AI Assurance & Governance

Golden set, evaluations, security tests, a risk report and evidence-based go/no-go criteria.

Partner co-delivery

A specialist AI workstream for software houses, integrators and consultancies — openly or white-label.

How we work

From diagnosis to a system that runs in production.

We run every deployment along the same repeatable engineering path. We don't start with the model — we start with data and risk. Every stage ends in a concrete result, not a presentation.

Discovery & audit

We map processes with AI potential, inventory “shadow AI”, classify systems by AI Act risk, and pick the use cases with the highest return.

Result: A prioritised roadmap, risk classification and an architecture recommendation.

Data & preparation (ETL)

We collect, clean and version your sources — documents, databases, files — into one controlled knowledge base. The hardest work happens before the model, not inside it.

Result: Clean, versioned data ready for RAG and integrations.

PoC on real data

We build a proof of value on your actual data before scaling anything. We check whether AI solves the real problem, not just whether it looks good on a slide.

Result: A working prototype and a fact-based decision: scale, or change the scope.

Evaluations & guardrails

We force answers strictly from cited sources, measure quality, and add hallucination control, data security and human oversight. The part a prompt trainer can't build.

Result: Measurable answer quality and security, documented before production.

Production deployment

We build the backend on AWS in a production-grade architecture, connect AI to the systems you already run (ERP, CRM, DMS), and add web and mobile apps where needed.

Result: A system running in production that people actually use.

Handover & care (governance)

We hand over documentation and train your team, and in the retainer we maintain RAG quality, governance and AI Act compliance as data, models and regulations change.

Result: Your team can use the system; quality and compliance hold over time.

Industries

Four industries we go deepest in.

In finance, manufacturing, healthcare and hospitality, we deploy AI where answers must come from controlled data and remain under human oversight. The common foundation is RAG or automation on AWS, source-level permissions, quality tests and governance matched to AI Act risk. Each industry still needs different data, integrations and accountability boundaries, so the dedicated industry pages explain the use cases, system limits and implementation priorities in detail.

Finance

Automation of analysis and customer service in a tightly supervised industry. AI Act and regulator compliance from day one.

Manufacturing

Technical knowledge bases, maintenance and documentation support. AI where downtime costs real money.

Healthcare

Knowledge-driven medical systems — the domain where we built our own product. Data security above all.

Hospitality

AI that grows guest numbers rather than cutting jobs: talk to each guest in their own language, tell the story of the place and its surroundings, reply instantly and personally. Multilingual podcasts and videos reach guests who have never been to you.

Looking specifically for RAG on company documents?

See how we deliver RAG

Need AI that does more than answer — that acts in your systems?

See how we build AI agents

Deploying AI under the EU AI Act? Start with a compliance audit.

EU AI Act compliance audit

Free tool

Is your organisation ready for a bounded AI PoC?

Answer 24 questions about process, data, shadow AI, ownership, risk and governance. The result points to the next step without claiming automatic conformity.

Open the AI Readiness Scorecard

Open worksheet

Model the full TCO of a GenAI solution — without fake audit pricing.

Structure inference, retrieval, data, evaluation, monitoring, operations and error costs. A separate brief captures what is needed to scope an audit individually.

Open the GenAI TCO worksheet

Open register

See every AI system and use case in one portfolio view.

Record owners, stage, data, integrations, supplier, volume, error impact and the next review date. Roles and risk remain explicitly marked for verification.

Open the AI system register

Open governance template

Separate accountability for an AI system from decision authority.

Complete a 7-stage × 7-role matrix and decision rights covering the decider, required input, evidence, escalation and response time. Every stage has exactly one Accountable.

Open the AI Governance RACI

Not sure which level to start with?

The answer is almost always the audit. It's the cheapest way to make AI decisions based on facts, not promises.

Book an AI audit