AI as a Service
Regulatory AI,
validated before use.
Tools, agents and implementations built around your products, your processes and the regulations you work under. They run in your environment, on your data, and a qualified person reviews what matters.
Built around your process.
Running on your data.
Released by a person, not a model.
Most AI in regulatory teams arrives as a chat window nobody has validated, holding data that should never have left the building. We build it the other way round: scoped to a defined use, tested against it, and kept compliant as models and regulations change.
What we build
Tools, agents
and implementations.
Every build starts from a piece of regulatory work that is costing your team time. These are typical examples; yours will be shaped around your range.
Tell us what’s slowing you downTypical builds
Tools for everyday regulatory work
Claims checked against your evidence, classification support, documentation sections drafted for review, and search across your own files and standards.
Agents that run routines for you
Regulatory change monitoring for your range, supplier certificate and agreement chasing, complaint triage for post-market surveillance, and evidence gathered ahead of an audit.
AI built into the systems you use
Connected to your quality system, document control and data, and wired into Slack, Monday.com, Airtable, Microsoft 365 or anything else with an API.
For device, IVD and cosmetics teams, including borderline products.
Four things true of every build
AI you can use.
AI you can defend.
- Validated
- Each tool is validated for its intended use before it goes live. You get the written record, ready for an auditor.
- Reviewed
- The AI drafts, checks and flags. A qualified person reviews and releases anything with regulatory consequence, and the review is recorded.
- Yours
- It runs in your accounts, on your data. Your data never trains a model or reaches another client, and you own the outputs and the IP.
- Assessed
- Every build is assessed against the EU AI Act risk categories and UK/EU GDPR, and the assessment is kept with the validation record.
A worked example
One build you
can click through.
A product, supplier and compliance system for a multi-product range: the portfolio map, supplier qualification, claims traced to their evidence and the board the team runs from.
Explore the demonstrationYour AI
AI as a Service
Built around your processes, running in your environment, kept validated by us. For when the work is specific to you.
Discuss your AIOur platform
SMEDTEC OS
Our regulatory operations platform, ready to use. For when shared tooling fits the work.
Explore the platformHow it works
A scoped build,
then a service that keeps it compliant.
- 01Scope
The work, the decision points, the data and the intended use, agreed before anything is built.
- 02Build
Tools and agents built around your process, in your environment, connected to your systems.
- 03Validate
Tested against the intended use, assessed for the EU AI Act and GDPR, and written up.
- 04Deploy
Into your team's daily work, with the human review step built in.
- 05Run
Monitored, re-validated when models, regulations or your use change, and improved over time.
You pay for the build, then for the service that keeps it running and validated. The infrastructure it runs on, such as the model provider, hosting and database, sits in your accounts and is billed to you directly, so you always see what it costs to run.
Questions
Before you
bring in AI.
Who owns the data, the outputs and what you build?
You do. Your AI runs in your own accounts. Your data, everything it produces and the IP in what we build belong to you. Your data is never used to train models and never shared with other clients.
Why a service, not a one-off build?
Because AI in a regulated business does not stay validated on its own. Models are updated, regulations change and your use of the tool moves on. After the build we monitor each tool, re-validate it when something material changes and keep improving it, so it stays fit for use and defensible at audit.
What does validated mean here?
Each tool is tested against a defined intended use before it goes live, and you receive a written validation record you can show an auditor. It is the same discipline your quality system already expects for software used in the QMS, repeated whenever the tool, the model or its use changes.
Does the AI make regulatory decisions?
No. It drafts, checks, searches and flags. A qualified person reviews and releases anything with regulatory consequence, and that review is recorded. Accountability stays with people.
How do you handle the EU AI Act and GDPR?
Each build is assessed against the EU AI Act risk categories and UK/EU GDPR before it is deployed, and the assessment is kept with the validation record. If the assessment changes what we should build, we tell you before we build it.
How is this different from SMEDTEC OS?
SMEDTEC OS is our regulatory operations platform: you use what we have built for every team. AI as a Service is your AI: tools and agents built around your own processes, running in your environment, which we keep validated and running for you.
Ready to scope
your first build?
A scoping call defines the use, the data and the validation it needs.