Short Answer
A regulated startup that does not yet have the cash to build a team can delegate the most repeatable and report-shaped work to role-based artificial intelligence workers: regulatory and quality drafts, literature and market research, competitive intelligence, and investment readiness. These workers do not decide; they report and recommend. No submission is filed, no document is sent, and no code is deployed until you approve it. Responsibility and the final say remain with the founder. The right framing is a support team that produces the output format of a person you cannot yet hire, but never signs on your behalf.
Serteser Consulting builds and hands over this role-based AI employee team, which runs locally and uses your own key, for medtech, biotech, and medical device startups. It is set up by a biomedical engineer who has lived through the world of MDR, FDA, TITCK, and ISO 13485.
Why a medtech-native team, not a generic AI agent
The market is now full of the promise of the "AI agent." Most are generic, domain-blind, and lack the one thing a regulated startup needs: subject-matter depth. A medical device founder's problem is not finding a model that can produce text. The problem is drafting an IEC 62304 software lifecycle document with the right headings, the right traceability matrix logic, and by asking the right questions.
This is where the difference lies. The value of this team is not in the model itself, but in the fact that the person who built it has taken a clinical artificial intelligence product to the field and has seen MDR and ISO 13485 processes from the inside. The agents' roles, prompt libraries, and checklists are shaped by this experience. A generic assistant tells you "regulation is important." This team prepares a reasoned draft on which class your software might fall under within the scope of Rule 11, and leaves the final assessment to a human.
A critical framing note: TITCK requirements are the national transposition of the European MDR regulation. This means that a team that establishes the MDR logic uses the same foundation on both the TR and EU sides. This is a serious lever for a solo startup, but it is never used with overclaiming. Classification is framed carefully in light of the MDCG guidances, and the final decision is yours.
Which work an AI employee can carry
The common feature of delegable work is this: repeatable and report-shaped. Not a creative leap, but disciplined production. Here are the realistic roles.
Regulatory and quality worker. Intended-use statement drafts within the MDR, FDA, and TITCK framework, ISO 13485 quality management system gap analysis skeletons, IEC 62304 software documentation templates, risk management (ISO 14971 logic) table drafts. All drafts, all awaiting human approval. Filing a submission or making a claim is never the agent's job.
Research worker. Literature review, clinical evidence mapping, summarizing standards and guidances, compiling the regulatory history of similar devices. It turns the reading load that eats up the founder's days into a source-linked and traceable report.
Competitive intelligence worker. Structured briefings on competitor products' public positioning, regulatory statuses, claims, and gaps. It offers the founder a concrete table for the question "where are we different, where have we fallen behind."
Investment readiness worker. Compiling market size research, drafts of competitor landscape slides, preparation notes for due diligence questions, data room content lists. Every figure that goes to the investor still passes through your verification.
These four roles give a founder who does not yet have a payroll the output format of a full-time expert. Not their productivity, but their format and first draft. What closes the gap in between is your expertise and your approval.
The boundary: it reports and recommends, it does not decide
The most important feature of this team is not what they can do, but what they do not do. The agents are not autonomous decision-makers. No document is sent, no submission is filed, no code is deployed, and no permission is granted until a human operator approves it. In every flow there is a gate, and at that gate, you stand.
This design choice is not marketing but a necessity in a regulated field. Responsibility remains with the founder. You are the one who signs before a regulator, gives a figure to an investor, and makes a promise in a hospital pilot. The agent carries the heavy work in front of these decisions, not the decision itself.
A second design principle is data sovereignty. The team runs locally and uses your own API key. Your company's regulatory data, clinical documents, and sensitive information stay on your own machine. We do not host this sensitive data. For a regulated startup, data residency and privacy are not a feature added later, but a constraint that comes from the start.
Finally, this is not a self-service SaaS. The team is built and handed over by Serteser Consulting: configuring the roles according to your product, shaping the prompts and checklists with domain knowledge, and delivering a structure the founder can use with confidence. The value is not in the software itself, but in the experience that places it into a regulated context.
If you would like to talk about building this first team for your startup, you can review the professional consulting services.