← All posts

How I use Claude to run operations for a regulated health-tech company

· ai-operations, claude, volumetryx

Short answer: Volumetryx.ai is my company — a clinical trial imaging platform running under FDA 21 CFR Part 11, HIPAA, and GDPR. Despite the compliance weight, I use Claude to handle five specific operational workflows that would otherwise require a team of four or five full-time operators. The five: SOP drafting, compliance documentation, customer outreach personalization, interview synthesis, and product roadmap prioritization. Claude does not make regulatory decisions, handle PHI directly, or sign off on compliance artifacts — but it can do 70% of the surrounding operational work, and that’s enough to run the company lean.

Here is what each workflow looks like, where Claude helps, and where it stops.

Why this matters beyond health-tech

Most founders I talk to think “AI for operations” means chatbots or marketing automation. That’s a small slice of what’s actually possible. The interesting question isn’t “can AI do the obvious tasks” — it’s “can AI handle the serious operational work in an environment with real consequences?”

I’m using Volumetryx as the test case because if AI-augmented operations works under FDA 21 CFR Part 11, it works for a digital agency or a DTC brand running on Shopify. The regulated environment is the harder case. The answer is yes, with specific limits.

Workflow 1: SOP drafting

Standard Operating Procedures are a core compliance artifact for any regulated health-tech company. Volumetryx has SOPs for data handling, user access, system validation, incident response, change management, vendor management, and a dozen other topics. Each is 5–15 pages of structured documentation.

Writing an SOP from scratch takes a senior operator 6–10 hours. Claude drafts a complete first version in 15 minutes from a conversation about the process, the regulatory requirements, and the internal structure. I then edit the draft, verify the regulatory references, and approve.

Time saved per SOP: 5–8 hours. For a company that writes 10–20 SOPs a year, that’s 50–160 hours of senior operator time recovered.

Where Claude stops: It doesn’t make regulatory interpretation decisions. If an SOP section touches a specific 21 CFR Part 11 requirement, I verify the interpretation myself or check with a compliance advisor. Claude is a drafting accelerator, not a regulatory authority.

Workflow 2: Compliance documentation

Beyond SOPs, regulated health-tech generates a steady stream of compliance artifacts: validation protocols, risk assessments, training records, audit logs, change requests, deviation reports, CAPA documentation. Most are structured documents with predictable sections and cross-references.

Claude reads the context (existing artifacts, process description, regulatory section) and drafts the new artifact in the correct format. For a deviation report, for example: I describe what happened in plain language, Claude produces the structured report with root cause, impact assessment, corrective actions, and references.

Time saved: 60–70% on documentation work. What used to take a full day takes 90 minutes.

Where Claude stops: Final sign-off on compliance documents is always mine. Claude does the structure and the first draft, a human verifies and approves.

Workflow 3: Customer outreach personalization

Volumetryx targets oncology biotechs running Phase I/II clinical trials. It’s a narrow, technical audience. Generic outreach doesn’t work — prospects need to see that you understand their specific trial design, therapeutic area, and timeline.

Claude handles the personalization at scale. For each prospect: I provide the company name, pipeline information, and recent news. Claude reads the context, drafts a 2–3 sentence personalized opener that references their specific trials, and outputs a version ready for the outreach template.

Time saved: From 15 minutes per message (manual research + writing) to 2 minutes (prompt + light edit). For 40 outreach messages a week, that’s 8–10 hours recovered.

Where Claude stops: I don’t send AI-generated messages without reading them. Every message gets a human check before it goes out. The quality of the research layer depends heavily on what I feed into the prompt — Claude is not a replacement for actually knowing the market.

Workflow 4: Interview synthesis

Customer development is core to Volumetryx. I run interviews with radiologists, clinical operations leads, and biotech program managers. Each interview is 45–60 minutes and generates 15–25 pages of transcript.

Pre-Claude, synthesizing a week’s interviews (5–8 calls) took me a full day of reading transcripts, extracting patterns, and writing up findings. With Claude: I load transcripts into a project, ask for pain-point clustering and objection analysis, and get structured output in 20 minutes. I verify the patterns against my own memory, add context Claude couldn’t see, and the write-up is done in an hour.

Time saved: 6–7 hours per synthesis cycle.

Where Claude stops: The output is only as good as the prompt and the context. If I ask for “insights” I get generic business-book output. If I ask for “verbatim quotes that show the pain pattern around data handoffs” I get usable material. The skill is in the prompt, not the model.

Workflow 5: Product roadmap prioritization

Volumetryx has more feature ideas than capacity to build them. Prioritization is a weekly exercise: balance customer needs, technical debt, compliance requirements, and strategic direction.

Claude helps me structure the tradeoffs. I load: current roadmap, recent customer feedback, known technical debt items, and compliance requirements. I ask for a scoring model (impact, effort, risk, strategic fit) and a ranked list with reasoning. Claude produces a structured analysis that I use as a starting point for the actual decision.

Time saved: 3–4 hours per prioritization cycle. More importantly: better decisions because nothing slips through the cracks.

Where Claude stops: I make the call. Claude surfaces tradeoffs and structures the analysis. The judgment — which tradeoffs matter, which customers are strategic, which technical debt is truly blocking — is mine.

What Claude cannot do in a regulated environment

This is the important part. AI in regulated operations has hard limits, and pretending otherwise is how you end up with a 483 observation or a data breach.

  • No PHI in prompts without proper BAA and infrastructure. Claude.ai under a standard subscription is not HIPAA-compliant out of the box. You need a business agreement and infrastructure that handles PHI correctly — typically Claude via Anthropic’s enterprise tier with a BAA, or through a HIPAA-compliant cloud layer.
  • No final compliance sign-off. Claude drafts. Humans verify and approve. Every regulated artifact that goes to an auditor has a human name on it.
  • No clinical judgment. Claude does not interpret trial data, suggest clinical protocols, or make medical decisions. Those are scope boundaries, not technical limitations.
  • No undocumented workflows. Every Claude-assisted process at Volumetryx is documented in an SOP. Auditors want to know what tools you use, what they do, and how you validate the output. “We use AI” is not an answer — “here is the SOP for AI-assisted SOP drafting” is.

What this means for your (non-regulated) business

If I can use Claude to run five core operational workflows at a health-tech company under FDA 21 CFR Part 11, your agency or your SaaS or your DTC brand can use the same patterns without the regulatory overhead. The workflows translate directly:

  • SOPs → playbooks. Drafting playbooks for agency client delivery, SaaS onboarding, DTC warehouse operations.
  • Compliance docs → internal documentation. Change management, decision logs, process documentation.
  • Outreach personalization → sales outreach. The exact same pattern, minus the regulatory constraints.
  • Interview synthesis → customer research. Same workflow, any industry.
  • Roadmap prioritization → feature prioritization. Same scoring, same tradeoffs.

The ceiling isn’t the technology. It’s whether you’re willing to restructure how operational work gets done. Most businesses aren’t — which is why this is still an unfair advantage for the ones that are.


Want to see what Claude-augmented operations looks like for your business? Book a free 30-minute operations call. I’ll walk you through the specific workflows that fit your stage. Book here.