Regulatory affairs is the work nobody notices until it goes wrong. A clinical evaluation report can run hundreds of pages. A dossier may contain years of submissions, labels, authority questions and product changes. The documents must be current, internally consistent and traceable to evidence. If the work drifts, a medicine or device can miss a market deadline - or struggle to remain on the market at all. Celegence built a business in this paperwork, then discovered something more interesting hiding inside it: a product roadmap.
The Chicago-area company was founded in 2017 by sisters Sonia A. Veluchamy and Punya Abbhi. Veluchamy had spent close to a decade as managing director of life-sciences software company ArisGlobal; Abbhi came from management consulting and technology implementations. Their original proposition was not a chatbot. It was tech-enabled regulatory service: combine an onshore-offshore talent model with people who understood pharmaceutical and medical-device rules, then use software to make delivery less expensive and less painful.
The customer wrote the expansion plan
Celegence first helped clients navigate changing requirements such as IDMP and the European Union's Medical Device Regulation. Then a pharmaceutical customer asked whether the same offshore team writing pharma documents could also prepare clinical evaluation reports for its devices. That request pulled the young firm deeper into medtech. The opportunity was concrete: a paying customer exposed adjacent demand, and Celegence followed it.
Today the company supports regulatory strategy and intelligence, publishing and submissions, RIMS and data work, labeling, technical and medical writing, and lifecycle maintenance. Its customers range from a manufacturer with one marketed product to global pharmaceutical, medical-device and in-vitro diagnostics groups. Acumed and Navignostics have been publicly named. Many larger clients remain anonymous, which is ordinary in this corner of health care. In late 2025, Celegence said it had worked with more than 100 pharmaceutical and device companies and employed more than 175 people worldwide.
CAPTIS began with the boring bit
The flagship platform, CAPTIS, began when Celegence's own team was assigned recurring clinical evaluation reports under EU MDR and IVDR. Writers had to search literature, screen articles, pull data, manage references and update long documents on a fixed cadence. The founders asked a wonderfully unglamorous question: which part does the team find monotonous and time-consuming? Literature review came first.
That starting point matters. Celegence did not ask a focus group to imagine an AI future. It observed paid production work, found a task with visible inputs and outputs, and let medical writers define the nuisance. Engineers and writers then joined the same workshops. CAPTIS expanded from literature review into data extraction, reference management, templates, document authoring, dashboards and post-market surveillance workflows. It can support clinical and performance evaluation reports, periodic safety updates and related MDR or IVDR documents.
The services-to-software loop
Celegence publishes time-saving studies that claim a 62 percent reduction in data-gathering time, nearly 30 percent less time on a systematic literature review, and 45 percent savings in searching and archiving full-text articles. These are company studies, not universal benchmarks. Still, they reveal the right unit of product thinking. The pitch is not “AI transforms compliance.” It is fewer hours spent locating, deduplicating and retyping evidence.
The first failure was useful
The cleanest proof of Celegence's judgment is a feature it did not ship. In an internal proof of concept, the team tested AI for mapping external adverse events to internal device-risk documentation. A subject-matter expert checked the results. They were not accurate enough. Celegence held the capability back and decided it needed more data before the system could be trusted.
“It wasn't accurate enough to push forward into the solution yet.”Punya Abbhi on an AI risk-mapping proof of concept
Another concern changed the product more visibly. Users wanted to know where an AI extraction came from inside an article. The question was not merely whether the answer sounded right; reviewers needed to inspect the evidence. Celegence responded with retrieval-augmented generation and source-level traceability. The writer can see the passage, accept it or edit it. This is a better picture of regulated AI than the usual robot-replaces-professional cartoon: software prepares, organizes and points; a qualified person remains accountable.
That boundary is also the company's difference from a general-purpose model or a pure technology vendor. Celegence's software is trained by current workflow knowledge because the company still performs the work. Against a traditional consultancy, the tools provide reusable infrastructure and a subscription product. Against an internal team, Celegence offers specialists and global delivery capacity without requiring the manufacturer to hire every niche skill. The compromise is deliberate. Services make the model less tidy than pure SaaS, but they also generate product feedback, trust and operating data.
Three products, one document lifecycle
CAPTIS gets the AI attention, but the acquired Qdossier business added a second useful layer. Dossplorer is a cloud eCTD viewer and dossier manager. It imports submissions, searches across them, supports comparisons and manages authority correspondence. Abbhi has described a flavoring-component search that might once have taken two weeks becoming a query measured in minutes. Dosscriber supplies structured Microsoft Word templates for eCTD-ready documents. Together, the products cover evidence collection, authoring, review, submission history and reuse.
The Qdossier acquisition in 2022 also explains what changed Celegence's mind about building everything organically. After five years of expanding services, the founders wanted a stronger European footprint and the ability to support companies of different sizes globally. Buying the Netherlands operation brought regional expertise and mature dossier tools at once. The price was not disclosed. Neither are standard subscription fees, outside funding or a valuation. Celegence makes money through consulting projects, managed services, dedicated delivery resources and standalone software subscriptions. The founders say nearly 10 percent of profit is reinvested into the technology suite each year.
What a founder can steal
First, begin with a service if the domain is complicated and buyers cannot specify the software they need. Paid work reveals the actual sequence, exceptions and handoffs. Second, ask operators which task they dread, not which feature they want. Third, pair the engineer with the person currently responsible for the output. Fourth, express value in hours, revisions and error reduction. Finally, establish a rejection rule. If an expert cannot verify the answer or the evidence is thin, the automation waits.
The copyable operating recipe
- Choose a recurring workflow with expensive, structured inputs.
- Use service delivery as continuous product discovery.
- Make every generated claim traceable to its evidence.
- Test internally before asking customers to absorb the risk.
- Keep human approval wherever judgment changes the outcome.
There are conditions where this approach will not work. Low-volume compliance may not justify a specialized platform. A company with clean data, mature systems and enough internal specialists may prefer to keep the work in-house. Generic tools may be sufficient for low-risk document assembly. And an onshore-offshore delivery model depends on strong governance, security, communication and quality controls; without them, lower labor cost becomes rework. AI is also a poor shortcut when the task is novel, the evidence is sparse or nobody qualified has time to review the result.
The market position is less cinematic than “AI for health care,” and more durable for it. Celegence sits between regulatory consultancies such as ProductLife Group, Freyr, PharmaLex and larger outsourced-service firms, and software platforms from companies such as Veeva, Generis, Ennov and ArisGlobal. Its wager is that manufacturers want both the tool and someone who knows why the tool might be wrong.
That wager has produced visible traction: a TOPRA Innovation Award for CAPTIS, a 2025 Inc. 5000 ranking at No. 3,367 with 117 percent three-year growth, and a top-10 global IVD manufacturer selecting the platform after a six-month competitive evaluation. The revealing detail in that sale was not the model. Celegence configured the customer's environment, tailored training, held weekly check-ins and changed features during the trial. Enterprise software, even with AI inside, remains a human business.
Celegence's lesson is not to make compliance exciting. It is to respect why the boring parts exist. The forms, references and review trails are evidence that someone checked. Automating them has value only when the checking becomes easier, not invisible. The company started by selling expert time, found the repetitive work inside it, and built software that knows when to hand the pen back.