Knowledge management in the pharmaceutical industry
By Elium
ICH Q10 names knowledge management as an enabler of the pharmaceutical quality system. What that means in practice, and why documentation alone is not enough.
Few industries depend on knowledge as literally as pharmaceuticals. What a company knows about a formula, a process, a piece of equipment and its history is what allows the product to be made the same way twice, defended in an inspection, and transferred to another site without starting the science again.
That is also why knowledge management, in this industry, is a regulatory term. It appears by name in the quality guidance the industry is inspected against.
What the regulator means by knowledge management
The International Council for Harmonisation (ICH) sets the technical standards that drug regulators in Europe, the United States and Japan have adopted. Its guideline Q10, on the pharmaceutical quality system, names two enablers of that system: knowledge management and quality risk management. Its definition, in section 1.6.1, runs to a single sentence:
Knowledge management is a systematic approach to acquiring, analysing, storing and disseminating information related to products, manufacturing processes and components.
The same section is equally specific about where that knowledge comes from. It lists prior knowledge, whether public or internally documented; pharmaceutical development studies; technology transfer activities; process validation studies across the product lifecycle; manufacturing experience; innovation; continual improvement; and change management activities.
Two things follow from that list, and both are easy to miss.
First, the scope is the whole lifecycle, from development through commercial life to discontinuation. Knowledge that only exists inside the development team has not met the definition.
Second, most of the sources named are activities, not documents. Manufacturing experience, innovation and continual improvement produce knowledge continuously, in the heads of the people doing the work. The guideline expects that knowledge to be acquired and disseminated. Existing somewhere is not the test. ICH Q10 dates from 2008 and the definition has not needed changing since.
What pharmaceutical knowledge is made of
Pharmaceutical knowledge combines chemistry, microbiology, analytics, materials science, process technology and toxicology, with cell engineering and genomics on top of that in advanced therapies. The scientific core is only half of it. To reach a patient, it has to be paired with project management, risk analysis, budgeting and a working understanding of regulatory expectations.
The difference between data, information and knowledge does real work here, because the three need different handling. A batch record is data. A trend report is information. Knowing which deviation patterns matter, and why the process was changed three years ago, is knowledge, and it is the part that leaves when a person does.
Why documentation alone is not knowledge management
Every pharmaceutical company already has a documentation system, built around good manufacturing practice: procedures, work instructions, test methods, specifications, reports, and the training records that go with them. It is rigorous, audited, and necessary.
It is also not a knowledge system, for three reasons.
It records decisions, not reasoning. A specification says what the limit is. It rarely says which experiments set it, which alternatives were rejected, or what happens near the limit. When the next change control opens, the reasoning has to be reconstructed from memory.
Traceability breaks at the change. When a specification or a process step changes, the document is revised. The reason for the change, and what was learned along the way, often stays with the people who were in the room.
It is filed for the auditor. A document stored correctly under a procedure number is findable by anyone who knows the number. A process engineer asking a question in ordinary words comes up empty.
The result is familiar: a company holds an enormous amount of documentation and still cannot answer, quickly, what it already knows about a given problem.
How knowledge actually moves in a pharmaceutical company
In practice, knowledge moves through a small number of channels, and most of them are informal:
- Controlled documentation: procedures, methods, specifications, reports
- Training programmes tied to good practice requirements
- Routine and non-routine meetings, including deviation and change control reviews
- Technology transfer between development and manufacturing, and between sites
- Seminars, conferences and professional bodies
- Team updates, internal demonstrations and presentations
- Mentoring, shadowing and paired work with experienced staff
Only the first two are usually managed. The rest carry a large share of what the organisation knows and leave almost no searchable trace. That imbalance is the practical knowledge management problem in pharma, and scale makes it worse. Guidance and expectations are updated continuously: by the European Medicines Agency, the US Food and Drug Administration (FDA), ICH, the Pharmaceutical Inspection Co-operation Scheme, the World Health Organization and the pharmacopoeias. Each update usually arrives as an email to a distribution list. Time-sensitive information reaches an inbox, not the people who need it three months later.
What AI changed in 2026
Pharmaceutical companies are now putting AI into development, manufacturing and quality work, and the regulators have responded quickly. In January 2026, the FDA and the European Medicines Agency published ten guiding principles of good AI practice in drug development. Two of them are knowledge management problems under another name: data governance and documentation, and life cycle management.
For a quality or knowledge lead, the consequence is concrete. An AI assistant answering questions about a process is only as good as the content it reads. If the corpus contains three versions of a method with no indication of which is current, the assistant will answer confidently from the wrong one, and it will do so at scale. Governance, ownership and review dates stop being knowledge management hygiene and start being a condition of using AI at all. We have written separately on why AI projects fail on the knowledge foundation rather than on the model.
What good looks like
Servier, France’s leading independent pharmaceutical group, had the problem in its competitive intelligence function: signals sat in departmental silos, spread across separate document sites, newsletters and research folders, and rarely reached the people setting strategy.
The rebuild put that intelligence in one place with owners and structure. Servier now runs 1,000 active users across 140 countries on a single intelligence platform, with more than 9,000 assets centralised and searchable, and executive briefings are produced quarterly from the platform’s own content. Usage more than tripled over six years as the platform spread to every therapeutic area.
Nothing in that outcome required new science. It required deciding who owned what, and giving people one place to look.
Where to start
- Pick one lifecycle handover, usually technology transfer between development and manufacturing, or between two sites. It is where knowledge loss is most expensive and most visible.
- Name owners before you migrate anything. A space with no owner becomes an archive within a year.
- Capture reasoning next to the record. Add the why beside the what: rationale for a limit, alternatives rejected, what was learned during validation.
- Give regulatory updates a destination. A shared space with an owner and a review date beats a distribution list, because it is still there in six months.
- Set a review cycle and hold it. In a regulated environment, an out-of-date answer is a finding waiting to happen.
If you want to see what that looks like on one platform, book a demo and bring a question your teams keep asking.
Common questions about knowledge management in pharma
ICH Q10 defines it as a systematic approach to acquiring, analysing, storing and disseminating information related to products, manufacturing processes and components. Alongside quality risk management, it is one of the two enablers of the pharmaceutical quality system, and it applies across the whole product lifecycle, from development to discontinuation.
Because product quality depends on knowledge that is rarely written down in full: why a limit was set, what was learned during validation, which deviations mattered. That knowledge is needed at every technology transfer, change control and inspection, and it leaves the company when experienced staff do.
Section 1.6.1 defines it and lists its sources: prior knowledge, whether public or internally documented; pharmaceutical development studies; technology transfer activities; process validation studies over the product lifecycle; manufacturing experience; innovation; continual improvement; and change management activities. The guideline places it alongside quality risk management as an enabler of the quality system.
No. A good manufacturing practice documentation system records decisions and is organised for compliance and audit. Knowledge management also has to carry the reasoning behind those decisions and make it retrievable by someone asking a question in ordinary words, months or years later.
In practice it is a governed platform where product and process knowledge is published in structured spaces, each with a named owner and a review date, searchable across content and attachments, with permissions that let confidential development and manufacturing knowledge be shared with the right people. It sits alongside the document management and quality systems rather than replacing them.
It raises the requirement on the underlying content. The FDA and the European Medicines Agency published ten guiding principles of good AI practice in drug development in January 2026, including data governance and documentation, and life cycle management. An AI assistant reading an ungoverned corpus will give confident answers from superseded documents, which is why ownership, versioning and review cycles come first.
Related reading
Ready to put your team's knowledge to work?
See it on your own knowledge.
See how Elium gives your teams and your AI agents one source of answers they can trust.