Medical science liaisons carry advanced degrees. A typical MSL field team includes pharmacists, PhDs, and physicians. They have spent months in product training. They know the mechanism of action, the pharmacokinetic profile, the primary and secondary endpoints from the pivotal trial, and the approved indication down to the specific patient population language.
The knowledge gap in MSL teams is not a knowledge deficit. It is a retrieval problem, and treating it like a knowledge problem leads to training programs and interventions that solve the wrong thing.
What MSL Training Programs Are Built To Do
Standard MSL onboarding programs are designed around knowledge acquisition: learn the product, learn the clinical data package, learn the competitive landscape, learn the compliance requirements for scientific exchange. Training lasts weeks to months. At the end, the MSL is evaluated on whether they can discuss the clinical program accurately and within label.
This model makes sense for the knowledge that needs to be internalized: scientific background, product mechanism, high-level trial results, compliance principles. An MSL who has to look up what CDK4/6 inhibition means during a conversation with an oncologist is not prepared for that conversation.
But the model breaks down for the long tail of clinical detail. A single clinical study report for a complex oncology product can run to thousands of pages. The prescribing information for a product with multiple approved indications might have 40 or 50 pages of dosing guidance, warnings, and clinical study summary tables. A publication package for a product with five years of post-approval data includes dozens of papers, poster presentations, and secondary analyses. The expectation that an MSL should hold all of this in working memory, ready to retrieve on demand during a physician conversation, is not realistic and probably not desirable.
The Retrieval Problem That Training Does Not Solve
Consider the kind of question that a KOL asks in a substantive scientific exchange. Not "what is the mechanism of action" but "in the subgroup analysis from the extension study, what was the observed rate of treatment-emergent peripheral neuropathy in patients who had received prior taxane therapy?"
That question has a specific answer in a specific table in a specific document. An MSL who has done their job knows that this information exists and has a strong prior about where to look. But they probably cannot retrieve the exact number from memory, and they should not try to. Giving an approximate number from memory in a scientific exchange when an exact number is in the document is an invitation for a compliance problem. Either the approximation is wrong, or the physician asks for the source and the MSL cannot provide it.
The right answer in that moment is "let me find that for you in the CSR" or, if the meeting was well-prepared, having already found it before the meeting and knowing the exact page. Neither of those outcomes requires the MSL to have memorized the table. Both require fast, reliable access to the right document.
How Fragmented Document Systems Make the Problem Worse
The retrieval problem is compounded by where medical affairs documents live. Clinical study reports are typically in Veeva Vault or a similar regulatory content management system. Prescribing information may be there too, or in a separate document management system. Publication packages, reprint management, and key data slides are often on SharePoint, with different folder structures per product and per region. Standard response documents live in a medical information system. Training materials and field resources are somewhere else again.
An MSL preparing for a KOL meeting who needs to pull data from four of these sources is navigating four different systems, four different search interfaces, and four different file-naming conventions. None of these systems are indexed together in a way that lets the MSL search across all of them with a single question.
The fragmentation is not usually the result of poor planning. It reflects the organizational reality that different functions (regulatory, medical information, publications, field medical) own different document types and maintain them in systems optimized for their own workflows. The MSL is downstream of all of them and inherits the complexity.
What the Research on Expert Knowledge Retrieval Tells Us
There is a body of research on expert knowledge retrieval in high-stakes professional contexts, including medicine, law, and aviation, that describes a consistent pattern. Experts are not better at storing more information. They are better at knowing where to look and at recognizing when they do not know something well enough to rely on memory. The performance gap between expert and novice professionals often comes down to metacognitive accuracy: knowing the limits of your own recall.
A senior MSL with five years on a product knows they cannot recite the exact p-value from a subgroup analysis without looking it up. They know which document it is in and roughly where. They know how long it will take to find it. A new MSL on the same product may not yet have the metacognitive calibration to know they are working from an imprecise memory rather than an accurate one.
The implication for MSL support tools is that the goal is not to give MSLs a better memory. It is to give them fast access to precise information so they are not forced to choose between admitting they need to look something up (which creates friction) and citing a detail from memory that might be approximately right (which creates compliance risk).
Fast Retrieval as a Complement to Expert Knowledge
The framing we use internally at Argon is that source-grounded retrieval is a complement to expert knowledge, not a substitute for it. An MSL who does not understand the clinical program cannot ask a useful question of a retrieval system. They will not recognize whether the returned answer makes sense in context. They will not know whether the retrieved passage answers the physician's actual question or a related-but-different question.
Expert knowledge is what makes retrieval useful. Retrieval is what makes expert knowledge deployable in real time, under the time pressure of a physician interaction, without the MSL having to carry thousands of pages of clinical detail in working memory.
We are not saying memorization has no value. Some knowledge should be internalized: the mechanism of action, the approved indication, the primary endpoint results, the key safety signals. These are the things that come up in every conversation and where an MSL who has to pause and look them up signals underpreperation. But the long tail of clinical detail, the subgroup analyses, the pharmacokinetic parameters by patient population, the exact label language for edge case dosing scenarios, belongs in a retrievable index, not in working memory.
The Productivity Implication for Medical Affairs Leaders
From a medical affairs leadership perspective, the relevant question is not "how much does our MSL team know?" but "how quickly can our MSL team find the right answer when a physician asks a question that is not in the talking points deck?"
That second question is harder to assess in a training evaluation but more predictive of field performance in substantive scientific exchange. The best KOL interactions are not ones where the MSL recites rehearsed content. They are ones where the physician asks something unexpected and the MSL has the tools and judgment to address it accurately.
Closing the knowledge gap in MSL teams means building systems that support expert retrieval, not programs that expect expert memorization. Those are different problems with different solutions, and the organizations that recognize the distinction tend to invest in the right places.