Author:
To come to important decisions, life sciences leaders increasingly rely on insight from a widening array of sources, which are being created at unprecedented speed. Between expert interviews, surveys, advisory panels, AI-generated synthesis, and third-party data, the volume and variety of data can be a boon to discovery, giving teams richer research outcomes and greater scale, which in turn expands the imagination for what’s possible in a given R&D program.
But the trend also makes it harder to ensure and audit the quality of data. Unqualified participants, unverified credentials, undisclosed conflicts, and fraudulent respondents can sneak into the workflow and compromise research, which comes with high stakes in healthcare. Flawed inputs can skew pricing, forecasting, positioning, and launch decisions worth millions, not to mention the human implications of drugs or technology designed on flawed research. In a society that’s increasingly concerned about the influence of AI on our life decisions and future, this represents an area where trust and safety are paramount.
So, how do teams establish confidence in the respondents and sources behind their research?
It’s a problem I’ve grown familiar with during my many years helping enterprise clients make decisions with far-reaching implications. Those decisions require a careful mix of expert reasoning, domain-specific context, noise reduction, and confidence. Today, it’s something I’m thinking about carefully in my new role as Chief Research Officer at GLG, where we’re working to scale access to trusted human expertise through a variety of mediums. I’m particularly interested in how we can enrich the quality of our global network of experts – including the most hard-to-reach populations – while making sure our partners can tap into its full value. In addition, we’re working on new techniques to leverage the graph of interconnected insight from that network.
While we’re generally known for our one-on-one expert calls, GLG has built out an array of research tools our clients in the life sciences turn to, including large-scale surveys and in-depth engagements. As we watch the space, especially with an eye for compliance and safety to stay ahead of risks, we’ve noticed data integrity often breaks down due to participants bringing inflated, outdated, or altogether fraudulent credentials.
The risks aren’t necessarily new, but AI has complicated them and raised the stakes as adoption reaches deeper into workflows. Fake respondents are now cheaper to produce and more sophisticated, and, with data flowing downstream at a seemingly instantaneous clip it can be harder to disentangle the implications of bad dataLeaders will need to lean on a combination of intuition, critical thinking, and methodological tactics to tell whether something is off in a survey program. Now is the time to sharpen the expertise on your team as AI enablement scales to ensure no one defers too much authority to the outputs of increasingly powerful machines.
Vigilance is important, but it’s even better to have confidence in your data providers in the first place. Trust isn’t just about assurances, brand, and track record. It’s about infrastructure and process.
What we call provenance, for example, has become a key motif in the research space and represents a structural defense against such risks. It’s a fancy word for a simple concept: how far back can you trace an insight to its precise source? Life sciences leaders count on provenance as much as anyone – and it matters more than ever given the proliferation of AI in their research and decision making.
When it comes to human data, the profile of the source has always been inextricable from the quality of insight, especially in a scientific or clinical field. In the AI era, anonymity carries even more risk, because a single data point can be summarized, blended, and passed along until no one remembers where it came from.
This is why I believe provenance should become table stakes for any research provider, even though most in the space still consider it a premium offering. In practice, that means two things. First, respondents should have to clear identity, credential, and compliance screens before they take part in an engagement. And then attribution – researchers should be able to trace data back to a vetted source, even when it’s one among many in a survey. We hold ourselves to that standard at GLG, but it’s one every provider should be able to meet today, and every buyer should be able to test.
Leaders should scrutinize research partners on this point. How rigorous are they about ensuring respondents are who they say they are – and how precisely can they trace information back to the source? Are the experts in the network current in their knowledge? Are they engaged in relevant debates with other experts in their field? Counterpoints can be useful.
The payoff is worth the extra effort and cost. Strategy teams can decide which assets to pursue knowing the experts behind their diligence are who they say they are and product teams can shape and price a new therapy with input from the right clinicians and KOLs – and defend those choices to leadership. Confidence will enable the speed and scale that teams need to compete in the new information landscape, especially as new LLMs and domain-specific small language models enter the field. GLG is taking aggressive steps here with our new model eval work via GLG Frontier, where our validated experts will train and test new models to ensure they produce high resolution, expected results – for life sciences teams as well as the patients, consumers, and communities they serve.
