A few months ago, I met someone at a life sciences conference who told me about her new assignment: Use AI to automate her team’s weekly documentation. Leadership was enthusiastic. And at least in theory, she was, as well. The problem was that nobody in her organization had cleared time off her plate for her to learn the necessary tools or rebuild the applicable workstream. As a result, this AI project sat behind everything that was already due. Several months on, it remained on the sidelines.
I hear a version of that story constantly. In fact, it’s one of the patterns my colleague Brett Ramos and I set out to understand this year. Brett, senior director of commercial operations and analytics for Neurelis, and I conducted primary research among 100 commercial leaders across finance, IT, sales, marketing, and analytics at biopharma companies. We presented the results at the Pharmaceutical Management Sciences Association (PMSA) meeting in spring 2026.
The fallacy that eagerness equates to uptake was just one piece of conventional wisdom that didn’t stand up to scrutiny. What follows is a MythBusters-style rundown on the four adoption-related narratives that proved unfounded, along with a persona-based approach for meeting people where they are when planning AI rollouts in biopharma organizations.
4 adoption myths
One of the clearest lessons was that enthusiasm does not necessarily translate into regular use. Nearly 80% of survey respondents said they actively look for ways AI could improve their work, with only 8% worrying it would reduce the value of their own role. Yet, just 34% had actually built AI into how they work. That gap looks less like resistance and more like something structural getting in the way.
Department was not a reliable predictor, either. We found people in analytics who had barely experimented with AI alongside salespeople asking for direct access to the data warehouse so they could build their own agents.
Age was no more definitive. We found quiet experts across every age bracket, and people still waiting for a practical reason to use the technology. More training didn’t automatically close that gap, either; the bigger question was usually whether someone had found a use case that actually saved them time.
None of this means AI isn’t worth the effort. According to a widely cited MIT report, 95% of enterprise AI pilots fail to move the needle, which is easy to read as a verdict on the technology itself. Yet results of a separate report from Snowflake suggest that when AI does get adopted, 92% of early adopters report positive returns, and life sciences companies are earning back $1.43 for every dollar invested.
The organizational problem
It’s less that the technology doesn’t work and more that getting an organization to actually use it differently is the harder problem. Indeed, adoption tracks with individual mindset more than role, and this is a big part of why a single, company-wide AI rollout rarely works.
We found three major attitudes: The 34% who said AI is built into their work sat alongside another roughly one-third who self-described as tinkerers with the technology; the last third were either waiting for proof or using AI only when required.
Even that self-assessment may be generous, as people tend to rate their own adoption more favorably than their habits would suggest. Still, these three different starting points were often sitting inside the same team, sometimes the same function. We considered what it would take to design a segmented approach to AI adoption in an organizational setting.
To that end, we grouped responses into five personas, chosen from eight the survey identified. These five represented the ones most worth designing around given typical time and change-management constraints.
5 personas
The “governor” was focused on compliance and explainability. “Enthusiasts” valued experimentation, even if the technology had yet to earn their trust. For the “anxious skeptic,” job implications were of paramount importance. “ROI-ers” wanted a quantified pilot before supporting a broader investment, and the “swamped” understood the potential but had no time to learn another tool.
These are not fixed identities. Someone may display enthusiast tendencies while experimenting with a familiar task and governor-like qualities when the same tool touches patient or commercial data. That distinction matters when planning a rollout.
A compliance-focused employee should help set guardrails early, not receive a memo after the fact. Someone interested but inexperienced may need a tightly bounded pilot. And for the swamped, the most useful intervention may just be protected time and a workflow that removes a burden instead of adding one.
The practical lesson
Leadership often assumes that clearing the technical path is the hardest part of an AI rollout, but that isn’t usually where things get stuck. In the survey, the top barriers clustered closely together: compliance and regulatory uncertainty (52%), data quality or system limitations (47%), insufficient training (44%), and a lack of approved use cases (43%).
None of those separated cleanly from the rest, so there’s no single lever to pull. Those barriers are at least as much about trust, bandwidth, and change management as they are about technology.
Which brings me back to the woman I met at that conference. She did not sound resistant to AI. She sounded busy. No amount of enthusiasm could create the time required to learn a new tool and redesign a weekly process around it. And that may be the most practical lesson for companies planning their next rollout: Adoption requires capacity, not just encouragement.
