The information edge is mostly gone. SRC looks at how AI shifts competitive advantage differently across catalogue products, specialised kits, and capital equipment — and what it means for commercial hiring in life sciences.
What AI Is Actually Doing to Biotech Sales Jobs
Last year, O2 — the UK mobile carrier — built an AI “granny” called Daisy. Her entire job was to answer scam calls and keep fraudsters on the line for as long as possible so they could not reach real victims. The longest call ran around 40 minutes. Daisy could hold a conversation. She could not hold a relationship.
If AI can do that, the question of what it does across the rest of a commercial job becomes hard to ignore. At SRC, we work with biotech commercial teams every day, and some version of that question comes up most days: what is AI actually doing to the work of selling in life sciences?
Door-to-door gave way to the phone. The phone gave way to the internet. This looks like the next turn — and it does not affect every role the same way.
What the Data Shows
Anthropic’s March 2026 report on AI’s labour market impact breaks usage down by job role. One of the roles it covers is essentially the life-science tools rep: “Sales Representatives, Wholesale and Manufacturing, Technical and Scientific Products.” Observed AI usage in that role: 0.73%.
The tempting reading is that the capability is sitting right there and almost nobody is using it yet. The realistic reading is that adoption is simply early — or that real-world capability is not yet as ready as the theoretical charts suggest. Either way, the gap will close, and it will not close uniformly across every product segment.
Below is our view on where that shift is and is not happening — and what it means for the people doing the job and the companies hiring them.
Where AI Shows Up First: Marketing and Business Development
Marketing has always been expensive, because the hard parts were genuinely hard. Mapping a market took time. Finding the right labs took people. Writing a message that spoke to a specific researcher’s recent work, then getting it in front of them, took both. That difficulty was the whole justification — it was what separated a good campaign from a wasted one.
Most of that is now something a single person with the right tools can do in an afternoon.
Anyone can generate a personalised outreach message in seconds — one that references a researcher’s last five papers and connects your product to exactly what they are working on. Market mapping that used to keep a team busy for weeks can now be done in a day, though the output still needs significant cleaning before it is usable.
The problem is that every competitor has the same tools. When everyone can craft the perfect personalised message instantly, the perfect personalised message stops working. A researcher who receives five of them in a day learns to delete all five. The cost of generating noise has dropped to zero, which means the noise never stops.
Reaching someone is easy now. Being worth reading is not.
What is clear is that standing out now requires either the creativity that comes from years of knowing what works, or the technical ability to automate what you used to do well — and increasingly both. Some companies are already hiring one experienced marketing manager to run the whole function, rather than paying for a team of specialists to cover each part separately.
How AI Affects Different Product Segments
AI does not hit life science commercial roles uniformly. The impact depends heavily on what you are selling. Here is how it plays out across three key segments.
Catalogue Products: The Segment AI Changes First
Restriction enzymes, common antibodies, generic kits — the consumables a lab can order without speaking to anyone — are getting more automated by the month.
Platforms like ZAGENO now pull vendor catalogues into one place and rank them on price, lead time, and a lab’s own purchasing history. AI flags anything that looks like odd spend. Large pharma companies have been doing this in-house for years: Novartis has run an AI-powered “Buying Engine” since 2021, centralising indirect purchasing across the whole company using the same recommender techniques behind streaming services.
Keith Robison — who spent ten years at Millennium Pharmaceuticals and now works at Ginkgo Bioworks — made the stakes plain in his April 2026 post “If My Agent Can’t See Your Catalog, Will It Exist Much Longer?”:
“If their agent cannot see your catalogue, then for the purposes of the sale, you are simply not there.”
For smaller suppliers the risk is immediate. Innovative Bioscience, a US fetal-bovine-serum company, has written about adjusting to this reality — using AI to track pricing against national distributors, running a knowledge base to steer customers to the right product, and flagging supply problems early. The smaller you are, the more clearly you can see the risk: keep up with your customers’ tooling, or quietly disappear from their search results.
The agents are not perfect yet. While ordering compounds for his lab, Robison had one tell him a product was in stock when it had actually been discontinued. Another compound was matched to the wrong CAS number — the kind of error that does not announce itself, because a CAS number looks equally authoritative whether it is right or wrong. For now, the human stays in the loop not to do the searching, but to catch the confident mistake.
What this means for the job: Catalogue selling is the segment AI changes first and hardest. The role that answers basic questions and closes the order has little left when an agent handles both sides of that interaction. As catalogues get cleaner and agents get better, that gap closes in one direction only.
Specialised Kits: The Knowledge Edge Erodes Slowly
Specialised kits are where things get more interesting. A scRNA-seq library prep, an ATAC-seq kit for frozen tissue, and a CITE-seq panel are not interchangeable. Two products that look similar on a spec sheet can give very different results from the same sample. The only real way to choose between them used to be to read what other labs actually got — and that took a PhD student the better part of a week.
Now you can ask which comparisons exist and get a reasonable summary of what was used, how the results differed, and where the protocols diverged, in minutes.
For years, that comparative knowledge lived with your own people — the reps and field application scientists who had seen things work and fail in dozens of settings. When an agent can put together a reasonable version of that picture from published papers in minutes, the rep’s advantage gets smaller.
It does not go away entirely. The rep who knows where that published picture is wrong — in a specific and costly way — is still valuable. The rep whose main value was holding information the buyer did not have time to find is in a harder position.
There is also a structural shift for vendors. The work that shapes the published record — scientific affairs, KOL relationships, publication support — was never unimportant, but it was easy to underfund when the link between that work and winning deals was not visible. It is more visible now. Buyers arrive having already read the data. A kit with strong published evidence wins before anyone picks up the phone.
What this means for the job: Nobody loses a deal overnight. But the competitive advantage is slowly moving out of the sales conversation and into the available data on the internet — where the rep has no say.
Capital Equipment: Preparation Changes; the Deal Does Not
Buying a confocal, a sequencer, or a high-content imaging platform involves too many people for any agent to stand in for. Facilities worries about power and footprint. IT worries about data storage. Finance looks at depreciation and service contracts. A long line of end users all want different things. The committee is too complex, and the politics too human, for an AI to navigate.
What AI does change is how prepared everyone turns up.
Buyers can now pull apart a request for proposal, model total cost of ownership across service, consumables, and downtime — rather than just the sticker price — and produce comparison summaries that have actually read the spec sheets. Importantly, forums like Reddit get cited too. Scientists share unflattering things about support teams and application specialists in places that never appear in published papers.
Robison is clear on how he uses AI here: preparation, not decision. Faced with two cell sorter models he did not know well, he had an agent run a head-to-head on the specs — not to make the call, but so he would know which questions to ask and would not walk in needing the basics explained.
The less obvious implication for vendors: more transparency, not less. An agent doing the buyer’s research will quietly cross off anyone who looks like they are hiding something.
For reps, the shift is uncomfortable if your edge was knowing more than the buyer. It is fine if you were genuinely useful regardless — reading the committee, knowing when to stop pushing on price, being able to talk to a PI who wants the best data and a core director who just needs something their team can use.
What this means for the job: Big-ticket sales change too, but more slowly. No agent yet understands the politics of a buying committee or knows when to back off.
The New Part of the Job Nobody Has Named Yet
Something worth paying close attention to: your customer’s AI is now part of your market.
When a researcher asks Claude or ChatGPT which product fits their experiment, that shapes the deal before you are even in the picture — and the answer is not always accurate. If a model has decided your product is not strong in an area where it actually is, that opinion is sitting on every laptop in your territory. You can only fix it if you know it is there.
So there is a new part of the commercial job now: knowing what AI models say about you and your competitors, and correcting it where it is wrong. This is not marketing in the traditional sense. It is a new form of competitive intelligence.
The Skills That Still Matter
The three segments do not move together, and that is the whole point.
Catalogue selling is getting hollowed out — once an agent handles both sides of a repeat order, there was never much of a relationship there anyway. Specialised kits are harder to call: nobody loses a deal overnight, but the advantage is slowly moving out of the sales conversation. Capital equipment changes too, but more slowly, because no agent yet understands the politics of a buying committee.
A rep used to have an information edge over the buyer. That edge is mostly gone. What is left is the part of the job that was never about information: knowing when to push and when to back off, and being able to speak to both a PI who wants the best possible data and a core director who just needs something their team can actually use.
Daisy could hold a conversation. She could not hold a relationship.
What This Means for Commercial Hiring in Biotech
At SRC, we work with biotech commercial teams navigating exactly these shifts. The candidates who thrive are not necessarily the ones who know the most — they are the ones who know where human judgment still changes the outcome.
The roles that remain durable are the ones where relationships, reading a room, and scientific credibility matter. The roles most exposed are those where the primary value was access to information the buyer did not have time to find themselves.
If you are building or rebuilding a commercial team, we can help you think through what that looks like for your specific portfolio.
Many thanks to Keith Robison (Ginkgo Bioworks) for his time and for a conversation that genuinely shaped this piece.
