Access was the advantage. Now it is learning speed.
Budget used to decide who could do serious marketing work. AI made that access cheap. What separates teams now is how fast they learn what works.
In life science marketing, what decides it now is how fast you find out what works, and we have wasted enough hours to know the difference.
________________________________________________________________________________________________________________
Key takeaways
- AI made research, content and outreach cheap, so budget no longer decides who can do serious marketing work.
- The advantage now is how fast a team learns what works for its own business.
- Reading 244 forum conversations took two hours. Review-based personalisation earned a 24% cold email reply rate against a 3% to 8.5% benchmark.
- Watch for three traps: hype mode, random acts of AI and fear mode.
- Run ten-day sprints: one workflow, one metric, a human approval before anything goes out, then keep it or kill it.
Our co-founder Tom Claes was about fifteen when digital marketing arrived. He was building WordPress sites and running Google Ads for local bakeries and supermarkets, which felt like a serious operation at the time.
Plenty of companies waited it out. It looked like hype, or it looked messy, and nobody was sure where to start. The people who got curious anyway learned faster than everyone else, and the companies that hired them got ahead. The same pattern is running again, except the word in front of marketing has changed from digital to AI.
One caveat before anything else. A fool with a tool is still a fool. If your positioning is wrong, AI will help you be wrong faster and at higher volume.
Budget used to be the moat
Serious competitive research needed analysts. Content at scale needed writers. Outreach automation needed expensive software. Smart segmentation needed a data team.
Budget decided who did the work, and bigger companies outspent smaller ones. Access was the advantage. Most of our clients are challenger brands launching against multinationals with ten or twenty times their marketing budget, so that gap was the whole problem.
It is not the problem anymore. A two-person startup can now run research that used to need a five or six figure budget. The tools are cheap and some are free.
A controlled trial published in Science found ChatGPT cut professional writing time by 40%, and it helped the weakest writers most. A study of 5,172 support agents found novices gained 34% and top performers gained almost nothing. Cheap tools raise the floor.
If everyone has access, access is not the advantage. The advantage is how fast you learn what works for your business. That holds for any life sciences marketing team, whatever its budget.
244 conversations in two hours
Two examples, both from the same client.
We work with a medtech company that has built an alternative to the tenaculum, the instrument used to place an IUD. They are launching in the US.
The obvious way to write that campaign is to guess what women feel about IUD pain. The better way is to read what they have already said.
We identified the subreddits where the conversation actually happens. OBGYN, birth control, women's health, IUD, fertility. We added two keywords, pulled the most engaged recent discussions, and extracted the recurring themes.
244 conversations. Five themes. Two hours.
By hand that is three to five days of reading, and by the end you have seen everything and remembered a fraction of it. What came back instead was the five themes and the actual language people use, which is what you write ads with.
The themes were not noise. A trial published in Contraception found women rated IUD insertion pain at 64.8 on a hundred-point scale. Their clinicians rated the same procedure at 35.3. In 2024 the CDC told clinicians to counsel patients about pain beforehand and to offer lidocaine. The forum was saying what the clinical record already showed, and saying it in words you can put in an ad.
If you sell into labs instead of to patients, the same method works with thinner data. We have used it to find how scientists talk about a specific instrument, and the communities are smaller but the language is just as useful.
The tool matters less than people think. The time compression is the point.
What a 24% reply rate looks like
Same client, different problem. They wanted to reach gynaecology clinics in a single European city and had nothing to start from.
We scraped Google Maps for the clinics, which gives you profiles, contact details and patient reviews. Then we scraped each clinic website for the people working there and found their work emails.
The personalisation is the part worth copying. We did not open with the clinic name. We used the sentiment in their patient reviews instead. The first line ran something like this: “I came across your Google reviews and noticed your patients often highlight your kind and knowledgeable care.”
We sent postcards too. Hyper-personalised, and written and posted by a robot.
369 contacted. 324 delivered. 78 replied. 24 said they were interested.
That is a 24% reply rate. Published 2025 benchmarks put a good cold outreach reply rate between 3% and 8.5%, depending on the industry and how clean the list is. Hunter.io puts the average at 4.5%.
Researching and writing one personalised email by hand takes fifteen to thirty minutes. Building a list like that takes about three hours in total, including the mistakes we made learning it. By hand the same list is two to three weeks of opening websites and copying names into a spreadsheet.
The list is only half of it. Decide what a buying signal looks like for you, then work backwards to the data that proves it. When we map medtech companies in Boston, we prioritise the ones that recently raised, because a funding round means there is budget to spend.
Three traps, and we have fallen into all of them
None of this is automatic.
Hype mode. Our feeds are full of marketers demonstrating a new AI tool. It is genuinely interesting, and it is a trap. You get convinced the use case applies to you, you start setting it up, and three hours later you have built something you will never open again. Exciting, not operational.
Random acts of AI. You use it for emails, for social posts, for outreach, for research. No process sits behind any of it, so you cannot judge whether the quality improved or the impact was real. Plenty of activity, no progress.
Fear mode. The team avoids it because it feels risky, or confusing, or likely to hurt quality. That is a reasonable instinct. It is also a decision, and your competitor may be making the opposite one.
Run a ten-day sprint
So how do you use this without falling into the three traps?
Pick one workflow. Choose one metric. Let AI draft and have a human approve before anything goes out. Compare it to how you did it before. Then keep it or kill it.
There is now a compliance reason for that approval step. Since August 2026 the EU AI Act has required disclosure when AI generates deepfakes or published text on matters of public interest. The duty falls away where a human reviewed the content and took editorial responsibility. Most marketing copy sits outside that scope today. The scope will not shrink.
We run this as a standing session. Every couple of weeks someone presents a tool or workflow they tried in their own discipline. Designer, copywriter or marketer, they say plainly whether it helped.
Killing things is the point. Most of what you try will not beat the baseline, and finding that out in ten days is the whole return.
Where we landed on the data
There is a question underneath all of this, and it came from inside the team. When we showed the Reddit work at one of our internal sessions, someone asked what our position was on the data. Those posts are public. They are also extremely personal. People are describing medical procedures and pain, and we were putting it in a spreadsheet.
The honest answer at the time was that we did not know. Here is the one we hold now.
Publicly posted is not the same as consented to. Under GDPR, health disclosures are special-category data with the highest level of protection, and the exemption for data someone has made public is narrower than it sounds. The internet research ethics guidelines put it plainly. The more sensitive the data, the higher the duty of care, whatever the technical access.
So we use the themes and never the individual posts. We quote nobody, because a quote is searchable back to the person who wrote it. We keep nothing that identifies anyone.
Treat it like an intern, not a strategist
Every one of those was a judgment call. AI made none of them.
It does grunt work. It gives you a first draft. It pulls data. It does not tell you what matters or what to prioritise, and you should not ask it to.
Compressing time is not the same as cutting corners. The fundamentals of a life science marketing strategy have not moved, and neither has the go-to-market strategy underneath it. You still need real interviews, real market research and real judgement. AI removes the hours that were never the valuable part.
The skills compound, and this is where the two halves of the argument meet. On any single task, AI narrows the gap between a novice and an expert. Across firms and over years, it widens the gap between the ones running a learning loop and the ones who are not. OECD data shows the gap between the best and worst adopting countries widened from 14 points in 2021 to 24 points in 2024.
Raising the floor is not the same as levelling the field. Run the loop for a year and you will be measurably better at this than the teams who did not. That is exactly what happened to everyone who got curious about digital in 2001.
Frequently asked questions
How is AI changing life science marketing?
It made access cheap. Research, content and outreach that once needed analysts, writers and software now cost little or nothing. The advantage has moved to how fast a team learns what works for its own business.
What is the best way to start using AI in marketing?
Run a ten-day sprint on one workflow. Choose one metric, let AI draft, have a human approve everything before it goes out, and compare the result with how you work today. Then keep it or kill it.
Can you use public forum posts for marketing research?
Use them with care. Under GDPR, health disclosures are special-category data, even when someone posts them publicly. We use the recurring themes, never quote individual posts, and keep nothing that identifies anyone.
Does the EU AI Act require you to label AI-written marketing copy?
Since August 2026, the EU AI Act has required disclosure when AI generates deepfakes or published text on matters of public interest. The duty falls away when a human reviews the content and takes editorial responsibility. Most marketing copy sits outside that scope today.
Pick one workflow this week
Run the sprint on it. Ten days, one metric, an honest comparison against how you do it today. Good candidates: tracking how AI engines cite your brand (see the AEO guide), turning a webinar recording into follow-up content (see the webinar guide) or a small ChatGPT Ads test.
Or send us the workflow you think is worth automating, and we will tell you whether it is.
Let’s get in touch.
We build brands that make your science impossible to overlook, and worth more once it’s understood. Let’s talk about your rebrand.