It starts with excitement.

Someone in the leadership team has just run ChatGPT on a few old GVD sections. It spits out something halfway usable. Heads nod. Ideas spark. The conversation shifts. “What if we could use AI to speed everything up?”

Fast-forward two weeks and a few brave analysts are testing prompts. Some decent summaries come back. Some gibberish. Sometimes both in the same sentence.

Everyone agrees it’s promising, but no one is quite sure where it fits.

This is where most consultancies are right now.

Standing on the edge of AI adoption, hoping it will be a silver bullet.

But here’s the truth: generic AI isn’t built for this work. And if you try to force it in, it won’t just slow you down… it might make things worse.

Why? Because market access is complex in all the wrong ways for off-the-shelf AI

This isn’t writing marketing copy or summarising news articles.

Market access work is a mix of clinical nuance, pricing context, regulatory frameworks, and payer logic. It lives in long-form documents, regional value statements, stakeholder maps, and years of strategic knowledge that isn’t public and isn’t always consistent.

It’s high-stakes, highly technical, and often ambiguous.

Which is exactly the kind of environment that trips up general-purpose models.

Problem 1: No ontology, no understanding

General AI tools don’t understand what matters in your world.

They don’t know the difference between a G-BA ruling and a NICE Final Appraisal Document. They don’t understand why a comparator in Spain matters but not in Sweden. They can’t prioritise which clinical endpoint is most persuasive for a French affiliate.

Without a structured ontology, without a framework that mirrors how market access consultants actually think, AI is just another tool guessing its way through jargon.

Problem 2: Hallucinations that look credible

One of the most dangerous things about generic AI is how confidently it gets things wrong.

It will make up references. Misstate trial endpoints. Attribute claims to the wrong source. And it will format all of that beautifully. So unless you check every detail, it will pass muster. Until it doesn’t.

This can go beyond just factual errors, all the way to reputational risk.

Problem 3: No traceability

In market access, you can’t just present an insight.

You need to back it up. Show where it came from. Trace it to a study, a payer quote, a regulatory precedent.

Generic AI doesn’t give you that.

You get a paragraph, maybe a source name if you’re lucky, but nothing that holds up in front of a client or affiliate.

So you do what every smart consultant does.

You redo the work to check it.

And now the AI isn’t saving time, it’s wasting it.

So what’s the alternative?

The answer isn’t ditching AI. It’s building it on the right foundation.

That means:

• A structured ontology trained on how real consultants think
• A controlled, high-quality evidence base grounded in trusted sources
• A model that ranks information like a consultant, not like a search engine
• Outputs that link directly back to source so you can trace every claim

That’s what makes AI not just usable, but trustworthy.

Not just fast, but safe.

Market access needs its own infrastructure

Trying to run market access work on general AI is like trying to launch a new medicine with a general marketing plan. It might work at a surface level, but you’ll miss what matters.

The firms that are getting this right aren’t using AI to replace consultants.

They’re using it to remove friction from the work that gets in the way.

The re-researching. The reformatting. The rechecking.

And they’re doing it with systems built for their world.

Not just bolted on from someone else’s.

AI will change this industry. That much is certain.

The only question is whether it will work for you. Or against you.

Let’s make sure it’s the former.