Most pharma companies are using AI. The ones getting results have made a different decision about where it lives.
That is not a critique of tools. The tools are good. The question is whether AI sits on top of your commercial process or inside it. Those are not the same thing, and the difference is not philosophical. It is structural. A tool is something you pick up for a task and put down when the task is done. An operating layer is something the process runs on. It is embedded before the work begins, not applied to the output after.
What "Inside the Process" Actually Looked Like
At Novartis, I owned a 20+ step international HCP strategy and content framework spanning Oncology, Neuroscience, Cardio-Renal-Metabolic, and Immunology. Across the steps I owned, I embedded AI into more than half of them. Not at the end to polish the output. At the judgment points where consequential decisions were being made.
Here is what that looked like in practice.
When a customer story was built for a brand, the foundational messaging architecture that everything downstream depended on, 10 or more cross-functional participants from market access, medical, field, and regional leads brought different views to the room. AI was used to tighten the draft language, to test messaging against simulated customer responses, and to resolve alignment conflicts in a way that felt less like someone losing. The AI was not polishing a finished document. It was inside the process at the moment the decisions were being made.
When a messaging platform was built with five to seven people, including TA, Medical, and key market representatives, AI-based customer twins were used to generate iterative feedback before committing to formal research cycles. The traditional model was: draft, then wait for research, then revise. The AI-embedded model was: draft, test virtually, refine, then validate formally. The sequence changed. The AI was inside the sequence, not appended to the end of it.
When the implementation guide for regions and markets was built, compliance checking was previously dependent on me personally reviewing every document. That does not scale across 10 or more brands. AI was embedded to do compliance checking on process, template, and content flow. The AI sometimes caught what manual review missed. The leadership bottleneck disappeared.
Across all of these, the pattern was consistent: AI at the judgment point, before the decision was made, not after the output was produced.
What Stayed Human
The steps that stayed entirely human were not the ones I forgot to automate. They were the ones that required human judgment for specific reasons.
Every formal PMR testing step remained human. The legal and regulatory significance of formal market research requires human design, human oversight, and human accountability for the findings.
MLR finalization step remained human. The Medical-Legal-Regulatory review process is not a document check. It is a compliance function with legal liability attached. AI does not sign off on promotional materials. A physician-attorney-regulatory chain does.
Field force training remained human. Not because AI cannot support training development. It can. But ownership of that function sat elsewhere in the organization.
This is not a hedge. This is what regulatory fluency looks like in a commercial marketing operation. Knowing where AI belongs inside the process requires knowing exactly where it does not belong. Both answers matter.
Why This Is a Leadership Decision
The commercial leader decides whether AI is inside the process or on top of it. The AI platform does not make this decision. The vendor does not make this decision. The brand team does not make this decision by accumulating individual tools.
The decision is architectural. It requires someone who can look at a commercial process end to end, identify the judgment points where AI can change the quality or speed of the decision, and redesign the workflow before the work begins.
Most commercial organizations have done the opposite. They have acquired tools, trained teams to use them for specific tasks, and measured AI adoption by counting how many people opened the tool this week. That produces AI on top of the process. The outputs get polished faster. The process itself does not change.
The difference shows up in outcomes. When AI is embedded at the judgment points, the quality of the decisions improves because the feedback loop is inside the decision, not downstream from it. Alignment conflicts get resolved through a structured process, not by waiting for someone senior to break the tie. Compliance checking happens automatically as the document is built, not as a final gate before submission.
This is not a faster version of the old process. It is a different process.
The Question to Ask Before Your Next AI Investment
The diagnostic is simple. Where in your commercial process is AI actually embedded, and where is it applied on top after the work is done?
Most teams, if they answer honestly, will find that AI is applied on top in almost every case. The tools are real. The outputs are faster. But the underlying process is the same one that existed before the tools arrived. The workflow was not redesigned. The judgment points were not identified. The AI was handed to the team as a capability and the team used it the way they use every other capability: task by task, after the work is already in motion.
That is a reasonable place to start. It is not where AI creates a structural advantage.
The structural advantage comes when the process itself is built around AI at the points where it changes the quality of the decision. That requires a leader who understands both the commercial process and the AI capability well enough to see where they fit together. Not a tool selection decision. Not a training program decision. A process architecture decision.
If your commercial team is still treating AI as a tool you pick up for specific tasks and put down when the task is done, you have not yet built the operating layer. That is the conversation I am having with commercial leaders right now. Here is how I work with pharma and MedTech teams.