When Experience, Evidence, and AI Disagree in Pharma, What Should Leaders Do?

Three Answers. One Decision.

Consider a hypothetical launch review, six months before anticipated approval. The clinical profile is promising, the market is attractive, and expectations are rising.

Yet payer access is unresolved, competitor moves remain uncertain, and patient persistence is still an assumption. Leadership must commit resources anyway.

The commercial leader wants to move aggressively. Experience suggests that underinvesting at launch could leave the company struggling to catch up.

The analytics team sees a narrower early opportunity. Its scenarios suggest concentrating resources on patients most likely to gain access and start treatment.

Then an AI-assisted analysis arrives. Using available claims patterns, payer signals, patient journeys, and launch analogs, it points somewhere else: the biggest constraint may not be awareness; it may be access and persistence.

If that interpretation holds, the next dollar of promotion could create less value than the next dollar invested in payer evidence or patient services. Three credible answers and one decision.

The scene is hypothetical, but it illustrates a dilemma that also arises in portfolio reviews, market-access investments, and resource allocation. Pharma executives have always made consequential decisions before they knew all the facts.

Now AI adds another source of interpretation, often quickly and with persuasive confidence. The executive challenge is to understand what each source contributes, where each can fail, and which assumption should determine the next commitment.

System 1: Experience Sees the Pattern

Daniel Kahneman’s distinction between fast, intuitive thinking and slower, deliberate reasoning offers a useful starting point. System 1 is fast and driven by pattern recognition.

In pharma, an experienced market-access leader may sense that a value proposition will meet payer resistance, or a commercial leader may recognize an uptake pattern that a forecast overlooks.¹

This judgment has value. Years of experience can reveal context and patterns that a spreadsheet misses.

But experience can also become an anchor. The last successful launch may not be the right analog.

Competitive intensity may have changed and utilization management may be tighter. A familiar market can create false confidence.

The discipline is to expose intuition’s assumptions. Before presenting the forecast or AI recommendation, capture the team’s initial view.

Ask:

• What do we believe and why?
• What has to be true for that belief to hold?
• What evidence would change our mind?

System 1 should start the conversation. It should not end it.

System 2: Evidence Tests the Story

System 2 is slower and deliberate: the work of examining forecasts, segmentation, payer research, patient-flow models, scenarios, and financial analysis. These are modes of reasoning rather than separate departments.

Analysts use intuition, and commercial leaders reason carefully. Its purpose is not simply to add rigor. It is to make the strategy confront evidence.

That distinction matters, as a sophisticated forecast can still be built on an optimistic assumption.

Research can answer the questions the organization chose to ask while missing the question that matters. A polished strategy deck can create confidence without materially improving judgment.

For executives, the test is straightforward: did the analysis challenge the strategy, or merely validate it? If the launch plan assumes rapid adoption, step outside the model.

Examine comparable launches and ask:

• How quickly did physicians actually adopt?
• What access restrictions emerged?
• Where did patients abandon therapy?
• How did persistence evolve?
• Which assumptions separated the successful launches from the disappointments?

Good analysis makes the strategy’s vulnerabilities visible before the company commits capital.

System 3: AI Widens the Field of View

Steven D. Shaw and Gideon Nave propose extending dual-process thinking through Tri-System Theory, in which artificial cognition operates alongside intuitive and deliberative human reasoning. Their proposed System 3 offers a useful way to examine AI’s influence on human judgment.²

With appropriate tools and access to reliable data, AI can help examine claims, prescribing behavior, payer policies, patient journeys, competitive signals, and historical analogs. It can suggest overlooked explanations and alternative scenarios.

Its usefulness depends on the information it can actually access and the checks applied to its output. In the hypothetical launch meeting, the AI analysis raises a simple but disruptive question: What if awareness is not the bottleneck?

That question could change the investment decision, as it merits investigation rather than immediate acceptance. Shaw and Nave describe “cognitive surrender”: accepting AI-generated answers without sufficient scrutiny.

In their controlled reasoning experiments, accurate AI assistance improved performance, although erroneous assistance could reduce it below a no-AI baseline.

These were reasoning tasks, not pharmaceutical launch decisions. The findings support scrutiny of AI advice; they do not establish improved launch outcomes.²

Agreement also needs scrutiny. If the AI analysis uses the forecast’s assumptions and analogs, it may reproduce the same weaknesses.

Two matching answers are not necessarily independent confirmation. Ask the analytics lead to identify the data and assumptions the outputs share.

Recent research reported in Harvard Business Review found that leading AI models often favored similar, fashionable strategic recommendations even when the business context changed.³ This raises a related risk: cognitive convergence.

For pharmaceutical executives, agreement across AI tools should not substitute for testing whether a recommendation fits the company’s particular market, access constraints, and competitive position.

Ask AI to challenge the strategy, search for contradictory evidence, and reveal assumptions the team may be overlooking. Then verify the sources, examine the reasoning, and decide how much weight the recommendation deserves.

AI should have a seat at the table. It should not have the final vote.

The Strategic Value of Disagreement

Now the three views collide. Experience favors a broad launch, while the analysis supports a focused launch and AI suggests that incremental investment belongs in access and patient support.

The pressure to align the team can make a quick compromise attractive. Choose the forecast, settle the budget, and move on.

But a quick resolution can leave the decisive assumption untested. Instead, ask: Why do the three systems disagree?

Perhaps leadership is anchored to an earlier launch. Perhaps the forecast underweights utilization management.

Perhaps the AI relies on analogs that no longer fit the market. Each possibility points to a different assumption worth testing. The recommendations may also pursue different outcomes.

Experience may emphasize physician reach, the forecast may emphasize treatment initiation, and AI may emphasize continued use.

Leaders need a common outcome and time horizon before comparing them. If functions favor different investments because their performance measures reward different things, leadership must address those incentives too.

The framework helps organize the review. Experience supplies an initial judgment. Deliberate analysis tests its assumptions.

AI helps explore what the team may have missed. Human judgment determines what the evidence warrants, what uncertainty remains, and what consequences the organization is prepared to accept.

Investigate the disagreement most likely to change a consequential decision. Further debate adds little when the evidence already supports action.

Make the Decision and Design the Next One

The leadership team still has to act. Suppose it chooses a focused launch, commits the capacity that cannot wait, and preserves some funding to expand if access and treatment initiation exceed expectations.

Persistence can inform later decisions once sufficient follow-up is available. Preserving flexibility has value, but delaying essential capacity also has costs.

The team should make that trade-off explicit. Approval of a launch plan can end the strategic debate just as the organization starts learning from execution.

Leadership locks the budget, the field force deploys, and attention shifts to delivery. Under uncertainty, the assumptions behind the plan still need review as evidence arrives.

Before launch, the budget owner should specify the signals that would justify expansion or reallocation, who will monitor them, and when the team will review them.

If awareness is high but prescribing is below plan, examine physician confidence, clinical suitability, and workflow barriers before increasing promotion. If prescriptions rarely become treatment starts, investigate reimbursement and fulfillment barriers.

If initiation is strong but continued use is weak, establish why patients stop before assuming more patient support will help. Access problems may justify additional evidence generation or payer engagement, depending on their cause.

Allocate the next investment against the constraint the evidence identifies. Keep appropriate treatment and patient outcomes in view alongside commercial performance.

The operating rhythm becomes decide, execute, measure, adapt. Assess each signal over an appropriate observation period: coverage and prescription conversion may provide early information, although persistence takes longer.

An immature data window can make incomplete follow-up look like poor performance.

Five Questions for the Next Executive Review

This framework does not require another committee. It can begin with five questions in the next launch, portfolio, or investment review:

1. What does our experience suggest, and which assumptions support that judgment?
2. What does the evidence support, including findings that challenge our preferred answer?
3. What does AI suggest we may be missing, and can we verify its sources and reasoning?
4. Where do the recommendations diverge, and which difference could change our investment?
5. What evidence would cause us to change course, who will monitor it, and when will we review it?

The last question is crucial. Once leaders commit capital and reputations become attached to a strategy, changing course becomes harder.

Record why the team chose one path and rejected the alternatives. That record helps distinguish a flawed assumption from poor execution when results arrive.

Define the triggers before the evidence arrives.

Judgment is the Leadership Advantage

The meeting ends with a decision. The leadership team understands the assumptions behind it, how experience, evidence, and AI diverged, and what it will monitor as execution begins.

That is the real opportunity. AI can increase analytical capacity and speed. Better strategy still depends on how its output is used.

Experience can become overconfidence. Analysis can reinforce a preferred answer. AI can encourage cognitive surrender when its recommendations go unchecked.

The advantage comes from bringing all three into a disciplined decision process. Executives must recognize when experience rests on a relevant pattern, when analysis genuinely tests the strategy, and when an AI-generated conclusion extends beyond the evidence.

Leadership retains responsibility for choosing among the options and owning the consequences. The value of AI grows when it helps the organization identify assumptions it would otherwise overlook.

Before committing the next dollar, the team should be able to explain the assumption it is funding and the evidence that would justify funding something else.

About the Author

Thani Jambulingam, PhD, is a professor of pharmaceutical and health care business at Saint Joseph’s University. His work focuses on pharmaceutical strategy, commercialization, market access, launch excellence, and AI-enabled decision-making in health care.

References

1. Kahneman D. Thinking, Fast and Slow. Farrar, Straus and Giroux; 2011.
2. Shaw SD, Nave G. Thinking—fast, slow, and artificial: how AI is reshaping human reasoning and the rise of cognitive surrender. Preprint. SSRN. 2026. doi:10.2139/ssrn.6097646
3. Romasanta A, et al. Researchers asked LLMs for strategic advice. They got “trendslop” in return. Harvard Business Review. March 16, 2026. Accessed October 14, 2026. https://hbr.org/2026/03/researchers-asked-llms-for-strategic-advice-they-got-trendslop-in-return