The five stages of pharma competitive intelligence workflows

Obesity and diabetes licensing commitments reached $22 billion in just the first quarter of 2026 – already surpassing all of 2025’s $20.3 billion total. In the same period, Pfizer completed its $10 billion acquisition of Metsera after a bidding war with Novo Nordisk, and Roche invested $1.65 billion upfront to co-develop petrelintide with Zealand Pharma, adding an amylin analog to its CT-388 position. These aren’t isolated deals. They are decisions made at speed, in a space with more than 100 active development programs, where the difference between identifying an asset before a competitor does and arriving second can be measured in billions.^{1, 2, 3, 4, 5, 6}

The Competitive Intelligence (CI) teams supporting those decisions are, in most cases, working with more data than ever. The question is whether that data is reaching decision-makers in a form they can use. If you work in pharma CI, you’ve already explored some of what’s hiding in plain sight in the data you’re tracking. This post addresses what happens next: how intelligence moves – or fails to move – through five distinct workflow stages, and what the teams that consistently influence strategy do differently at each one.

Stage 1: Finding – traceability is the new coverage problem

Historically, CI’s main search challenge was coverage: not missing key signals such as a competitor’s trial registration, a related patent filing, or a regulatory submission that could alter an asset’s risk profile.

That concern hasn’t gone away. But the dominant failure mode has shifted. AI-assisted search tools now surface results quickly and in volume. What they do not automatically provide is a traceable path from each result back to a primary source — a registry entry, a patent filing, a peer-reviewed journal article, a regulatory submission — in a form that holds up in a portfolio review.

Consider what a credible CI package on an obesity asset requires today: integration across ClinicalTrials.gov, patent databases across multiple jurisdictions, peer-reviewed pharmacology literature, regulatory filings from FDA, EMA, PMDA and others as well as earnings and conference data. Each source type uses different data structures and reporting conventions. When CI teams shortcut that integration with general-purpose AI summaries, they can produce outputs that look complete but fall apart when a Business Development (BD) director asks where a specific claim comes from.⁷

The diagnostic question: Can you trace every claim in your current competitive landscape to a named primary source?

What high-performing CI teams do differently: They treat source traceability as a non-negotiable output standard, not a post-hoc documentation task. Before a landscape is delivered, the team can point to the specific registry entry, patent record, or journal publication behind each claim. This is not about being conservative – it’s about being credible when the questions come, and they always come.

Stage 2: Analyzing – pattern detection is not interpretation

Automated tools are effective at pattern detection: flagging that six companies have filed in a target indication, identifying clustering in a mechanism of action and tracking trial status across a competitive set. At the scale of 101,000+ pipeline drugs and 6.9 million+ patents, identifying those patterns is genuinely hard work, and automation helps.

Interpreting patterns is a different task. Whether clustering in each indication represents genuine target validation or a crowding risk that should change the investment thesis for a specific asset requires contextual judgment applied to a specific question.

The GLP-1 space illustrates the distinction directly. With more than 100 obesity compounds in active development, the raw count tells a CI manager very little. What matters is how those assets differ: which have distinct mechanisms, which target the same efficacy endpoints as semaglutide and tirzepatide, and which are intended for patients not adequately served by those therapies.

Roche’s CT-388, for instance, is differentiated by near-equimolar GLP-1/GIP receptor engagement (roughly 1:1 versus tirzepatide’s 9:1 ratio), a distinct peptide scaffold with independent IP, and Phase II data showing 22.5% placebo-adjusted weight loss at 48 weeks without a plateau — a profile that shaped both the company’s BD strategy and Pfizer’s decision to pursue a different mechanism through the Metsera acquisition. Mapping that differentiation requires depth of coverage across trials, patents, and clinical data — not a summary of headlines.^{6, 8, 9, 10}

The diagnostic question: When you identify a competitive pattern, does your team have a defined process for reaching an interpretation, or does that step happen informally?

What high-performing CI teams do differently: They assign each analysis a named decision question before the work begins. The question “what are the GLP-1 competitors?” produces a different output than “which GLP-1 assets represent a realistic differentiation threat to our Phase 2 program before 2027?” The second question constrains the analysis in a way that forces interpretation, not just enumeration.

Stage 3: Generating insights – the missing point of view

The gap between analysis and insight is a point of view. Analysis describes what’s happening in a competitive space. An insight names what it means for a specific decision and attaches a recommendation.

Most CI deliverables stop at description. Competitors are tracked accurately, trial progression is documented, the data is right — and then the “so-what” is left to the reader.

The CI teams that consistently shape portfolio and BD decisions go a step further: they explain the implication. Instead of saying, “here are five late-stage entrants in ALK+ NSCLC,” they say, “here is why the Phase III timing of two of them changes the risk profile of our licensing decision on Asset X, and what we should do if their data reads out before ours.”

Getting there requires setting the decision question at the outset. “Competitive landscape in oncology” is a topic. “Should we accelerate enrollment in our phase 2 before Competitor A’s data readout?” is a decision question. Those two starting points produce deliverables that are used very differently in a portfolio committee. McKinsey’s research on external innovation in pharma found that companies that identify high-potential assets early and make confident decisions quickly consistently outperform, with external innovation outperformers achieving 3.4 to 8.2 times greater returns on sourced assets. That kind of decision speed requires CI that arrives already connected to a recommendation.^{1112, 13}

The diagnostic question: Do your CI deliverables consistently include a stated recommendation, or do they present data and leave the interpretation to the reader?

What high-performing CI teams do differently: They align with their stakeholders on the decision question before building the deliverable — not after. In practice, this means a CI manager has a 20-minute conversation with the BD or portfolio lead before starting the landscape, asking: “What specific decision is this supporting, and when does it need to be made?” That conversation shapes everything that follows, from what data gets prioritized to how the output is framed.

Stage 4: Communicating – format determines whether CI gets used

A CI team that has executed the first three stages well can still produce output that does not influence a decision. The most common reason is format: a comprehensive briefing document that is organized around a topic area rather than structured to answer a specific question.

BD and portfolio leaders engage most directly with CI when it connects to a decision they are actively working through. A landscape that covers everything happening in an indication over the past 18 months is a useful background. It’s rarely what gets opened the morning of a licensing committee meeting. A two-page brief that answers “What is the probability Competitor X files before us in this indication, and which trial readouts between now and Q3 are the ones to watch?” is more likely to be in the room when the decision is made.^{12, 14}Every CI deliverable should connect to a named decision, a named stakeholder, and a timeframe. Without those three elements, a deliverable is a research update, not intelligence.

War games deserve specific attention here because they work differently from static deliverables. In a well-structured simulation, BD and R&D leadership do not receive a CI team’s conclusions — they test their own assumptions against simulated competitor responses. That active participation produces aligned assumptions that survive in the strategy meeting.

A consistent finding from pharma competitive simulations is that action steps rarely get implemented when senior leadership is not involved from the outset and is instead briefed only after the scenarios are built.

The design implication is straightforward: build senior involvement into the structure of the exercise before the simulation runs, not afterward.^{15, 16, 17}

The diagnostic question: For each CI deliverable your team produced in the last quarter, can you name the specific decision it was designed to support?

What high-performing CI teams do differently: They maintain a rolling “decision calendar” — a structured view of the BD, portfolio and R&D decisions coming up over the next 90 days, as well as the CI inputs each one requires. This replaces the reactive request pattern (“We need a landscape by Friday”) with a forward-looking production schedule anchored to actual decision timelines. It also makes it easier to say no to low-priority requests without creating friction.

Stage 5: Scenario modeling – from reporting to strategic advisory

Most CI functions stop at describing what is happening. Scenario modeling asks what might happen under specific conditions and what the organization should do in each case. It’s the output that distinguishes a CI function that reports on the competitive environment from one that shapes strategic decisions.

The Pfizer-Metsera acquisition is a useful example of the kind of decision scenario modeling is built for. Pfizer had already halted its oral GLP-1 program (danugron) in April 2025 after elevated liver enzymes in trial participants. The Metsera acquisition, completed in November 2025, was a pivot –buying a monthly injectable GLP-1 platform to re-enter the obesity market with a different mechanism and dosing approach. A CI team supporting that decision would have needed to model at least three scenarios: the competitive landscape if CT-388 (then still in Phase II) proved best-in-class; the landscape if Novo Nordisk consolidated its obesity position through its own acquisition activity; and the landscape if Metsera’s Phase II data didn’t replicate at Phase III scale. Pfizer eventually paid $10 billion after a bidding war with Novo Nordisk drove up the price from an initial $7.3 billion – which suggests the competitive scenario around Novo’s strategic interest in the asset was live and consequential.^{5, 8}

Scenarios are only as defensible as the data they’re built on. A portfolio committee that asks “How confident are you in this?” needs the CI team to point to primary sources: patent expiry timelines, trial design records, regulatory precedent. The quality of those sources determines whether a scenario is treated as a strategic input or a speculative one. In a deal environment where speed matters as much as accuracy — Q1 2026 obesity deal commitments were already at $22 billion — a scenario your team can’t defend quickly loses credibility before the conversation has started.^{1, 7, 18}

The diagnostic question: Does your CI function regularly produce forward-looking scenarios with named conditions and recommended responses, or does your output primarily describe the current competitive state?

What high-performing CI teams do differently: They separate scenario building from landscape building. A landscape is a snapshot; a scenario requires a trigger condition (“If Competitor X phase III data reads out positive by Q2”) and a recommended response. High-performing CI teams identify the three to five trigger conditions that would most change a specific strategic decision, assign probability assessments to each, and update those assessments as new data arrives. That structure makes CI outputs directly useful in planning conversations rather than as background reference material.

What AI changes – and what it doesn’t

AI is accelerating Stages One and Two measurably. Aggregating trial data, identifying patent clusters, summarizing a competitive set — tasks that previously took a CI analyst weeks can now take a fraction of that time. That’s real, and it changes how CI teams should think about capacity allocation.^{7, 18}

Stages Three through Five are not materially changed by AI. Insight generation, stakeholder communication, and scenario modeling still depend on analytical judgment, stakeholder knowledge, and the credibility of the underlying data. The CT-388 analysis described earlier — identifying the asset’s differentiation from tirzepatide on receptor engagement ratios and IP profile — is the kind of contextual interpretation that follows from structured data and analytical judgment, not from automated summarization.⁹

The specific risk AI introduces is at stage one. McKinsey has noted that pharma companies are exposed to AI hallucinations resulting from poor or incomplete data. An AI tool working from unverified or incomplete sources produces summaries that are fast and confident but can’t be traced to a primary source. For a CI team whose credibility with BD and portfolio leadership depends on the defensibility of its outputs, that’s a material risk.¹⁹

The answer isn’t to avoid AI — it’s to be deliberate about what the AI is fed. Cortellis draws from seven major global patent offices, 700+ peer-reviewed journals, regulatory agencies including FDA, EMA, PMDA, and TGA, and 100+ scientific and medical conferences. That coverage — continuously updated, structured, and traceable — is what makes AI-assisted CI defensible rather than directional. The tool amplifies what’s underneath it. What’s underneath it determines whether the output holds up.

The diagnostic question

The five stages aren’t a sequential pipeline that breaks in one place. There are five independent points where a CI function can lose the value it’s built.

For many teams, the primary failure point is stage four: the delivery format isn’t anchored to a named decision. For others, it’s stage three — analysis produces descriptions rather than recommendations. For teams investing in AI tooling, the current risk may be stage one — speed has increased, but source traceability has not kept pace.

Each failure point has a different fix. Stage Four is a process and format problem. Stage Three is a question-setting problem. Stage One is a data foundation problem.

Identifying which stage is costing your team the most influence right now is where a CI workflow improvement conversation should start.

Learn more about how Cortellis Competitive Intelligence helps pharma and biotech teams move from reactive tracking to proactive strategic foresight: Cortellis Competitive Intelligence & Analytics | Clarivate

References

Manalac, T. (2026, April 16). Obesity’s explosive growth continues as Q1 deals exceed total 2025 value. Retrieved from BioSpace: https://www.biospace.com/deals/obesitys-explosive-growth-continues-as-q1-deals-exceed-total-2025-value
Manalac, T. (2025, March 12). Roche Makes Another Obesity Play With Potential $5.3B Amylin Pact With Zealand. Retrieved from BioSpace: https://www.biospace.com/deals/roche-makes-another-obesity-play-with-potential-5-3b-amylin-pact-with-zealand
Sunny, K. C. (2025, November 13). Pfizer completes up to $10 billion acquisition of Metsera. Retrieved from Reuters: https://www.reuters.com/legal/transactional/metsera-shareholders-vote-10-billion-acquisition-by-pfizer-2025-11-13/
(2025, March 11). Roche enters into an exclusive collaboration & licensing agreement with Zealand Pharma to co-develop and co-commercialise petrelintide as a potential foundational therapy for people with overweight and obesity. Retrieved from Roche: https://www.roche.com/media/releases/med-cor-2025-03-12
War for Obesity Biotech Metsera. Retrieved from MedCity News: https://medcitynews.com/2025/11/pfizer-metsera-acquisition-bidding-war-novo-nordisk-glp1-obesity-pfe-mtsr-nvo/
Alvarez, D. (2026, January 8) Biopharma Trends 2026. Retrieved from BCG: https://www.bcg.com/publications/2026/reimagining-business-models-biopharma-trends
K-Dense Web. (2026, April 3). Pharma Competitive Intelligence in One Session: CT-388 and the GLP-1/GIP Obesity Landscape. Retrieved from K-Dense Web: https://www.k-dense.ai/blog/ct388-competitive-intelligence-obesity
GlobalData Healthcare. (2026, January 23). JPM26: Pfizer’s Metsera deal supercharges its obesity strategy. Retrieved from Yahoo! Finance:
K-Dense Web. (2026, April 3). Pharma Competitive Intelligence in One Session: CT-388 and the GLP-1/GIP Obesity Landscape. Retrieved from K-Dense Web: https://www.k-dense.ai/blog/ct388-competitive-intelligence-obesity
(2026, April 2). The GLP-1 Drug Pipeline: What Employers Should Expect Next. Retrieved from Excellus: https://excellusforbusiness.com/the-glp-1-drug-pipeline-what-employers-should-expect-next/
McKinsey & Company. (2025, June 20). Pulse check: Key trends shaping biopharma dealmaking in 2025. Retrieved from McKinsey & Co. : https://www.mckinsey.com/industries/life-sciences/our-insights/the-synthesis/pulse-check-key-trends-shaping-biopharma-dealmaking-in-2025
Tayi, A. (2025, December 18). vibe bio. Retrieved from Strategy & Due Diligence in Pharma Business Development: https://vibebio.com/blog/strategy-due-diligence-pharma-business-development/
Carlucci, S. (2026, April). Building your internal competitive intelligence network. Retrieved from LinkedIn: https://www.linkedin.com/posts/1carlucci_competitiveintelligence-pharma-biotech-activity-7437366914814738432-_g0_
Biopharma Vantage. (2026, May 9). AI in Pharma Due Diligence: Leveraging for Licensing and M&A. Retrieved from Biopharma Vantage: https://www.biopharmavantage.com/ai-pharma-due-diligence-licensing-ma
Life Science Leader: https://www.lifescienceleader.com/doc/using-war-games-to-predict-competitors-moves-0001
Bernard, S. (2017, February 20). War Games: Win Your Brand with Action. Retrieved from PharmExec: https://www.pharmexec.com/view/war-games-win-your-brand-action
How to Win Your Market with Competitive Wargames. (2025, October 1). Retrieved from Sedulo Group: https://sedulogroup.com/blog-post/what-is-wargaming-competitive-strategy/
McKinsey & Company. (2024, January 9). Generative AI in the pharmaceutical industry: Moving from hype to reality. Retrieved from McKinsey & Company: https://www.mckinsey.com/industries/life-sciences/our-insights/generative-ai-in-the-pharmaceutical-industry-moving-from-hype-to-reality