Part 3: Agentic AI and the New Economics of Transaction Risk
Investment Committee Takeaways
- Agentic AI is changing diligence from a fixed workstream into a continuously operating risk engine.
- The core question is no longer whether a portfolio looks strong; it is whether the portfolio supports the valuation under realistic market, reimbursement, regulatory, and competitive pressure.
- Patent analytics can identify signals, but investment judgment still turns on whether those signals affect exclusivity, pricing power, FTO, exit value, and deal structure.
- The advantage belongs to transaction teams that can keep pressure-testing assumptions until capital deployment becomes irreversible.
Introduction
In Part 1 of this series, we discussed how AI commoditized the first layer of patent diligence by standardizing issue spotting across prior art, claim scope, and portfolio analytics. In Part 2, we examined how AI reshaped patent prosecution and why long-term portfolio value increasingly turns on specification architecture, continuation flexibility, and preserving optionality as technologies evolve.
Part 3 is about the deal model. Agentic AI is not merely making diligence faster. It is changing how sophisticated investors identify, sequence, and act on risk before capital is deployed.
For years, patent diligence was bounded by deal timing, budget, and human synthesis capacity. At some point, diligence stopped because the incremental question was too expensive, too slow, or too hard to integrate into the transaction process. That constraint is weakening.
From Portfolio Review to Investment Risk Engine
Sophisticated investors are no longer using AI merely for search, summarization, or patent scoring. They are beginning to use continuously operating agentic systems to generate hypotheses, monitor changing datasets, rerun valuation assumptions, surface emerging asymmetries, and identify the next layer of transaction risk in near real time.
That changes the diligence question. The traditional question was: “Is this a strong patent portfolio?” The better investment question is: “Does this portfolio continue to support the valuation as the market, reimbursement environment, regulatory framework, and competitive landscape evolve?”
That distinction matters in healthcare AI, diagnostics, biologics, therapeutics, and platform-technology transactions where intellectual property is often treated as a core enterprise asset.
The Analytics-Exclusivity Gap
A portfolio can score well and still underperform as a moat. That is the analytics-exclusivity gap.
Patent dashboards may show citation strength, family breadth, geographic coverage, and active prosecution. Those are useful signals. They are not the same as pricing power, FTO comfort, design-around resistance, reimbursement durability, or exit leverage.
A diagnostics company may begin as a biomarker-identification business and later become clinical workflow infrastructure. A biologics platform may move from a single therapeutic hypothesis into manufacturing optimization, combination therapy, or patient stratification. A healthcare AI company may pivot toward reimbursement analytics or broader infrastructure integration. The patent architecture often remains anchored to where the company started rather than where the business is going.
That disconnect is easy to miss in static portfolio analytics. It becomes visible when diligence recursively pressure-tests the assumptions behind the investment thesis.
When the Portfolio Scores High but the Moat Runs Shallow
Consider a diagnostics company with twenty-plus granted patents and pending applications covering a proprietary biomarker panel used in oncology treatment selection. AI-driven portfolio analytics score the portfolio favorably: strong citation metrics, broad geographic coverage, active continuation families, and claims that map to the current commercial product.
On a management presentation, this looks like a well-protected franchise. In traditional diligence, the analysis might stop there.
Recursive pressure-testing can reveal a different picture. The granted claims may be tightly drafted around a fixed combination of biomarkers identified during early discovery work. The assay methodology, sample preparation steps, algorithmic interpretation, and clinical decision rules may be only partially captured in the specifications. The portfolio may protect the original panel, but not the evolving platform.
When the diligence team models design-around paths, the picture can change quickly. Competing reference laboratories may have published validation data using overlapping biomarker panels and different scoring methods. A competitor continuation may claim a partially overlapping panel using a different algorithm. The target may need to construe its claims narrowly to distinguish prior art, creating the very design-around corridor competitors are already exploiting.
The data may all be available. The issue is synthesis. Analytics can surface references, claims, and filing activity. Investment judgment is needed to connect claim scope, prosecution history, competitor strategy, clinical convergence, and the company’s actual revenue model.
In that setting, the gap between portfolio score and actual exclusivity can represent hundreds of millions of dollars in enterprise value.
When Reimbursement Reality Changes the Meaning of the Portfolio
A related pattern arises when patent analytics and reimbursement reality diverge.
Assume a precision-medicine company has a strong-looking portfolio covering a companion diagnostic tied to a specific therapeutic. The patents cover the assay methodology, biomarker target, and clinical algorithm used to guide treatment selection. Claim mapping confirms coverage across the commercial product. The portfolio appears to support premium pricing and market exclusivity.
But the revenue model depends on favorable Medicare reimbursement under an existing Local Coverage Determination. When an agentic diligence system cross-references the patent portfolio against CMS coverage trends, MolDx developments, and recent LCD revisions in adjacent diagnostic categories, the investment picture may change. If coverage criteria are tightening and prospective clinical-utility data may be required, the evidence package that once supported reimbursement may not support the next review cycle.
The patents may remain valid and enforceable. But the commercial value of the patents depends on a reimbursement environment that is shifting beneath the deal. If reimbursement contracts, the revenue base supporting the valuation contracts with it. The patent moat may still exist, but the market it protects may be less attractive.
Traditional patent diligence may not surface that issue because it is not a patent-law problem. It is an underwriting problem that changes the meaning of the patent portfolio.
What Recursive Derisking Actually Looks Like
Recursive derisking means that diligence does not stop at the first answer. Each answer generates the next investment-relevant question.
An agentic system may identify continuation weakness. That finding may trigger analysis of competitor filing behavior, reimbursement evolution, product-roadmap drift, and future workflow adoption. That analysis may expose design-around risk, specification limitations, or FTO concerns several years into the projected hold period.
A practical framework is:
Recursive Derisking Capacity = Agentic AI × Human Judgment × Remaining Transaction Time
- Agentic AI expands the range and depth of inquiry.
- Human judgment determines which questions matter to the investment thesis.
- Remaining transaction time determines how many layers of uncertainty can be explored before capital deployment becomes irreversible.
The function is not linear. The later a material asymmetry is surfaced, the more consequential it may be for valuation, structure, indemnity, earnout mechanics, or willingness to close.
When Specification Depth Matters More Than Portfolio Size
Patent count is easy to measure. Specification depth is harder to underwrite. Yet in life sciences transactions, specification depth often determines whether the portfolio can continue supporting the business as the technology evolves.
A biologics-platform company may hold an impressive portfolio of composition-of-matter and method-of-treatment patents. But if the original specifications were drafted around a narrow therapeutic hypothesis, they may not support later continuation claims directed to combination therapies, new indications, or adjacent clinical applications. The continuation chart may look robust, while the specification runway is actually running out.
Agentic systems can flag the pattern by comparing granted claim scope against disclosed-but-unclaimed embodiments, public pipeline disclosures, and competitor activity in adjacent therapeutic areas. But determining whether the written-description requirement under 35 U.S.C. § 112(a) supports future claims remains a legal and scientific judgment call.
For investors, the question is not how many patents exist. The question is whether the portfolio can continue producing useful claims during the hold period and at exit.
The New Transaction Architecture
Historically, diligence largely preceded closing. The transaction timeline assumed there was a practical point where further inquiry produced diminishing returns.
Agentic AI changes that assumption. Continuously operating systems can monitor legal, scientific, reimbursement, competitive, and regulatory developments while a transaction is moving. That may push investors toward continuous adaptive diligence rather than static diligence.
In practical terms, this may affect staged diligence checkpoints, rolling valuation reassessments, exclusivity analyses, signing and closing mechanics, earnout terms, indemnity allocation, and pre-closing monitoring protocols.
The economics of “one more question” are changing. As the marginal cost of sophisticated inquiry declines, transaction teams can keep testing the investment thesis longer. The challenge becomes deciding which signals deserve action before the window closes.
Late-Stage Discovery and Deal Leverage
Late-stage findings involving reimbursement displacement, continuation weakness, design-around feasibility, competitive convergence, specification limits, or platform drift can materially alter valuation, earnout structure, indemnity allocation, licensing economics, or willingness to close.
For example, a rolling agentic analysis may flag newly published competitor continuation applications filed only weeks before signing. If those applications could mature into claims affecting the target’s next-generation product, the finding may alter FTO comfort, negotiation leverage, indemnity terms, or earnout structure.
The sophisticated investor of 2026 may no longer ask only: “Have we completed diligence?” Increasingly, the better question is: “How long should we keep pressure-testing this transaction before capital deployment becomes irreversible?”
The New Competitive Advantage
The competitive advantage is no longer simply access to AI-driven patent analytics. Sophisticated investors increasingly have access to similar tools. The advantage lies in how intelligently the transaction team structures recursive inquiry, how effectively agentic systems surface strategic asymmetries, and how quickly experienced professionals determine what matters commercially.
That is where transactions are won, repriced, restructured, or abandoned.
Recent public transactions reflect the stakes. XtalPi and DoveTree announced an AI-driven drug-discovery collaboration reportedly valued at up to approximately $6 billion. Novartis expanded its collaboration with Monte Rosa Therapeutics in a transaction reportedly valued at up to approximately $5.7 billion. Those transactions were not driven simply by patent count or citation metrics. They were driven by whether the underlying patent architecture supported future therapeutic pathways, commercialization optionality, platform scalability, reimbursement leverage, and long-term strategic exclusivity.
Why Agentic AI Increases the Value of Human Judgment
Agentic AI may increase the value of sophisticated human judgment rather than reduce it.
As analytics become commoditized, advantage shifts toward interpretation, prioritization, and transaction judgment. Recursive derisking produces more questions, more data, and more potential risk signals than prior diligence models. The challenge is deciding which signals change the investment thesis and what to do with them in the time remaining.
That requires practitioners who understand patent law, science, reimbursement, competitive dynamics, commercial strategy, and deal structure. Agentic AI makes those practitioners more powerful. It does not make them less necessary.
In 2026, access to AI-driven analytics is no longer the edge. The edge is knowing how to use agentic AI to pressure-test the investment thesis before capital deployment becomes irreversible.
Next in This Series
Part 4 will examine how these same dynamics are reshaping patent litigation and PTAB proceedings—and why the strategic calculus of enforcement is being rewritten by the same agentic tools transforming transactions.
Disclaimer
This article is for informational purposes only and does not constitute legal advice. The views expressed are those of the authors and do not necessarily reflect the views of their respective firms or clients.
About The Authors
Daniel J. Holmander
Daniel J. Holmander is a registered U.S. patent attorney and Co-Chair of the Intellectual Property Group at Adler Pollock & Sheehan…
Michel “Mike” Morency, Ph.D.
Dr. Morency has over 25 years of legal practice with general corporate law firms, primarily in the life science industry. He…
