Most pharmaceutical leaders are right to be skeptical about AI. But market access is the one area where the evidence is already in — and the results are hard to ignore. Real benchmarks, global HTA context, and what HEOR and market access teams need to know in 2026.

The pharmaceutical industry has committed billions of dollars to artificial intelligence over the past three years. The results have been, for the most part, disappointing. A ZoomRx survey found that 83% of life sciences professionals consider AI to be overhyped. Deloitte reports that only 9% of biopharma leaders have seen real return on their AI investment. And yet McKinsey estimates that generative AI could unlock between $60 and $110 billion in annual value for the sector.

The contradiction is real and deserves a specific answer: it is not that AI fails. It is that it has been applied in the wrong places. This article explains why pharmaceutical market access is the exception — the function where generative AI is already delivering measurable, verifiable, and reproducible results — with direct implications for HEOR teams, market access directors, and medical affairs professionals operating in Latin America and the United States.

Why this matters now

ISPOR ranked AI as the number one trend in HEOR for 2026–2027, up from position 3. Yet only 7 out of 5,000 HTA submissions to agencies such as NICE, CADTH and TGA indicated AI use in their preparation. The gap between professional interest and practical adoption is the single largest differentiation opportunity in market access today.

1. Why AI skepticism in pharma is justified — in most areas

Pharma leaders' skepticism is not irrational. It has an empirical foundation. AI has failed systematically in the areas where the most investment has gone:

Drug discovery: timelines run for decades, variables are near-infinite, and success is binary. AI can accelerate parts of the process, but the ROI horizon is too long to justify current investment levels.
Clinical trials: study design depends on regulatory criteria, heterogeneous populations, and endpoints that shift with the scientific context. AI helps in recruitment and monitoring, but does not transform the core process.
Manufacturing: benefits are real but incremental. Supply chain optimization requires a data infrastructure that most companies do not yet have.

In all these cases, the pattern is the same: open-ended processes, long time horizons, uncontrollable variables. That is precisely the context where generative AI struggles to create verifiable short-term value.

The root of the problem

McKinsey describes the issue precisely: simply adding AI to existing processes does not generate value. In a ZS survey of pharma technology executives, the dominant fear is that generative AI will follow the path of omnichannel: "it showed up everywhere — except in our bottom line." The concern is legitimate — but it assumes that all pharma functions have the same process profile. They do not.

2. What makes market access different: the process profile AI actually needs

The right question is not whether AI works in pharma. It is which type of process AI works for. ISPOR's 2026–2027 HEOR trends report identifies AI as the most significant transformation in the field — precisely because market access brings together the conditions that make generative AI effective:

Pharma function	AI investment	ROI horizon	Suited for generative AI?
Drug discovery	High	10+ years	No — open-ended, unpredictable
Clinical trials	High	5–10 years	No — multiple uncontrollable variables
Manufacturing / supply chain	Medium	2–5 years	Partial — process optimization
Market access / HEOR	High	6–18 months	YES — structured, evidence-based process

Pharmaceutical market access — building value dossiers, cost-effectiveness models, systematic literature reviews, budget impact analyses — meets precisely the criteria in the last row:

Structured: follows standardized methodologies (ISPOR, CHEERS, GRADE) with defined, verifiable steps.
Evidence-based: operates on verifiable data extracted from indexed literature, epidemiological databases, and regulatory sources.
Repeatable: the same process is replicated for each product, indication, and market — with known and bounded variation.
Time-defined: outcomes (coverage, pricing, payer negotiation) are measurable in months, not years.

3. Real benchmarks: what AI is actually achieving in market access today

These are not projections. The results are already measurable and documented in scientific literature and implementation reports:

Market access process	Traditional timeline	With AI	Documented benchmark
Systematic literature review	3–6 weeks	2–4 days	~80% reduction in screening time
Cost-effectiveness model (ICER)	4–6 weeks	3–5 days (local adaptation)	Error margin under 1% vs. published models
Full value dossier (GVD)	4–8 months	6–8 weeks	60% time reduction documented
Local market adaptation	2–4 weeks/market	2–5 days/market	Scalable to multiple markets simultaneously

80%

Screening time reduction

A systematic review that previously took four weeks can be ready in three days. That is not marginal efficiency — it is the difference between reaching the payer within the decision window and arriving after the competitor has already closed the deal.

4. The Latin American context: why the window is open right now

The case for AI in market access holds globally. But in Latin America it carries an additional dimension that makes the window especially urgent.

4.1 Regional HTA agencies are maturing their methodologies right now

Latin American health technology assessment agencies, coordinated through the PAHO RedETSA network — which convened 37 institutions from 19 countries at its 16th meeting in Buenos Aires in June 2025 — are evolving their methodologies at an unprecedented pace:

HTA Agency	Cost-effectiveness threshold	Key criteria	MEA acceptance
Brazil — CONITEC	3× GDP per capita (ref.)	Local evidence, real-world data	Yes — Zolgensma MEA (Mar 2025)
Colombia — IETS	Up to 3× GDP (debated)	Budget impact + national evidence	Partial — criteria evolving
Mexico — CENETEC/IMSS	Not formalized	Budget efficiency + comparative price	No — centralized procurement prioritizes cost
Argentina — IECS	Not formalized	Multiprofessional criteria, methodological quality	Emerging — no formal framework yet

CONITEC signed the region's first outcomes-based access agreement (MEA) for gene therapy in March 2025 (Novartis/Zolgensma). That signals a clear direction: greater methodological sophistication, higher demand for local evidence, and more competitive advantage for those who arrive prepared.

4.2 Brazil's new health data infrastructure creates unprecedented opportunities

In July 2025, Brazil launched the National Health Data Network (RNDS), a real-world data infrastructure that opens unprecedented possibilities for local evidence generation. Simultaneously, Brazil's AI Law (PL 2338/2023) is advancing in Congress and the National AI Plan allocates approximately one-third of R$23 billion to the health sector. Companies that build AI capabilities for market access in Brazil now will have access to data their competitors cannot use retroactively.

4.3 The regulatory gap is a window, not a warning sign

No HTA agency in Latin America has published specific methodologies for evaluating AI-generated evidence. CONITEC has no guidelines. IETS has no guidelines. That is not a red flag — it is the opportunity to set the standard before it becomes mandatory.

Trust capital

Companies that adopt transparent AI practices aligned with international frameworks such as ELEVATE-GenAI and ISPOR CHEERS-AI will build credibility with agencies before formal requirements exist. That trust capital is priceless once regulators start asking the question.

5. Which market access functions benefit most from generative AI

Evidence synthesis (systematic reviews and meta-analyses)

This is the use case with the strongest evidence base. LLMs can screen thousands of references, extract efficacy and safety data, and generate structured summaries in a fraction of the traditional time. Precision is comparable to expert human reviewers when used with appropriate validation protocols.

Economic modeling (ICER, BIM, threshold analysis)

GPT-4 has demonstrated the ability to replicate published cost-effectiveness models with an error margin below 1% in under 15 minutes. The most valuable practical application is not building models from scratch — it is adapting existing models to local parameters: epidemiology, unit costs, and willingness-to-pay thresholds for each market.

Value dossier construction (GVD and local dossiers)

The ISPOR Working Group presented an AI-assisted co-authoring tool for the Disease Overview section of Global Value Dossiers at the 2025 Montreal Annual Meeting. The result: 80% of AI-generated interpretations required no editing. idalab's EPRI tool documented a 60% reduction in HTA dossier writing time.

Differentiated value communication by payer profile

One of the most common errors in market access is using the same value message with different decision-makers. A public payer at CONITEC and a medical director at a private insurer in Mexico operate on completely different decision criteria. AI can generate calibrated versions of the same value argument for each profile, based on the specific criteria of each audience.

What AI does not replace

AI in market access does not replace the methodological judgment of the HEOR expert, the contextual knowledge of the local market, or the relationship with the payer. What it does is eliminate the volume bottleneck that prevents small teams from competing with large-organization resources. A team of three people with AI can produce the output that previously required ten.

6. Methodological rigor: ensuring AI-generated evidence is accepted

The question that worries market access teams most is not whether AI works. It is whether the evidence it produces will pass HTA agency scrutiny. The relevant standards in 2026:

ELEVATE-GenAI: the first standardized reporting framework for LLM use in HEOR research, published by the ISPOR Working Group. It establishes criteria for transparency, traceability, and validation.
CHEERS-AI: an extension of the CHEERS checklist for health economic evaluations incorporating AI and machine learning methods.
PALISADE: a specific checklist for machine learning methods in HEOR and RWE research.

NICE's position is the most explicit of any global HTA agency: AI use must be declared, transparent, and reproducible, and results must meet the same quality standards as traditional methods. In Latin America, where no agency has published specific guidelines, alignment with these international frameworks is the best available practice.

A note on regulatory acceptance

The fact that CONITEC and IETS have no specific AI guidelines does not mean they will accept anything — it means they apply their existing methodological criteria. A dossier with AI-generated evidence that meets ISPOR standards will be evaluated by the same criteria as a conventional dossier. The risk is not using AI: the risk is using AI without documented methodological rigor.

Frequently asked questions

Why is market access different from other pharma functions for AI application?

Because it meets the conditions that generative AI needs to create real value: a structured process, based on verifiable evidence, repeatable, and with a defined evaluation horizon. Drug discovery and clinical trials are open-ended processes with decade-long horizons. Market access produces measurable outcomes — coverage, pricing, payer negotiation — in months.

What type of evidence can AI generate for an HTA submission?

AI is particularly effective for: (1) clinical evidence synthesis via assisted systematic review, (2) adapting cost-effectiveness models to local contexts, (3) constructing the narrative of the value dossier, and (4) sensitivity analysis in economic models. In all cases, AI outputs must be validated by a methodological expert before submission.

How does the Latin American HTA context differ from Europe or Canada?

European agencies (NICE, G-BA, HAS) have formalized methodologies and explicit thresholds. In Latin America, formalization varies: CONITEC in Brazil is the most methodologically advanced; IETS in Colombia operates with evolving criteria; Mexico has no formalized threshold. This means that in LATAM, the quality of the value narrative and local adaptation carry more relative weight than in systems with rigid thresholds.

What methodological frameworks are relevant for AI-generated evidence in market access?

The most relevant in 2026 are ELEVATE-GenAI (ISPOR), CHEERS-AI for health economic evaluations, and PALISADE for machine learning methods. NICE has published an official position requiring that AI use be declared, transparent, and reproducible. For LATAM, where no local guidelines exist, alignment with these international standards is the recommended practice.

Is there evidence that LATAM HTA agencies are accepting AI-generated evidence?

There is no formally documented precedent in the region yet. What does exist is the NICE (UK) precedent: in August 2025, NICE launched two HTA Innovation Laboratory projects exploring AI in economic modeling and HTA process automation. The trajectory is clear — agencies are actively evaluating how to incorporate AI, not whether to incorporate it.

Conclusion: the window is narrow and it is open now

The 83% of pharma executives skeptical about AI are right — about the wrong contexts. AI has underdelivered in long-horizon R&D, in manufacturing without adequate data infrastructure, and in digital transformation initiatives unanchored to real processes.

But market access is different. It is the pharma function that best aligns with the strengths of generative AI: structured, evidence-based, repeatable, and with short-term measurable outcomes. The benchmarks already exist: up to 80% less time in systematic reviews; cost-effectiveness models adapted in days with less than 1% error; value dossiers built in weeks instead of months.

6–18 months

Market access ROI horizon

In Latin America, HTA agencies are maturing their methodologies right now. The next 24 months will define the rules of market access in the region.

In Latin America, the argument carries additional urgency. HTA agencies are maturing their methodologies right now. Brazil's new health data infrastructure and the evolution of CONITEC, IETS, and RedETSA signal that the next 24 months will define the rules of market access in the region. Companies that arrive with solid evidence, calibrated narratives, and trained teams today will have a competitive advantage for years.

The window is open. It will not stay open indefinitely.