"Next Best Action is the dominant initiative at every single customer": this is the summary by Mark Miller, managing director of Deloitte Consulting for life sciences, cited in the latest industry panels. The Next Best Action AI market reached $5.8 billion in 2024 and projections take it beyond $30 billion by 2033, according to reconstructions by IntuitionLabs and Tellius published in early 2026. Between October and December 2025 the two CRMs that together cover more than 40% of the global pharma life sciences market — Veeva with Vault CRM and Salesforce with Agentforce Life Sciences — made native AI agents for the medical representative workflow generally available, closing a decade of unscaled pilots and vendor promises.
The concrete question is no longer whether the technology exists. It is whether 2026 will be the year in which Next Best Action keeps the productivity promises it has been making for ten years. The answer depends less on model quality and more on how companies address three constraints: the actual adoption level among medical representatives, the asymmetry between the American and the European model in available data, and the boundaries of Italian compliance with the Farmindustria Code of Conduct.
What Next Best Action really does today
The most mature layer in 2025-2026 is pre-call. The system provides the representative with a priority-ranked list of HCPs with an explicit explanation: "Dr. Rossi, cardiologist, last F2F 47 days ago, historical email open rate 70%, new ESC guideline on PCSK9i published 12 days ago, suggested message on efficacy in the mixed dyslipidemia subgroup." Added to this is a recommendation of channel mix among F2F, email, webinar and remote engagement, and a prediction of the best time for contact. Veeva's Pre-call Agent announced on December 3, 2025 and Salesforce Agentforce's hands-free narrated briefing are the most recent native implementations.
The in-call layer is less mature. Veeva introduced in December 2025 the Voice Agent based on Apple Intelligence for note dictation and the Free Text Agent which performs a real-time compliance check on unstructured notes before saving. The Media Agent enables recall of approved promotional content with natural language during conversation. Real-time coaching capabilities during remote detailing — tools such as Quantified or Retorio — are still predominantly pre-visit, for roleplay and training. The post-call layer is by contrast fully operational: automatic follow-up triggers via personalized email, routing of sample or literature requests, updating of HCP segments based on engagement signals collected, and feedback to the recommendation engine which learns from acceptance or rejection of previous suggestions.
Aktana's +14% and the data that qualifies it
The most cited case study in the industry is the one published by Aktana with a top-10 biopharma: +14% sales among medical representatives who engaged with Next Best Action suggestions, +22% coverage and frequency on priority HCPs, -20% pre-call planning time, 60-80% tool adoption. Annualized incremental revenue is estimated between $2.5 and $8 million, on a lift directly attributable to NBA between 2 and 6.5% of sales. ZS Associates reports on a sales ops transformation with a top-20 pharma a 40% reduction in cycle time and 93% adoption, and quantifies in aggregate +40% customer engagement and +10% sales productivity attributable to scaled AI. IQVIA declares for OCE+ a 30% outperformance versus benchmark using its own recommendation engine. McKinsey documented in January 2024 productivity improvements between 20 and 60% in AI pilots at Pfizer, Novartis, AstraZeneca and Novo Nordisk on HCP engagement and regulatory documentation, and quantified the value of generative AI for pharma and medtech at $60-110 billion per year, of which $18-30 billion attributable to commercial operations alone.
The figure that qualifies all others is the adoption rate. Aktana reports on the LEO Pharma deployment 78% network engagement against a benchmark of 60-70%, and 48% acceptance of recommendations against a benchmark of 30-50%. The point is sharp: the +14% in sales cited above exists for reps who use the tool and accept the suggestions. For the share of the network that remains below the critical acceptance threshold, the system does not generate lift, it generates alert fatigue. The commercial lever of Next Best Action in 2026 is therefore not in AI model shopping but in process design, change management and feedback loop mechanisms that grow the acceptance rate above 50%.
The 2025-2026 newcomer: native AI agents in the CRM
Veeva announced AI Agents for Vault CRM in November 2025 and made them generally available on December 3. The first wave includes Pre-call Agent, Voice Agent, Free Text Agent and Media Agent. The technology stack is based on Anthropic and Amazon models hosted on Amazon Bedrock, with the option to build custom agents via Vault AI on Bedrock or on Azure AI Foundry. Pricing is bundled: AI Agents are included at no additional cost for Vault CRM customers. The 2026 roadmap includes Safety and Quality agents in April, Clinical Ops and Regulatory in August, Clinical Data in December.
Salesforce brought Life Sciences Cloud into general availability on October 11, 2025 and rebranded it Agentforce Life Sciences as part of the Agentforce 360 launch two days later. Functionality includes pre-call planning, voice-activated account prep, smart summaries, hands-free narrated briefing while in transit, talk-to-text for visit close-out. AstraZeneca selected Agentforce Life Sciences as a unified global platform on December 4, 2025; Novartis announced a five-year global rollout on December 17. Chiesi joined in April 2026 with a global perimeter of 3,300 users, Fidia as early adopter back in September 2025. Pricing is metered: Agentforce 1 Edition starts at $550 per user per month with a consumption-based Flex Credits system. On the AI model front, Anthropic launched Claude for Life Sciences in October 2025 and Claude for Healthcare at JP Morgan Healthcare 2026; Bristol Myers Squibb signed an enterprise deal with Anthropic dedicated to commercial; Novartis confirmed the use of Azure OpenAI copilots for HCP email generation.
Why Next Best Action is different in Europe
The factor that separates European NBA from American NBA is prescription data at the individual physician level. IQVIA Xponent — the product that tracks prescription volumes per HCP covering 93% of the retail channel in the United States — has no equivalent in Europe. National legislation and the GDPR framework make individual physician prescription inaccessible, and European NBAs must reconstruct proxies: aggregated health claims, IQVIA OneKey reference data (25 million HCPs, 6 million HCOs in 117 countries), specialty and affiliation, sample orders and sales, engagement signals such as email opens, portal visits, congress attendance. The data architectures that work in the United States, where NBA explicitly optimizes on prescription volume of the target physician, do not transpose directly into the European market.
GDPR adds a second layer. HCP data is personal data even when pseudonymized, and European pharmaceutical companies must base processing on legitimate interest or explicit consent, with HCP access and deletion rights that the system must be able to fulfill. The EDPB included transparency and consent as a focus of the 2026 coordinated enforcement action, and in 2025 sanctions in the healthcare sector grew by 26% compared to the previous period. For those designing a European NBA, this means that the engagement model must be built explicitly on the physician's voluntary opt-in to the system and on traceability of every message sent, not on passive aggregation of prescription data. It is a difference of architecture, not of tuning.
The Italian constraint: 2024 Code of Conduct and change management
The Farmindustria Code of Conduct, last revised October 9, 2024 and again on February 21, 2025, has introduced relevant clarifications for Italian Next Best Action. Article 3.25 regulates activities toward non-prescribers; article 3.26, revised in 2024, regulates information to the public and reactive responses by non-commercial personnel. The operational constraint for an NBA system is twofold: every message suggested to the representative must be classified unambiguously as promotional or non-promotional, and the classification trail must be auditable years later. An NBA system that explicitly optimizes "brand X sales to Dr. Y" sits in an area of reputational and disciplinary risk incompatible with the Italian code: law and code forbid direct incentives based on prescription volume of specific drugs.
The Italian operational context is documented by IQVIA data published in January 2024: 36.2 million scientific information activities in 2023, in line with 2022 but still -29% compared to the 50.8 million of pre-pandemic 2019. The channel mix is 82% face-to-face, 11% email, 3% phone, with F2F rising to 83% in January 2024. The Italian pharmaceutical industry closed 2024 with 56 billion euros of production, +12% revenue, +21% investment, +16% export and 71,000 sector employees, according to the 2025 Farmindustria Assembly. For an Italian company that has to put Next Best Action into production in 2026, the critical point is not the choice between Veeva and Salesforce. It is process design: how to translate global blueprints into Farmindustria constraints, how to measure actual acceptance of recommendations beyond the critical 50% threshold, how to involve the sales network in a model in which AI suggests but does not replace the professional relationship. The +14% in sales exists, but is not included in the license.
