AI models are versatile products: they can double as both finished goods and raw materials. Models can serve as the foundations for other models through fine-tuning, architecture modifications, in combinations with other models, as synthetic data generators for other models. They can be used to create higher-abstraction compound products without explicit knowledge of the original. They can be continually modified and repackaged with new injected data and new value added. They can even be formulated jointly by multiple parties without sharing ingredients.
When a product has such high flexibility, value chain dynamics gain a tremendous amount of fluidity. Multiuse potential lends itself to non-competing partnership and collaboration design; adaptability creates opportunities to repurpose capabilities used for downstream value creation into upstream commercial assets with controllable competitive risk. In biomedical AI, this flexibility is compounded by the proliferation of different data modalities and the integration of wet-lab workflows, creating a vast space of technical and commercial possibilities. As AI efforts advance, biopharma and tech players alike are sizing this opportunity space and deciding where to position themselves. Their choices are beginning to shape a new and dynamic value network and creating the potential for earlier revenues, diversified value propositions, and innovation mechanisms that sit at the intersection of technical and commercial design.
But as this new option space is starting to remodel the biopharma value network, reality lags behind opportunity. Non-AI-native biotech companies looking to integrate AI into R&D face a complex set of questions and decisions: where the highest ROI is, whether the goal is innovation or acceleration of existing processes, whether to build or buy or partner, what their internal readiness is for adoption. They often face bottlenecks in hiring specialist AI expertise and have concerns about outsourcing AI builds that involve proprietary data. Capability-focused AI-first companies face a corresponding set of strategic choices that can both restrict and create adaptability in a rapidly expanding competitive landscape. Early choices on data collection mechanisms (outsourcing, internal, or co-development) can carry long-term effects on market positioning and optionality in conflict with short-term needs. Prioritization of marketing versus technology-market fit exploration at the wrong time can lead to later pivots or miss timely opportunities. Proving partnerability can be challenging in an industry with limited open benchmarks and high concerns about data sharing.
AI's technical versatility creates unique pathways for companies navigating this landscape and willing to combine technical and commercial innovation. Non-AI-native bioplatforms most often consider AI as an enablement layer or for internal innovation only, while outward-facing commercial activity on AI capabilities is generally carried by AI-first companies. This is a natural consequence of resource allocation on each side and a perception that AI as a commercializable asset implies foundation model-scale investment. But as with every industry, products come in all scales; AI capabilities can be differentiated through unique biological insight that played into data collection and processing or the development of fit-for-purpose training workflows. Lower generalizability limits the market, but the investment needed for development will be correspondingly lower and proportionate to the expected internal payoff. A model intended to serve a niche use case could be developed with design choices that facilitate somewhat broader applicability or repurposing toward "derivative" use cases at low marginal development cost and acceptable overall spend.
Not every use case and prospective capability is a candidate for repurposing or broader applicability, but where relevant it can change considerably the calculus on resourcing, investment value propositions, and build vs. buy decisions and avoid the common pitfall of investing in AI development enough to commit but not enough to succeed. The cost of hiring and maintaining technical expertise can be better justified if development brings in earlier revenues through external licensing, while outsourcing or partnering on a prospective capability can prevent future commercialization of an attractive asset. Wet-lab-focused companies are in fact well-positioned to act as opportunistic AI capability prospectors: they have the intuition and expertise to identify biologically valuable adjacent use cases and determine whether these cases create unacceptable competitive overlap in therapeutics development. But it is essential that this exploration be accompanied by a technical one to inform on build optionality and feasibility.
Strategic innovation should not stop at product. As AI becomes increasingly more integrated in biopharma R&D, the difference between "AI-first" and "wet-lab-first" companies will likely erode; instead, we are more likely to differentiate companies by model scale and by how they view value realization from AI: upstream through capability licensing, downstream through internal enablement for therapeutic development, or both. For reasons ranging from capacity constraints (a company not equipped to fully capitalize on its digital stack in therapeutics development) to capability designs that enable multiuse and non-competing applications, AI value transfer across the industry is likely to keep increasing. With the pace of AI technical development outpacing the ability of biopharma BD to form partnerships, innovation on the engagement mechanisms that facilitate this value transfer is critical for ensuring that new capabilities can be adopted quickly when the fit is right. There is a perceived risk of engaging with newcomers when the data shared is highly proprietary and the outcome is uncertain, but there are technical and commercial means available to reduce it. Performance milestones that unlock increased data sharing, use of confidential computing platforms, and development of custom autonomous agentic workflows for partner-side privacy-focused exploration of data-model fit can lower engagement barriers. AI capability-providing companies that productize these into cohesive engagement packages can have a stronger entry point to a competitive and reputation-focused market, but this requires balancing technical feasibility with commercial demands and building with both performance and privacy in mind.
Companies looking to innovate in life science AI have access to a variety of strategic levers that draw on unique features of AI as capabilities and products but that are underutilized by the industry today. While the playbook looks different for each company, the prerequisite is the same: removing the silos between technical thinking and strategic direction. I founded Coescent to help wet-lab-first and AI-first companies alike close the technical-strategic AI innovation gap. Coescent is focused on AI-driven strategy development and value discovery across the spectrum of biotech R&D, guided by an approach that cuts across technical, scientific, and commercial dimensions. Coescent is fundamentally multimodal: the complexity of data modalities and needs across the drug discovery and development process makes it imperative to consider the potential of AI integration across all relevant biological, chemical, and natural language-based modalities involved in a company's R&D platform.
As the bioAI value network continues shifting and forming in the coming years, companies can proactively take steps to enhance their place in it through strategic innovation and new mechanisms for AI-driven value realization. The guiding principle of Coescent engagements is that technical bottlenecks can be solved with creative strategy and strategic quandaries can be resolved through technical insight. Every engagement is an opportunity to innovate on both fronts.