Advertise Inquiries | About Us | | EventsSubscribe
Biotech industry elites gathered in San Diego on June 22 for the second annual AI Summit at 2026 BIO International Convention (BIO 2026). Executives and researchers packed the kickoff session to understand how computational tools are moving beyond theoretical promises into daily operational realities. STAT News health tech reporter Brittany Trang moderated a robust discussion with Sajith Wickramasekara, Co-founder and CEO of Benchling, a leading cloud-based biotech R&D platform.
The conversation centered on newly released data from a comprehensive Benchling survey of cutting-edge life sciences companies. Wickramasekara detailed how organizations currently utilize these advanced tools and where the industry still faces significant roadblocks. This inaugural session grounded the broader summit in evidence and practical use cases rather than speculative hype. The insights provided a critical foundation for the seven marquee sessions happening throughout the afternoon at the global convention. Attendees sought clarity on strategic resource allocation and the true return on investment for their rapidly expanding technology budgets.
Addressing the gap between industry excitement and practical application in drug discovery, Wickramasekara noted that frontier model capabilities currently exceed widespread adoption in the field. While acknowledging that hype often overshadows daily operational reality, he maintained an optimistic stance, suggesting that the entire ecosystem stands to gain significantly from existing foundational tools. The challenge for leadership lies in pinpointing exactly where these technologies can deliver immediate, tangible value.
Trang pointed out survey results showing where organizations focus their computational efforts. Protein structure prediction leads the entire sector with a massive 73% adoption rate among surveyed scientific models. General purpose large language models (LLMs) closely follow at 72%. When breaking down specific research workflows, literature review and knowledge extraction show the highest concentration with 76% of companies actively using AI tools in this area. Scientific reporting and internal communication also demonstrate heavy usage at 66%. Meanwhile, more specialized scientific applications trail behind these leading use cases. Docking and binding prediction models see a 52% adoption rate. Target identification sits at a solid 58%.
Wickramasekara observed a surge in biopharma innovation centered on antibody and protein engineering. By leveraging models from pioneers such as VantAI (rebranded as Proxima in early 2026), Latent Labs, and Chai Discovery, scientists can now bypass months of conventional wet lab experimentation. He described this shift toward generative design as a powerful alternative to physical screening, noting its potential to serve as a digital surrogate for live mouse models. While adoption for the generative design of proteins and molecules currently stands at 36%, other critical areas like molecular property prediction and ADMET analysis are still in the early stages of maturity, with only 28% of surveyed firms utilizing AI in those domains. Looking ahead, he expects major pharmaceutical companies to refine their own proprietary models by capitalizing on their unique internal datasets and experimental expertise.
Effective implementation of these advanced tools necessitates a fundamental overhaul of conventional software deployment methods. Wickramasekara cautioned against the use of phased rollouts in which IT departments try to dictate specific prompts or strictly limit usage. He argued that such restrictive measures are ineffective, as it is nearly impossible to predict who the natural early adopters will be. Given the diverse ways computational tools can be applied, companies should instead offer widespread access to observe how staff members incorporate the technology into their specific functions. Management should identify these instinctive power users and highlight their successful processes to encourage adoption throughout the organization.
Financial anxieties often stall these broad implementations before they even begin. Executives frequently express deep fears about massive and unpredictable cloud computing bills. Wickramasekara suggested that biotech firms should aggressively adopt these tools until budget constraints are explicitly raised by finance departments, noting that productivity improvements are difficult to quantify prior to organic integration.
The Benchling CEO also stressed that successful adoption requires both employee enthusiasm and visible leadership involvement. Executives should not simply act as cheerleaders from the sidelines, they must personally use the technology and demonstrate practical workflows during staff meetings. To illustrate this commitment, Benchling dedicated an entire week to allowing staff to pause standard operations and redesign their daily workflows using collaborative AI models.
Data from the survey elucidated the difficulties encountered during pilot stages: 36% of organizations cited poor integration with current workflows and change management as a major obstacle, with another 47% viewing it as a minor issue. Notably, despite the common assumption that executive backing is a critical hurdle, 45% of surveyed companies indicated it played no role in the failure of their pilot programs.
Despite the massive enthusiasm surrounding new computational capabilities, data integrity remains a major structural barrier. Trang highlighted that over half of surveyed organizations identified data availability and quality as primary challenges during their pilot programs. Wickramasekara explained the core fundamental problem underlying this frustrating statistic. He noted that developers train standard large language models on public internet data like textbooks and publicly available articles.
However, the vast majority of valuable scientific information remains proprietary, highly heterogeneous, and hidden deep inside corporate firewalls. Consequently, these general models lack the specific scientific context required to excel at complex reasoning out of the box. Algorithms require highly structured digital environments to function properly and generate accurate predictions. Wickramasekara detailed how his company spent a decade trying to digitize laboratories because structured records of molecules and experimental decisions provide incredibly fertile ground for computational analysis. Companies must build a robust and clean data foundation before they can feed information into diverse computational initiatives. Without this critical infrastructure layer, organizations will struggle endlessly to take full advantage of the heavy investments they have made in advanced algorithms.
While many survey respondents look forward to generating new intellectual property (IP) and reducing development expenses, these high-level objectives remain largely unfulfilled for most organizations so far. Data from the presentation clarified the expected timeframe for these financial gains. Currently, only 23% of companies report a tangible impact on cost reduction, though a significant 56% anticipate achieving this within the next year or two. A similar trend exists for revenue and IP generation, where only 16% see current results compared to 39% who expect them in the near future.
Wickramasekara reframed the return on investment conversation by focusing on operational speed rather than immediate baseline cost reduction. He pointed to extensive inefficiencies embedded in the drug design and development lifecycle. Although computational tools are unable to accelerate the biological growth of physical cells, they offer a powerful means to alleviate the administrative and analytical burdens that typically slow down experimental workflows.
This emphasis on speed corresponds with findings from the survey, which reveal that 50% of firms are currently benefiting from faster science and time efficiency, with another 33% anticipating these benefits in the next one to two years. Wickramasekara highlighted a notable instance with Prime Medicine and their efforts to submit an Investigational New Drug (IND) application. To satisfy new regulatory standards for cell and gene therapies, the firm had to validate 35 distinct products for a single complex treatment. By employing an automated scientific assistant, researchers were able to synthesize historical data and develop subsequent studies in only a few days, a process that would typically take several months. Furthermore, a leading pharmaceutical firm utilized specialized digital agents to analyze 10,000 past experiments prior to starting new oncology mouse models. The software identified that 18 of the 20 intended studies had been completed years earlier. Such advanced search functions help mitigate the significant loss of institutional memory that often occurs in large firms during staff transitions.
The Benchling kickoff discussion at the AI Summit set the stage for broader industry themes explored during BIO 2026. A primary area of interest for leadership is the progression toward agentic science within laboratory settings. These systems are advancing from data processing to independently establishing objectives, refining strategies, formulating hypotheses, and managing experimental protocols. This evolution redefines the technology, transitioning it from standard software into an autonomous digital partner equipped to manage intricate clinical trials, fast-track protein engineering, and revolutionize early-stage drug development.
The AI Summit also highlighted quantum computing as an imminent reality for drug discovery rather than a distant science fiction concept. Industry experts marked the 2025 to 2026 period as a decisive turning point for biopharma quantum applications. A landmark achievement occurred in March 2026 when IBM and the Cleveland Clinic finalized an innovative hybrid quantum-classical workflow to simulate macromolecules for chemical, materials science, and medical research. Using this system, researchers successfully modeled the electronic structure of the 303-atom Trp-cage protein, a synthetic 20-amino-acid miniprotein featuring hydrogen bonding and a hydrophobic core typical of much larger biological structures. The scalability of this approach was further demonstrated in May 2026, when a collaborative team from IBM, the Cleveland Clinic, and RIKEN simulated a record-breaking 12,635-atom protein complex. These advancements represent a significant leap into the era of quantum utility, positioning hybrid quantum systems as functional, tangible tools for modern biomedical inquiry.
Researchers and executives also discussed the critical need for federated learning frameworks to protect sensitive health information. These privacy preserving systems allow allied institutions to train shared models across decentralized servers without ever pooling raw patient data centrally. For biotech executives navigating strict regulatory landscapes, this method represents the ultimate key to balancing patient data privacy with the need to build powerful computational capabilities.
Furthermore, attendees examined massive financial movements currently reshaping the global landscape. The core challenge for many organizations has officially shifted toward artificial intelligence platform commercialization and data monetization. Companies must now figure out how to monetize their proprietary data, achieve strict standards for model transparency and rapidly accelerate the translation of computational results into clinical assets. Massive industry megadeals validate this new commercial focus. For instance, the recent $2.5 billion partnership between Insilico Medicine and SK Biopharmaceuticals demonstrates exactly how computational platforms now command premium valuations within the broader market.
As the biotech sector rapidly integrates advanced generative AI models, automated digital agents, and predictive computational platforms, organizations face critical strategic choices regarding their technology infrastructure. During the kickoff session of this year’s BIO AI Summit, Wickramasekara suggests that companies should avoid reliance on a single provider. Instead, he promotes a multi-model strategy that utilizes high-end frontier models for complex scientific reasoning while leveraging open-source options for day-to-day tasks. This approach is particularly economical because open-source models generally remain only 6 to 12 months behind commercial industry leaders, allowing companies to scale their operations effectively without depleting their research funding.
Ultimately, the competitive divide between traditional biotechs and specialized tech-bio firms matters far less than the collective advancement of the industry. The ultimate metric of success remains moving life-saving therapies to patients faster. If automated workflows and digital predictions can yield even a modest increase in notoriously low clinical success rates, the entire global pharma market will experience unprecedented expansion. Both established legacy companies and computational startups have an extraordinary opportunity to lead this new era of digitally-native drug discovery.
Leave a Reply