What is the governed middle? Bridging dry lab AI and wet lab execution

By Anirban Mudi, Lead Platform Product Manager, IDBS

Across pharmaceutical R&D, in silico science is producing insight faster than ever. Yet for many biopharma organizations, a persistent question remains: if the dry lab is learning faster, why isn’t the organization producing better evidence faster? The answer rarely lies in algorithm quality. It lies in how insight travels between the computational and physical worlds of science, and whether it reaches the bench in a form that supports governed execution and long-term reuse.

At IDBS, we focus on making time a competitive advantage for our customers. From first hypothesis to validated result, our goal is to compress cycle times with speed and certainty – without compromising quality. That requires a lab loop designed for governed learning, not just faster modeling.

Dry lab, wet lab and the lab loop

Figure 1. The IDBS Lab Loop

The dry lab is where in silico experimentation takes place: computational chemistry, bioinformatics, prediction and hypothesis generation supported by AI. The wet lab is where those hypotheses are tested through synthesis, bioassays, analytical testing, stability studies and process development under real scientific and regulatory constraints.

The connection between these two environments is the lab loop (Figure 1). As the image illustrates, it is a continuous cycle. Models in the dry lab generate hypotheses. Those hypotheses are translated into experimental execution in the wet lab. Results are captured as evidence, and that evidence feeds back to refine the next round of models, decisions and experiments. For AI to contribute meaningfully in regulated environments, its role in shaping hypotheses, experimental choices and decisions must also move through this loop in documented form, so it can be reviewed and scrutinized alongside the resulting evidence. In regulated pharma, this loop matters because the objective is not simply faster insight. It is faster, governed learning that produces evidence that is traceable, attributable and defensible.

Why AI progress outpaces the loop

AI has dramatically strengthened the dry lab side. Yet while AI experimentation is nearly universal across life sciences, only a small minority of organizations report it delivering sustained business value at scale.¹ Many also lack confidence that their data practices can support AI beyond pilots.²

The bottleneck is what happens after insight is generated. Three failure modes typically appear:

• Insight arrives informally. Outputs from the dry lab often show up as slide decks, notebooks, documents or informal recommendations rather than structured, executable intent.
• Context is lost. Wet lab teams must reinterpret these outputs into protocols and workflows. Critical context – such as why a candidate was ranked higher, which assumptions drove a DoE or why a drug candidate was deprioritized – is often implicit or missing.
• Feedback breaks down. Results cannot be traced cleanly back to the originating model, and execution data returns stripped of the context needed to make it reusable for the next round of learning.

Is your data ready to support the lab loop? Take the AI data readiness assessment to find out where your foundations stand.

Where the gap becomes a regulatory risk

In regulated pharma, a broken lab loop is a compliance exposure, not just an efficiency problem. Many biopharma QC labs still operate with disconnected systems and limited automation.³

When AI recommendations enter the scientific workflow, they become part of the decision record. In GxP environments, that record must meet ALCOA+ standards. In practice, AI cannot remain an external source of recommendations. Its role in generating, shaping or influencing decisions must be documented within the workflow itself, so reviewers can understand not only the outcome but how AI contributed to it. If an AI output influences a batch decision or stability protocol, regulators will expect to see how it was generated, what data it relied on and who approved it.

The FDA’s draft guidance on AI in regulatory decision making formalizes this expectation, requiring a risk-based credibility assessment for each AI model, documentation of its context of use and evidence of how credibility was established and maintained.⁴ Most lab environments are not designed for this level of accountability. AI outputs sit outside the governed record, creating gaps that remain invisible until inspection.

This risk is compounded by shadow AI, with scientists adopting unvetted tools outside governed systems. Ungoverned AI introduces invisible drift-decisions shaped by unvalidated models and untraceable recommendations that may not surface until a deviation or audit finding.

Learn how IDBS builds AI-ready, governed data foundations.

The need for the governed middle in a lab loop

Research has begun to demonstrate scientific loops in which AI proposes experiments, automation executes them and results feed back with human oversight.^5,6 These models are promising. In regulated pharma, the challenge is not simply to close the loop. It is to govern the lab loop so that dry lab insight is translated into wet lab execution, and wet lab outcomes return as trusted, reusable evidence.

A governed lab loop ensures that insight from the dry lab is captured as structured, auditable intent rather than informal guidance. Execution preserves that intent, deviations are recorded explicitly and outcomes can be traced back to upstream assumptions. AI outputs that influence decisions become part of the regulated record. Feedback returns with execution context, and human review is built into defined decision points.

This is where the governed middle becomes critical. Many organizations invest in dry lab capability and wet lab infrastructure but underinvest in the layer that connects them. It is the place where scientific intent is preserved, execution is auditable and evidence becomes reusable across programs, sites and time. That same middle layer is what allows AI to integrate seamlessly into the lab loop. Instead of sitting outside the process as an informal source of recommendations, AI becomes part of a documented workflow in which its outputs, assumptions and influence on decisions can be reviewed and scrutinized as part of the scientific and regulatory record. When that middle is weak, the lab loop fragments into disconnected workflows, lost context and non-reusable results.^7,8 When it is strong, the lab loop becomes a governed system for compounding scientific learning.

For more than 35 years, IDBS has partnered closely with the world’s leading biopharma companies to solve hard problems in regulated, complex environments. We co-design solutions with customers so the governed middle reflects real workflows, real constraints and real decision points.

From faster insight to governed evidence

Closing the gap means designing the lab loop as a governed system. Connected execution delivers tangible results: fewer errors and deviations, improved compliance and shorter testing timelines.³ Organizations that capture disproportionate AI value do so by redesigning workflows and strengthening governance, not by deploying more models.⁹

Seen in that context, the governed middle is not just an architectural idea. It is the operational layer that turns insight into evidence. It is also the layer that makes AI participation visible, documented and reviewable within the governed scientific process, rather than leaving it to be inferred after the fact. IDBS Polar is one example of how that layer can be established through structured data capture, workflow governance and contextualized evidence.

While AI capabilities continue to evolve, the foundation-structured intent, governed execution and traceable evidence-is the key to unlocking AI insights you can trust. AI is only as reliable as the data, context and governance it sits on.

As part of Danaher’s connected ecosystem of expertise, technologies and capabilities, IDBS helps customers move from idea to impact with greater speed and certainty. We bring rigorous methods and cross-portfolio strengths to help you scale connected science globally – and keep it governed.

Ready to connect insight to evidence?

If you’re looking to turn AI-generated insight into governed, reusable evidence, the first step is understanding where your data foundations stand. Take the AI data readiness assessment to evaluate your organization’s ability to support AI across the lab loop.

To go deeper, explore how a structured, context-first approach to data underpins scalable and trustworthy AI:
Read the context-first AI data strategy whitepaper for biopharma. Or contact us to discuss how your organization can build the governed middle that connects insight to evidence.

IDBS, a Danaher company, helps biopharma organizations turn scientific data into trusted insight that accelerates innovation and improves patient outcomes.