The Intelligence Layer: Between Evidence and Execution

An op-ed by George Kwiecinski.

Most software sold to quality and regulatory teams falls into one of two categories.

1. Workflow applications, such as QMS modules, checklists, and CAPA trackers, organize work that a team already understands.

2. Data feeds deliver records, but often leave the user to separate signal from noise. One gives people a place to record decisions. The other gives them raw material.

Neither necessarily explains what the regulatory world is doing, what the evidence means, or how it should inform the next inspection, audit, supplier decision, or product decision.

The gap between reality and records

A system of record is not reality itself. Reality is the run you take in the park. The health application is where the distance, route, heart rate, and time are recorded. The measurements matter, but meaning appears only when someone interprets them in context.

Regulatory work has the same gap. An inspection occurs. An investigator documents observations. A company records its response in a quality system. The source document preserves evidence, and the workflow system preserves activity, but neither guarantees that meaning travels between them.

The difficult work sits in the middle: understanding the evidence, connecting it to related rules and entities, testing it against other records, and deciding what it means for the organization.

Intelligence is interpretation

I use intelligence layer to describe the combination of source documents, structured and unstructured data, domain interpretation, and interfaces that allow people and software to reason from the same evidence.

The data must come first. For regulatory intelligence, that can include FDA Form 483 observations, warning letters, recalls, medical-device adverse-event reports, rules, companies, facilities, investigators, and agency offices. These records become more useful when their relationships are structured and traceable rather than merely scanned and stacked.

More data, however, does not automatically create more intelligence. General AI systems can cover a very large but shallow pool of subjects. A narrowly engineered application can reason deeply within one small workflow. A useful regulatory intelligence layer must combine breadth with domain depth.

Breadth without depth and depth without breadth solve different problems. An intelligence layer needs both.

This distinction also explains why professional judgment remains essential. In a 2026 KeyPedia Labs survey of quality professionals, one respondent summarized quality intelligence this way:

“I’ve been in industry. I understand what the regulators will do.”

• Director of Quality at a Midsize Biotech

Experience creates valuable judgment, but individual memory is not a referenceable evidence base. An intelligence layer should help professionals test that judgment, compare it with the record, and explain the basis of a decision.

Four operating speeds

At the speed of light.

At the speed of compute.

At the speed of evidence.

At the speed of regulation.

Each phrase represents a different operating standard. The speed of light is a theoretical physical limit. The speed of compute is the pace of machine execution and technical iteration. Evidence must become usable before regulation reaches the organization through institutional review, response, or enforcement.

Regulated industries operate under a different constraint. Regulation is evidence-based, procedural, and slow by design. The organizations affected by it, however, make patient-facing and product-facing decisions continuously. Waiting for the next enforcement action is not a reasonable intelligence strategy.

Intent matters, but execution determines the outcome. A team may genuinely intend to comply, yet poor execution can still harm patients, interrupt supply, or create an avoidable enforcement risk.

The useful goal for regulated organizations is not speed for its own sake. It is the speed of evidence: making source material available, structured, and interpretable early enough for people to act before the institutional response arrives.

Our mission: make evidence usable before it becomes enforcement

Our mission is not to add another screen to regulatory work. It is to make the world’s regulatory evidence usable before it becomes an inspection finding, a supply interruption, or a patient risk.

That requires a durable intelligence layer built to:

• Keep the source record attached to every conclusion
• Connect observations to cited rules, companies, facilities, investigators, offices, and products
• Make evidence searchable, comparable, and reusable across decisions
• Work through a dedicated platform or supply structured intelligence to internal chatbots, dashboards, agents, and workflows
• Support professional judgment without pretending to replace it

Workflow applications will remain useful. Systems of record are necessary. Raw regulatory data is necessary. Our work is to build the interpretation layer that connects them and make it available wherever regulated decisions are made.

That is the mission behind KeyPedia: a structured, referenceable view of regulatory evidence that people and systems can use before the inspection, before the audit, and before the decision.