Clinical Ops in the Loop™ — The Method
The method behind everything we build.
Agents do the work. The right human owns the decision.
Salesforce's direction is unambiguous: agents operate the workflow, people supervise. In most industries, that is an efficiency story. In regulated research, it is a governance question — an agent drafting a consent record or flagging a deviation is touching regulated, patient-facing decisions at scale.
Clinical Ops in the Loop™ is our answer: a reference architecture for agentic AI in a regulated research environment. Before a single agent is activated, it determines which decisions an agent may execute, which it may only prepare, and which qualified human owns every consequential call.
The four design questions
Who owns the decision?
A named, qualified role — scientist, QA reviewer, investigator, coordinator — assigned by design, not discovered in an audit.
What may the agent do without asking?
Every action is classified: execute the routine, prepare the consequential, propose the judgment calls. The boundary is documented before go-live.
How is every action evidenced?
Every agent action, checkpoint, and override lands in an audit trail built for 21 CFR Part 11 — so when an inspector asks who decided, the answer is in the record.
How does the design evolve?
Agents improve, models change, studies amend. The method includes change control for the agents themselves — design authority, revalidation triggers, documented rationale.
Where It runs
Natively on Salesforce — Agentforce Life Sciences for the agents, Data 360 as the governed system of context, Shield for the evidence layer, MCP-based integration to the systems you already depend on.
Claude operates inside the Salesforce trust boundary as the reasoning engine, governed by the method.
The method carries Nurse Alicia's conviction — technology governed by someone who understands both sides. Read her story.
