Florence Healthcare

my love language: fewer logins

Checkbox governance doesn't work for AI. Like driving a car, governance has to be built in, frequent, responsive, and always "on." A policy that says "drive safely" isn't the same as a dashboard that flags problems in real time, or the driver that hits the brakes because when being cut off. Checkbox governance: a policy drafted, signed off, filed away, revisited only after an audit or an incident equates to the annual inspection sticker, not the dashboard. It answers "do we have a policy," not "do we know what our AI is doing right now, and are we prepared for what’s next?” "Always on" operational infrastructure: a living inventory of every model touching trial data, review thresholds before deployment, monitoring built into existing workflows, and clear ownership of who approves and who watches for drift. It's the dashboard running the whole drive, not a sticker checked once a year. This isn't just our framework. NIST's AI Risk Management Framework runs on the same logic. Four functions, Govern, Map, Measure, Manage, run continuously across a system's life, the way a driver reads the road for the whole trip, not just at the start. The EU AI Act goes further for high-risk systems. Article 9 requires an ongoing risk management process. Article 17 requires a documented quality management system across that same lifecycle. Florence isn't classified high-risk under the Act today, and while those articles aren't a binding requirement for us, we strive to hold ourselves to the same standard anyway. Why does it matter? It comes down to accountability. The difference shows up the moment something goes wrong. Checkbox governance has a policy that says "we take this seriously." Operational infrastructure has a paper trail showing what was reviewed, when, and by whom, like a black box recording the drive instead of a sticker confirming a check happened once. At Florence, we build with the dashboard, not the sticker. The rest of this series walks through what that looks like in practice.

"we already use florence"

Things that happen at every site that don't show up on a resume: - the EDC that logs you out mid-entry - the training system that makes you redo the same module for every study - the imaging system that opens in its own separate universe - the lab system nobody can find the login for Sponsors pick the systems. Vendors change study to study. Sites don't control the stack, they just adapt to it, over and over, one study at a time. Innovation built hundreds of tools to solve narrow workflow problems. Who's building for the site that has to run all of them at once?

Kinda chic to know exactly where every site, every signature, and every document stands.

Leanne Woehlke, Tala Fakhouri, Michelle Rowe, and Leslie Pimentel Byatt, MSML, MBA, PMP, CCRC, FACRP are coming to Research Revolution and their sessions will dig into where AI actually earns its place in clinical research: recruitment that doesn't take 9 months for 3 patients, a regulatory blueprint straight from Washington, a study start-up process without the bottlenecks, and governance that keeps humans in the loop when the bots start running the show. Swipe to explore sessions and get learn more https://hubs.li/Q04sj7n10 #RR26 #ResearchRevolution

when you send all the files and they say they didn't receive the email

Recently we wrote about how Florence uses AI across clinical trial operations. The most frequently asked question from sponsors and sites is “how do you govern it?”. In June 2026, the EU finalized its Digital Omnibus amendments to the AI Act. The high-risk compliance deadline shifts from August 2026 to December 2027 for stand-alone systems, and to August 2028 for AI embedded in regulated products, a category that covers many clinical trial tools reviewed under the MDR or IVDR pathway. However, the Omnibus did not touch Article 50. Transparency obligations: disclosing when someone is interacting with AI, marking AI-generated content and flagging deepfakes; all still apply from August 2, 2026. Article 50 is broad-reaching and applies to ANY AI system in the 4 risk categories. We all know that checkbox governance is self-limiting, but it is even more ineffective when working against two different clocks.  More runway on one obligation and a live deadline on another is when checkbox governance shows its limits. A single "are we compliant" checklist can't address two different clocks. Operational infrastructure tracks each requirement on its own timeline. Over the next few weeks, we're breaking down what that infrastructure looks like day to day starting with the distinction that matters most: governance as a checkbox versus governance as operational infrastructure. Read the full blog, Building a Clinical Research AI Governance Framework here: https://hubs.li/Q04s10Z40 Note: High-risk deadline shift based on the Council and Parliament's Digital Omnibus agreement (provisional May 7, 2026; Council final approval June 29, 2026). Article 50 transparency deadline (August 2, 2026) and scope per the EU AI Act text and the Commission's draft Guidelines, summarized at https://hubs.li/Q04s0QDw0.

when you have an urgent request and you have to figure out which sticky note login is the right one