We analyzed 50 AI tool assessments across 3 health systems. Here's what we found. 👇
-
-
-
-
-
+3
Accelerating the deployment of beneficial AI in healthcare.
External link for Onboard AI
109 N 12th St
Brooklyn, New York 11249, US
Healthcare AI governance is moving from theory to operating reality. That’s why Onboard AI is joining Digital Medicine Society (DiMe)'s Operationalizing AI Governance in Healthcare initiative, a multi-stakeholder effort focused on making AI governance more practical, actionable and usable for healthcare organizations. Health systems are already deploying AI across clinical and operational workflows. But the infrastructure around those decisions hasn’t kept pace... Too often, governance work still lives across spreadsheets, inboxes, static documents and one-time committee reviews. That makes it harder for organizations to answer the questions that matter most: - What was reviewed? - What evidence was considered? - What conditions were attached? - Who owns the risk? - What needs to be monitored or reassessed over time? This initiative, led by DIME and co-hosted by Consumer Technology Association, Qualified Health and the FDA is focused on turning fragmented guidance into practical tools healthcare organizations can use to govern AI responsibly at scale. At Onboard AI, this is the work we’re built for: helping healthcare teams bring evidence, risk, accountability and oversight into one durable system of record. Responsible AI adoption doesn’t happen through frameworks alone. It happens when people, process and technology come together in a way that makes decisions clear, reviewable and accountable over time.
AI is coming out of every nook and cranny across the enterprise. We're here to help you manage that.
It's exciting to see all the AI emerging in healthcare. At Onboard AI, we're seeing three major vectors: 𝟏. 𝐄𝐱𝐭𝐞𝐫𝐧𝐚𝐥 𝐯𝐞𝐧𝐝𝐨𝐫𝐬: this is your AI startup, large health IT company's new shiny AI tool, etc. 𝟐. 𝐈𝐧𝐭𝐞𝐫𝐧𝐚𝐥𝐥𝐲 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐞𝐝 𝐀𝐈: usually rougher around the edges, typically home-grown models or highly-specialized algorithms. 𝟑. 𝐄𝐱𝐢𝐬𝐭𝐢𝐧𝐠 𝐞𝐧𝐭𝐞𝐫𝐩𝐫𝐢𝐬𝐞 𝐫𝐞𝐥𝐞𝐚𝐬𝐞𝐬: Big volume. Dozens per week coming from EMRs, CRMs, and other SaaS tools you've used for a decade plus. Governance isn't just covering some AI once. It's covering all vectors across all updates. Systems need to be in place to inventory new tech, as well as capture updates and versioning. Does this new tool represent a material change to care workflows, or is it an incremental retraining with better data? This problem is really about asking the question, "How much time do we need to spend assessing this tool?" This is essentially the "Intake and Triage" problem... If we're in a world where every piece of tech uses AI, then "AI governance" quickly becomes just "governance".
Contextualizing Risks > Liabilities > Mitigations is key to AI Governance; but in healthcare, it's more complex. Different stakeholders, users and intermediaries carry different points-of-view across clinical, admin, ops, legal, regulatory, revenue, etc. Mapping different domain risks to a shared taxonomy of liabilities and mitigations is key and not easily done manually, or in conversations. Onboard AI codifies this process for you. Read more (article in comments).
What makes a good AI Assessment? AI Committees are being built across every healthcare institution. They're tasked with creating an AI Policy, a Charter and importantly, a Risk Assessment Questionnaire. Many AI Committees build their AI Questionnaires from authoritative sources like the National Institute of Standards and Technology (NIST), Coalition for Health AI (CHAI) or the FDA. In this article, co-authored by Jared Augenstein, we explore what a good AI Assessment covers and offer a downloadable template to use or compare yours with. https://lnkd.in/gnWRtTrR