Healthcare AI stops summarising and starts running the work

The assistant that cites your hospital, not the internet

Abridge announced support for local pathways as a new capability within the Abridge Assistant, and the Abridge Assistant is currently used by more than 300 enterprise health systems. The health systems using Abridge Assistant represent over 250 million patients. The scale is the context for the feature: a change to the assistant now reaches a large share of American hospital care. m:…5183

Local pathways are designed to translate medical guidelines into a shared, consistent approach to treating patients. When an institutional pathway is identified, the Assistant surfaces relevant content with a link to the original source. Clinicians often struggle to discover institutional guidance in the moment of care due to fragmented storage like PDFs, websites and intranets. The feature is aimed at that exact gap: the guidance exists, but the clinician cannot find it while the patient is in front of them, so the assistant brings it to the conversation with its source attached. m:…5183

The Abridge Assistant is connected to the patient's chart and the clinician's existing workflow, and it supports clinicians by summarizing changes across prior visits, rounds, labs and vitals. Health systems can enable the Assistant for every clinician through their existing Abridge deployment. Shiv Rao is the CEO and Co-Founder of Abridge and a practicing cardiologist, and Deepti Pandita is the Chief Medical Informatics and AI Officer at University of California Health. The assistant runs on enterprise-grade infrastructure that has passed security review and data governance requirements, which is the table stake for any tool that touches the chart. m:…5183

The rules nobody wrote down

Luminai introduced the Luminai Topology platform to centralize fragmented institutional knowledge for AI agents. Luminai Topology uncovers undocumented operational rules by learning from decision traces in operational data. Every hospital runs on rules that live in people's heads: which form goes where, which exception applies, who approves what. Those rules are the reason agents fail in healthcare, because the agent cannot read what was never written. m:…6117

Luminai Topology encodes documented logic and expert knowledge into structured rules for AI application, and provides an audit trail by linking every decision back to its source, such as documents or expert confirmation. It tracks rule changes across time, allowing agents to apply policies based on effective dates. The audit trail is the feature that matters most in a regulated setting: when an agent makes a decision, someone needs to see which rule it used and where that rule came from. m:…6117

Luminai Topology connects upstream decisions to downstream outcome data to suggest rule optimizations, and connects learning across different use cases to improve decision-making across the organization. It is model-agnostic, allowing automations to improve as frontier AI models evolve, and it ensures data privacy by not sharing data across different customers. Model-agnostic matters because the model will change next year; the rules and the audit trail are the durable asset. m:…6117

Capacity, not minutes saved

Healthcare systems currently rely on manual labor to move work forward despite digitizing information, Notable argues. A 'System of Work' coordinates people, AI agents and existing applications to complete operational workflows, and an AI agent that only reads documents saves time, but a coordinated workflow creates capacity. The distinction is the argument of the whole piece: summarising a document helps one clinician, while a workflow that moves a referral or a discharge forward helps the system. m:…5974

AI investments in healthcare should aim to shorten queues, resolve handoffs and return time to caregivers, according to Notable. Operators should measure referrals by time to appointment, referral conversion and unresolved backlog, and denials should be measured by resolution time, collections and cost per resolved case. Discharge coordination should be measured by time to resolve administrative barriers and staff coordination burden. The measures are operational, not technical: nobody cares how accurate the summary was if the patient still waits. m:…5974

Notable says its Health product delivers more than 3.2 million hours saved annually across its partners. NKC Health used Flow Compass to identify a configuration detail for headache questionnaires before go-live, and Flow Compass enables workflow planning, building, performance analysis and troubleshooting. The Notable Network allows healthcare builders to share expertise and reusable approaches. Hours saved is a vendor number, but the direction of the argument is worth noting: the vendors that win in healthcare will be the ones that prove capacity, not the ones that demo the cleverest summary. m:…5974