The secret to less front desk burnout
Read blog
Content
Rising digital demands and platform constraints have healthcare leaders rethinking how – and who – builds AI workflows. The greatest opportunity to turbocharge innovation comes from AI Builders inside the health system.
Builders on Notable’s Builder Partners Program can design, test, and deploy AI workflows that address healthcare’s biggest challenges. That ecosystem includes healthcare leaders such as Jeff Pan, Co-Director of the Center for Health Innovation at UC San Diego Health, and Brian Schuetz, Executive Director of MIT Health.
These AI Builders leverage Notable’s Flow Builder interface, enabling them to scale AI strategically and at their own pace.
Two different health systems, one Builders Program
Even though MIT Health and UC San Diego Health differ in size and structure, they both faced a common challenge: traditional systems stalled AI progress.
MIT Health is a small, ambulatory-only organization operating in a highly constrained EHR environment. Because of limited digital options, it was nearly impossible to meaningfully customize AI.
UCSD Health is a large, clinically-integrated network supporting value-based care across multiple payers, and using a large EHR. Complex, inefficient technology platforms limited fast-paced innovation.
In both cases, Schuetz and Pan are using their roles as Builders to create greater strategic value. “We knew we needed to make broad, sweeping efficiency improvements in every business unit: facilities, clinical, administrative, revenue cycle,” says Pan.
This strategic drive leads many organizations to ask: should I build, buy, or partner to meet my AI needs?
Takeaway 1: Crafting a tailored approach is only the first step; success depends on both having a flexible platform and the right mindset.
The AI Builder mindset
Builders have the ability to automate routine workflows and design AI Agents. Pan looks for Builders with three traits: an engineering mindset, familiarity with enterprise-wide healthcare technology, and an affinity toward learning operations.
That mix of skills has led to fast progress, with more than a dozen Agents and flows built for clinical, administrative, and revenue cycle areas at UCSD Health in just over two months.
UCSD Health is piloting a three-tiered Builder model with architects, configurators, and physicians. This approach allows for speed while keeping clinical oversight; the goal is for each tier to solve different problems, not create bottlenecks.
Takeaway 2: Small, empowered teams can support a wide range of use cases, but keeping momentum means staying consistent in design, safety standards, and execution.
Addressing AI safety and governance
MIT Health
MIT Health’s governance model is centered on making data useful rather than just accessible. In their structure, AI roles are clear: one Builder designs how AI blocks work and connect, another manages forms and tasks, and clinicians apply their judgment to ask the right clinical questions.
This way, compliance and safety become a strategic advantage.
For example, the emergency department discharge summary is a lengthy document. MIT Health uses AI to position that information through a physician’s lens, highlighting clinical relevance. Then, to deliver it safely and effectively to nurses, patients, and other downstream workflows, including care gap interventions.
UCSD Health
UCSD Health embeds key stakeholders (labor risk, legal, privacy, and regulatory) throughout the AI Builder process, from intake and evaluation to testing and deployment.
For example, during pre-endoscopy outreach calls, Pan expected some pushback to an AI Agent handling medication reconciliation and guidance. Instead, risk leaders supported AI deployment, saying the workflow reduced risk when combined with human monitoring and other safeguards.
How to establish AI alliances
Pan achieved the strongest buy-in by quickly demonstrating proof of value. “For our first voice calls, we used production snippets, not real patient calls, to run bulk tests across different scenarios,” he says. “Flow Builder made testing easier, giving us confidence to deploy capabilities far faster than other platforms.”
Accounting for upstream and downstream flow dependencies is equally important. Taking time early to get the entire department on board, answer questions, and showcase capabilities has paid dividends. Engagement, understanding, and turnaround have all improved.
A highly restrictive EHR environment has traditionally forced MIT Health to say no to good ideas. Now, with a flexible AI approach, they can react in real time and say yes, even if experimental.
For example, the team built an AI workflow to address ambulatory referrals outside of the network. AI collects ED discharge information, creates local summaries, and even includes overnight phone service data.
The ROI in AI
Traditional ROI models may not capture new AI capabilities, so it’s important to give AI space to demonstrate value in areas such as clinical insights, risk management, and other non-monetizable benefits.
MIT Health designed an AI flow to review every encounter note for sensitive exams and to determine whether a chaperone was present. Its value isn’t easily quantified in terms of monetized ROI.
“We’re now not just seeing the value in AI, but deploying it,” Schuetz says. “Closing the time between AI flow building and deployment, when it’s safe to do so.”
Takeaway 3: Calculating AI ROI isn’t always straightforward. Focusing solely on a single return statistic could miss critical ideas that bring long-term value.
Excitement in the AI learning curve
The term "learning curve" implies that progress isn't a straight line. Schuetz compares AI flows to a box of Legos, meaning you can build anything.
“Once you understand how the pieces fit, everything clicks: intake flows, data parsing, and suddenly you’re asking, ‘What can’t I do?’”
The best AI platforms balance ease of use with the depth to handle complex, real-world healthcare workflows. Schuetz says one of the biggest challenges is the sheer number of data elements they want to use and the many ways to use them.
Pan finds flow-building highly engaging. “We jumped in with three Builders, including myself, a full-stack engineer, and a data scientist,” he says. On day one, the team was using a beta version of Flow AI.
Both excitement and challenges are part of the AI Builder experience, and there’s value in every part of it.
The future of AI starts now, inside health systems
AI should be guided by strategy, not just speed. “Automation can amplify problems, so we focus on high-impact or strategic areas rather than trying to solve everything,” says Pan.
Looking ahead, UCSD Health is exploring ways to handle documents, such as reconciling complex health plan contracts in the revenue cycle or managing faxes.
The most effective AI flows go beyond the initial scope to connect individuals, functions, and departments. “The real power comes from people at the intersection of experience and strategic vision to build what clinicians truly need,” says Schuetz.
“Part of our journey is to be small and mighty. We hope to stay nimble,” he adds.
