AI Scribes Save Doctors 13 Minutes. That's Not the Bar We're Measuring Against.

The study everyone in health tech is talking about

A landmark study published in JAMA this April examined what happens when AI scribes are deployed at scale across five major academic medical centers. The findings were notable — and, for some, sobering: clinicians saved about 13 minutes of EHR time per day and documented roughly 16 minutes faster. Revenue ticked up modestly. Work outside of work hours barely budged.

The authors were careful about what they were measuring. But the broader conversation has been less careful — and some have taken the findings as a verdict on AI in clinical workflows generally.

We'd like to offer a different read.

Scribes and care intelligence are not the same thing

An AI scribe listens to a physician visit and generates a clinical note. It's a documentation shortcut layered onto a workflow that was already partially automated, already supported by structured EHR templates, and already operating on 15-to-20-minute interactions.

That's not what we built — and it's not the population the JAMA study was measuring.

The clinicians we serve are care managers and nurses at health plans and care management organizations. Their work looks fundamentally different:

• Calls run 30 to 60 minutes, often longer for complex members
• Post-call documentation is long-form, narrative, and highly structured — assessments, care plans, intervention logs, barrier identification
• Documentation can account for 20–30% of total call time, a far higher ratio than a physician's note relative to a clinical visit
• The balance between time spent with the member and time spent documenting the member is deeply out of whack — and care managers feel it every day

When AI deeply understands a care conversation — not just transcribes it, but comprehends the member's barriers, clinical context, emotional state, and the care manager's interventions — and then generates documentation tailored precisely to that organization's workflow and assessment structure, the math changes entirely.

What we actually see

In a formal time study conducted by a large multi-state health plan following their Laguna pilot, post-call documentation time fell by an average of 66% across all care teams. Some teams saw reductions of nearly 80%.

For context: the JAMA study found a ~20–25% reduction in documentation time for physicians. The difference isn't a product quality gap — it's a workflow gap. The ceiling for savings is simply much higher when documentation represents a larger share of total work time.

The productivity gains translated directly into capacity: hundreds of additional member enrollments per care manager per year, across teams. Time saved on documentation became time spent reaching more members. That's not a side effect. That's the point.

And adoption? Among credentialed users with active seats, engagement was essentially universal — ~100% of coaches opened the platform and completed the core documentation workflow. The JAMA study found that only about a third of adopters used their AI tool in the majority of their visits. That gap matters enormously — and it doesn't happen by accident.

By the numbers: a direct comparison

Key metrics: AI scribes for physicians (JAMA, 2026) vs. Laguna for care managers

Documentation time reduction by care team (customer time study)

Note: the behavioral health productivity gain was offset by other workflow factors during the measurement period. Laguna-generated summaries for behavioral health, wellness, and social work teams were still being customized and were expected to increase productivity further.

Why adoption actually happens

The JAMA authors flag low adoption intensity as a key reason the population-level results were modest. They're right. And they point to training and support as the lever.

We've learned the same thing. Here's what actually drives the numbers we see:

The documentation fits. Not a generic clinical note — documentation that matches each organization's specific workflow, assessment frameworks, and reporting requirements. When a care manager copies the output directly into their system without editing, that's not luck. It's calibration.

Onboarding is taken seriously. Every rollout includes structured training, champion development, and ongoing support. When coaches understand why the tool works and how to use it effectively, they use it on every call.

The product earns its place. This is the one that's hardest to measure but easiest to hear. Care managers tell us things like: "I don't think I could live without it now." And: "Whether you've been doing this job for years or are new, this is going to help us be quicker and make us better nurses." Tools with that kind of relationship don't struggle with adoption.

What the JAMA paper actually tells us

The Rotenstein et al. study is rigorous and important. Its finding — that broadly deployed AI scribes produce modest, real-world average time savings in physician workflows — is credible and worth taking seriously. It's an honest picture of what happens when you roll out a documentation tool at scale in a complex clinical environment with variable adoption.

But it's a study of a specific tool, in a specific setting, for a specific clinical role. Its conclusions don't generalize to care management. They don't generalize to platforms that do more than transcribe. And they don't generalize to deployments where the documentation is purpose-built, the training is serious, and the product has earned genuine affection from the people using it.

The bar we're measuring against isn't 13 minutes.

It's whether care managers can spend more time with members, reach more people, document confidently, grow professionally, and leave their shift feeling like the job is still worth doing.

That's the problem we built for. And that's the problem we're solving.