What clinicians taught us about ambient order staging

Last year, we introduced ambient order staging, giving clinicians a more efficient way to generate and stage medication orders simply by capturing them during a patient encounter. Suki has seen 72% growth in clinicians adopting orders month over month, and adoption continues to accelerate.

As adoption grew, so did our understanding of the nuances involved in generating structured orders. Unlike transcription, structured order generation requires interpreting the many ways clinicians verbalize medications in real-world practice. Insights from user interviews, surveys, and usage patterns informed the next generation of improvements, further reducing manual work for prescribing clinicians.

Challenge #1: Clinicians don’t verbalize every detail of the order

In theory, order generation sounds straightforward: listen to the conversation and create the order. In practice, clinicians don't verbalize every detail of an order out loud.

A clinician might say "let’s start tirzepatide" without stating the dose, frequency, or duration out loud. Those details may be entered manually later or selected from familiar order sets.

Rather than inferring missing details, we made a deliberate choice: only generate information based on information available from the transcript and encounter. Clinical safety is more important than producing a perfectly complete order.

What we changed

We continuously refined our models using real-world encounter data to better recognize clinical intent and accurately capture what clinicians actually say.

To reduce the need to verbalize every individual order, we developed order sets that allow clinicians to stage groups of commonly used orders and make adjustments as needed, preserving efficiency without sacrificing control.

What we learned

Clinicians need confidence that every element in a staged order comes directly from the encounter, not from AI assumptions. Building that trust meant finding ways to make orders more complete without sacrificing clinical safety, while also ensuring clinicians remain in control of every order they review and sign.

Challenge #2: Orders happen during the visit, not after

While documentation often is completed after the encounter, orders aren’t. Clinicians frequently review treatment plans with patients before the visit ends, making it impractical to wait for the note to be finalized before generating orders.

What we changed

We redesigned the workflow so orders can be staged immediately after the ambient conversation is over. Clinicians can review and send orders to the EHR immediately and finalize the rest of their note later.

Behind the scenes, this meant building a dedicated order-generation pipeline separate from note generation, ensuring orders aren’t delayed by the note generation process.

What we learned

Ambient workflows need to support what clinicians and patients need at the point of care, not just documentation. Orders must be available when clinicians and patients are together in the room, not minutes or hours later.

Challenge #3: Every EHR works differently

Supporting ambient orders across multiple EHRs introduced another layer of complexity. Each system has its own workflows, capabilities, and requirements for accepting structured data. Ensuring the data we generate is correctly formatted and consistently sent to each EHR was a major challenge.

What we changed

We invested heavily in ensuring orders are formatted correctly and accurately integrated into the appropriate fields in the EHR. This ensured clinicians didn’t have to copy-paste orders and instead allowed them to easily review and finalize so patients would have their orders ready by the time the visit was over.

What we learned

Depth and breadth of EHR integration are just as important as accurately capturing and structuring clinical data..We focused on optimizing the experience within existing EHR workflows, giving clinicians a faster, more intuitive way to review and sign off on orders.

Measuring success through usage

The most meaningful measure of success wasn’t the number of features we built; it was seeing how our users engaged with our product and how quickly adoption spread among our customers.

After redesigning the experience, we saw sustained growth in both adoption and weekly active usage. Clinicians didn’t simply try ambient orders once; they incorporated them into their daily routines. With 72% growth in orders adoption month over month, and 300+ organizations generating ambient orders, these trends tell us we’re solving a real workflow challenge and helping clinicians spend less time on manual order entry.

Looking ahead

This is just the beginning. As we continue to expand ambient order staging across additional order types and EHRs, we’ll continue building around the same product principles that guided us this year: clinical safety, workflows grounded in real-world practice, and keeping clinicians in control of every order they review and sign.

For more information, reach out to us here.