Ambience Healthcare on Epic: Toolbox, Haiku and What Your Team Owns

The Haiku build uses Epic's ambient framework. The clinician opens the encounter in Haiku, starts a recording, and Ambience returns a draft note structured to the organisation's templates. Patient and encounter context, including prior notes and diagnostics, is read from Epic so that the draft can reference it. The vendor calls this contextual intelligence; in integration terms it is Epic on FHIR reads plus the ambient module's write path, combined with the Showroom and Toolbox programme access on the vendor side.

Before the Toolbox build, and still at some sites, Ambience ran as an integrated application launched from Hyperspace. Ask which build you are being sold, because the two have different Epic version requirements and different analyst work. The general shape of these paths is described on the Epic page and the questions to ask are in our integration questions guide.

Not published: the exact interface inventory, the minimum Hyperspace and Haiku versions for the Toolbox build, and the site level implementation plan. All three come out during scoping; ask for them in writing before the contract, not after.

Three things, as listed on the vendor's site in September 2026. First, note drafting mapped to your templates across outpatient, ED and inpatient note types. Second, ICD-10 coding recommendations, with documentation generated to support the selected codes. Third, CDI prompts intended to close documentation gaps before the note is signed.

The second and third are where your compliance officer needs to be involved early. Documentation generated to support a code is exactly the pattern payers and the Office of Inspector General look for when they audit upcoding, and a system that does it at scale needs sampling, an audit trail and a policy for who is accountable for the code. That is not a reason to avoid the feature; hospitals lose real revenue to under-documented encounters, which our charge capture leakage calculator estimates. It is a reason to switch it on deliberately with a governance decision behind it. Our governance committee playbook sets out how.

For the documentation loop itself, the measures are the same as for any AI medical scribe: note turnaround, after hours charting, edit distance and clinician retention on the tool at ninety days.

• The security review and business associate agreement, covering audio retention, subprocessors and model training. The HIPAA and AI page lists what to insist on.
• Epic ambient security points per user class and any allowlist for shared encounters.
• Template mapping for each note type in scope, and the decision on which SmartSections receive generated text.
• Coding workflow design: who sees the ICD-10 suggestions, whether they flow to a coder queue or straight to the encounter, and how they are sampled.
• Mobile device management for Haiku, including microphone permissions on personal phones.
• Change control and named validation clinicians for each setting, since ED and inpatient workflows differ from clinic.

The usual surprise is the coding workflow item. Vendors sell it as automatic; your revenue cycle team will want it staged. Plan that conversation before the kick-off, not during it.

Pricing is on request only and Ambience sells to health systems, so smaller Epic groups on a hosted instance may not meet the minimum commitment. Our Ambience pricing page sets out what is known.

Second, the Toolbox build is recent. August 2025 is the announcement date; the number of sites live on the Haiku build is not published. Ask how many customers run the Toolbox build in production today, and on which Epic release, and get a reference among them.

Third, the coding emphasis cuts both ways. It is the differentiator against Abridge, Dragon Copilot, Nabla and Suki, all of which are also embedded in Epic, but it increases the compliance work on your side and it is the feature most likely to be scrutinised by payers. If you want documentation only, say so and negotiate accordingly. The AI medical scribe comparison lays out the field.

• Which build are we buying, the Toolbox Haiku build or the Hyperspace launched application, and what are the Epic version floors for each?
• How many sites are live on the Toolbox build, and can we speak to one on our release without the vendor present?
• Can coding and CDI features be disabled by user class or by setting, and what audit trail exists for accepted and rejected suggestions?
• Where is audio processed and stored, for how long, and is any of it used to train shared models? Put the answer in the BAA.
• What is the per clinician price at our volume, the minimum term, and the exit terms if adoption stalls?
• How many Epic analyst hours did a comparable site spend, and on what?

Our vendor HIPAA question list extends the compliance items.

Ambience is credible on Epic and its coding features are the reason a revenue cycle leader will push for it. An independent review keeps the clinical, compliance and financial cases separate long enough to test each one: does the note quality hold on your specialties, does the coding suggestion rate survive a compliance sample, and does the price beat the embedded alternatives on your Epic build. We take no vendor commissions and we run the comparison on your encounters rather than the vendor's.

Our vendor selection service runs that comparison; our governance and compliance service designs the coding oversight if you proceed. Book a call and bring your Epic release and your top ten specialties by encounter volume.