AI Vendor Evaluation

No. Kent Applied sells no software and takes no referral fees, so the recommendation can be any vendor, none of them, or a build.

Up to five. Three or fewer is $7,500, and four or five is $9,500. A longer list is cut to five in week one, against the requirements matrix.

No. Capability tests run on de-identified or synthetic cases you approve. Real patient data never goes to the vendors. We check how each would handle it by reviewing its documents: its BAA, its subprocessors (the other companies it sends data to), how long it keeps data, and whether your data trains its models.

With scripted visits. We write short visit scripts with fictional patients, your clinicians and staff act them out, and every scribe on your shortlist drafts a note from the same visits. The scripts include the situations where published studies found scribes make mistakes. In an audit of three commercial scribes, posted as a preprint that has not been peer reviewed, about one note in three had a verified error, most often in allergy and medication details. In a randomized trial at UCLA, the physicians who reported inaccuracies most often described details left out of the note. In a study of acted-out visits with an interpreter, the two scribes tested carried 55% and 60% of the interpreter's mistakes into the note. In a preprint covering 54,160 visits at one safety-net health system, notes from English visits were signed with at least 80% of the scribe's words left unchanged in 57.6% of visits, against 48.7% to 51.1% for visits in other languages. A 2025 report from the Peterson Health Technology Institute says some scribes have trouble summarizing a visit with a patient and several caregivers, because they can't always tell the voices apart. So the scripts cover medication and allergy changes, an interpreter visit, a visit in another language, a family in the room and visits from your own specialties. Your clinicians score each note without knowing which scribe wrote it: what it got wrong, what it left out, whether the medications and allergies are right, whether it fits your specialty's format, and how long it took to fix before signing.

Because a recording of a real visit is patient data. HIPAA's de-identification standard lists voice prints among the identifiers that must be removed, so we treat visit audio as patient data. Scripted visits give every scribe the same test with no real patient in it, and nothing about a real patient goes to the vendors or to Kent Applied. Your clinicians sign in to the scribes during the test; we never do.

Then the memo says so. No purchase, a build, or waiting for software you already own to add the feature are all valid results.

The memo shows the cost of building next to the three-year cost of buying. Kent Applied can do the build as a fixed-price Custom AI Tool, or you can take the memo to another team.

Yes. The memo lists the contract terms to ask for, and Tyler can join the negotiation calls. That's priced as a fixed add-on in your proposal.