Key points
- SQUIRE 2.0 gives 18 reporting items under four questions, why you started, what you did, what you found and what it means.2
- Give every result as a count over its denominator. In one randomised trial, a scribe was used in 33.5% of 24,696 eligible visits, and that share shapes every result after it.5
- CONSORT-AI adds 14 items for AI interventions, including how people and the AI interacted and an analysis of error cases.3
- A wai pilot runs for 90 days and ends with a report of what changed, with the method and denominator behind each number.7
Summary
Report the pilot as a quality improvement study. State the aim and baseline first, then give every result as a count over its denominator with an interval, with errors and limits in plain view. SQUIRE 2.0 is the base, and CONSORT-AI and DECIDE-AI add the items specific to AI.
Your partners settle the rollout in one meeting when every number in the pilot report carries its denominator and its interval. Health research already has the reporting standards, and a clinic can adopt them in an afternoon.
See it in your clinicThirty minutes on your own patients and doctors.
Request a demoSee how a 90-day wai pilot is set up around your team, and the report you receive at the end.
| Section | What to report | Example line |
|---|---|---|
| Aim and decision rule | The problem, the target and the rule agreed in advance | Adopt if the median reply time to enquiries falls by a third with no rise in complaints |
| Context | Sites, team, patient mix and dates | [n] sites, [n] front desk staff, [start and end dates] |
| Intervention | What the tool did, who used it and how staff were trained | Replies drafted by the tool and approved by front desk staff after a one-hour session |
| Denominators | Eligible items, items where the tool was used, exclusions | [n] enquiries eligible, [n] handled with drafts, [n] excluded as clinical |
| Baseline and comparison | Same measures over a matching period, or a concurrent group | Four baseline weeks at both sites, plus one site starting later as comparison |
| Results | Counts, rates and intervals for each measure | Median reply time and interquartile range before and after, bookings per 100 enquiries |
| Errors and unintended effects | Error cases reviewed, near misses and surprises | Drafts edited before approval, with the categories of edit |
| Limits | What the pilot cannot show | [n] weeks, [n] sites, no long-term retention data |
| Decision | The rule applied to the result | Rule met or not met, and the next step with the same measures |
Borrow a reporting standard
The EQUATOR Network lists SQUIRE 2.0 as the reporting guideline for quality improvement in health care, and a clinic pilot is quality improvement work.1
Its 18 items sit under four questions, why you started, what you did, what you found and what it means.2 It asks for the local problem and context, the intervention in enough detail to repeat it, and how each measure was taken.2
Denominators before percentages
Write each result as a count over a stated base, such as bookings over eligible enquiries, with both counts shown.
In a randomised trial of 238 outpatient physicians across 14 specialties, one AI scribe was used in 33.5% of 24,696 visits and the other in 29.5% of 23,653 visits.5 Report results per eligible visit and per visit where the tool was used.
SQUIRE 2.0 also asks for missing data, so state how many weeks or clinicians were dropped and why.2
Baselines, comparison and confidence
A concurrent comparison, such as clinicians not yet using the tool, gives the clearest answer. The minimum is a baseline measured the same way over a matching period, and SQUIRE 2.0 asks you to state how you judged whether the change caused the outcome.2
In the same trial, one scribe cut time in notes by 9.5% against usual care, with a 95% confidence interval of 1.8% to 17.2%. The other showed a 1.7% fall, with an interval from a 9.4% fall to a 5.9% rise.5 Report the first as a finding and the second as no measurable change.
For a small clinic, an interval around a proportion, or a median with its range, is enough.
What AI pilots add
CONSORT-AI adds 14 items for interventions with an AI component.3 A clinic can apply these directly, the training needed to use the tool, the setting, how inputs and outputs were handled, how people and the AI interacted, and an analysis of error cases.3
DECIDE-AI, for early clinical evaluation of AI decision support, has 17 AI-specific items and ten generic ones, agreed through a Delphi process that began with 123 experts.4 Its focus on human factors and safety at small scale fits a first clinic pilot.
In an emergency department pilot with five early adopters and 710 visits, AI-drafted notes went with more time in the notes section per adult patient, 4.3 against 1.8 minutes, and a much larger physician share of each note.6 Reported plainly, a finding like that lets the next site plan for it. To score note quality itself, see how to evaluate AI-drafted documentation.
Claims the data supports
A pilot supports claims about the people, weeks and workflows it measured. Report measured time for the clinicians who used the tool, over the weeks you ran, and leave annual projections out.
Another organisation's study is context for your result. Before and after figures without a comparison are associations, and the report says so. Set the decision rules in advance, as in clinic AI pilot success criteria, so the report tests them.
How wai reports a pilot
A wai pilot for the front desk and back office runs for 90 days around your team and workflows, and ends with what changed in revenue and hours saved.7 wai reports each result to the clinic with the method and denominator behind it, and every draft and approval is logged in an event log the clinic can export.8
Each clinic owns its report and decides what to share.
Where to start
Before the pilot, write the aim and the decision rule in two sentences, record four weeks of baseline with the definitions you will use later, and choose one comparison, a concurrent group if you can.
During the pilot, log usage, exclusions and errors weekly. At the end, fill in the template below, have a clinician outside the pilot read it, and put the denominators on page one.
Bring your baseline. We set up the pilot around it and report back against the same measures.
Frequently asked questions
Which reporting guideline should a clinic pilot follow?
How many patients does a pilot need?
Enough that the interval around the main result is narrow enough to apply the decision rule. Show the denominator and the interval so readers can judge.
Can we compare the pilot with last year's figures?
As a baseline, yes, if the definitions match. Seasonal and staffing changes affect results, so a concurrent comparison group is stronger.2
Should we publish our pilot results?
That is the clinic's choice. If you do, use the template above so readers see the denominators, intervals and limits.
Who receives the results of a wai pilot?
The clinic that ran it, with the methods and denominators behind each number. The clinic decides what to share.7
What should we do with errors found during the pilot?
Classify and count them by type. CONSORT-AI asks for an analysis of error cases, and that section tells the next site the most.3
Sources
- 1EQUATOR NetworkSQUIRE 2.0, reporting guideline recordRead 14 October 2026Read source
- 2BMJ Quality and Safety, Ogrinc and colleagues, 2016SQUIRE 2.0 (Standards for QUality Improvement Reporting Excellence), revised publication guidelines from a detailed consensus processRead 14 October 2026Read source
- 3Nature Medicine, Liu and colleagues, 2020Reporting guidelines for clinical trial reports for interventions involving artificial intelligence, the CONSORT-AI extensionRead 14 October 2026Read source
- 4Nature Medicine, Vasey and colleagues, 2022Reporting guideline for the early-stage clinical evaluation of decision support systems driven by artificial intelligence, DECIDE-AIRead 14 October 2026Read source
- 5NEJM AI, Lukac and colleagues, 2025Ambient AI Scribes in Clinical Practice, A Randomized TrialRead 14 October 2026Read source
- 6Annals of Emergency Medicine, Morey and colleagues, 2026Ambient Artificial Intelligence Versus Human Scribes in the Emergency DepartmentRead 14 October 2026Read source
- 7wai, 2026wai pricing, how the pilot worksRead 14 October 2026Read source
- 8wai, 2026Security at waiRead 14 October 2026Read source
A pilot report with every denominator on page one, and a rollout decided in a single meeting.
Fifteen minutes with a clinician who uses wai daily, then a 90-day pilot measured against your own baseline. We reply within one working day.
Partner with us
See it on your own patients, then build it with us
Thirty minutes on your own patients, doctors and rules, in your market. We reply within one working day, and the clinics that partner with us early shape what we build next.
Read how results explained works in a clinic