Choose a task with your team
We follow one repeated task, establish the current effort and exceptions, and agree on a deliverable the team can use. Each pilot has a named owner and a result we can measure.
Agent workflow pilot
Give your team less work to repeat.
An agent may be able to gather context, compare records, or prepare a draft. That is useful only if your team gains enough to justify the cost of running it, reviewing it, and correcting it.
I work with you on one focused pilot. We choose a task, build around sources your team trusts, and test the result with the people who will use it. Their experience and the measured results tell us whether to expand, revise, or stop.
At a glance
- Who it’s for
- Behavioral health and healthcare founders and operators first. Other founders and operating teams are welcome.
- Timeline
- One task, measured in use. We set the schedule when we agree on scope.
- What you leave with
- A pilot your team has used on one real task, an operating guide, and measured output quality, time saved, review effort, and cost to decide whether to expand, revise, or stop.
- Fee
- Fixed, agreed before kickoff.
What makes a pilot useful
Count the effort of using an agent’s output.
At One to One Health, I developed the initial AI infrastructure for provider note creation. AI prepared a draft for provider review, and charting time fell from 22 to 17.5 minutes across all encounters while clinician and patient satisfaction were maintained or improved.
That experience informs what I look for in a pilot: preparation that software can take on, an output someone can use, and a clear way to judge whether the change helped. Review effort matters. So do the occasions when the information is missing or the system gets something wrong.
Read the One to One Health storyPlanning a pilot
One task, followed through into use.
Build for the ordinary day and the awkward one
We connect approved sources and agree on what the system may do. We include human review and a way to recover when inputs are missing, a step fails, or a task is interrupted.
Use it and decide what comes next
We run the pilot with its owner and measure output quality, time saved, review effort, and cost. Your team gets an operating guide and evidence for a decision to expand, revise, or stop.
An example
From a meeting note
to work the team can review.
Illustrative walkthrough with sample content, not a customer system.
Sample workspaceMeeting → product decision
Approved input
What did the team actually say?
“We retype the same supplier details into two systems. Before we build anything, we need to know which record is right.”
- Source
- Sample discovery note
- Open question
- Which system owns the supplier record?
An agent uses the supplied source and preserves what is still unknown.
Prepared work
A useful next step, with its reasons.
Map the two entry points, confirm the source of truth, and measure the time spent re-entering details.
- Proposed deliverable
- A workflow map and one integration recommendation
- Why this step
- We need clear ownership before automating a write.
Preparation does not approve a build or grant access to a system.
Your judgment
Do we mean the same thing by the deliverable?
A designated owner reviews the map, corrects missing context, and agrees on one read-only pilot that compares records.
- Allowed
- Read approved samples and show differences
- Needs a later decision
- Writing changes into either system
Each handoff includes the accepted scope.
Underway
Build against the agreed scope.
Implementation follows the reviewed handoff. Checks cover conflicting records, missing fields, and repeated inputs.
- Review package
- Working comparison, source links, and test results
- Release
- Separate owner decision after review
A passing software test is one part of the delivery evidence.
What moved
Did the change help the team?
A designated owner reviews the output in the actual workflow. We record review time, errors, cost, and tasks no longer needed.
- Business result
- To be measured in the pilot
- Next choice
- Continue, revise, or stop based on the evidence
Recorded results and unresolved items remain available for the next task.
What I’m building
Where I’m putting these ideas to work.
Claims Native
Help a practice keep up with billing.
At TPN, a question about invoicing payers led us to build an internal claims filing and revenue capture platform. That experience is why I’m building Claims Native: claim preparation, checks, exceptions, and payment follow-up for practices, with people responsible for consequential decisions.
Early access
Visit Claims NativeInternal agent workspaces
Keep decisions and sources close at hand.
I’m building workspaces that bring sources, decisions, and drafts together so I can pick up a task without reconstructing the conversation. They are part of my own operating work.
Internal use
Scope and follow-through
Know what the pilot is there to tell us.
Before we start, we agree on the sources, allowed actions, human decisions, deliverable, measures, price, and delivery responsibility. If the workflow is still unclear, an assessment gives us a way to work that out first.
At the end, we review the pilot in the actual workflow and decide whether to expand, hand over, or stop. Ongoing operation needs its own agreement on support, recovery, access, and cost. Pilot results give us evidence for that decision.
Start with an assessmentWork with me
Which task keeps taking your team’s time?
Tell me what people do today and what they would be able to do if that work became easier.