The Deployment

Patricia Stone tells it this way: she had a board meeting coming and could not remember why a specific line in the financial model was built the way it was. The model was correct. She knew it was correct because Raymond had told her it was, twice, in sessions seven and nine. What she did not have was the reasoning, the chain of logic that connected the cost center structure to the Medicaid reimbursement pattern that Raymond had explained and she had understood in the moment and could not reconstruct three months later.

She asked the AI.

The AI returned Raymond’s explanation from session seven. Not a summary. The actual diagnostic reasoning he had walked through: the specific question he had asked first, the alternative structure he had considered and rejected and why, the downstream implication for the board reporting format. Raymond was in Cincinnati. He was not on the phone. His reasoning was still in the room.

Patricia reads this and says: I did not know this was what we were building.

Raymond Okafor is 66 and spent twenty years directing hospital finance at a regional health system in Cincinnati, specifically Medicaid reimbursement strategy and rural health center financial modeling. He retired knowing things that are not in textbooks. How Medicaid reimbursement actually flows through a rural community health center. Where the cost centers are hidden in an FQHC’s financial structure. What a payer contract should and should not concede when the institution is small and the payer is large. The specific patterns that distinguish a financially unstable community health center from one that is one good year of management away from stability.

He arrived at the health center in rural West Virginia for a twelve-week deployment. His Native, Julia Brennan, 25, had just finished her MPH at West Virginia University. She had not been inside a rural health center’s financial structure before. She had a facility with data modeling tools that Raymond did not have and a speed of analytical execution that would have taken Raymond weeks to replicate.

Patricia Stone runs the health center. She has run it for eight years. She knows her patients, her community, and her staff in the way that a director who has been in one place for eight years knows them. She does not know what Raymond knows about Medicaid reimbursement modeling or what Julia knows about visualizing that modeling for a board that does not work with financial data every day.

The deployment was twelve weeks. Each account of it is different. All three are accurate.

Raymond’s account begins with a specific recognition.

The health center’s cost structure had a pattern he had seen in four previous institutions. The care coordination costs were rising faster than patient volume, which should not be possible if the care coordination model was working as designed. The standard interpretation, which the center’s staff had already arrived at, was a coordination workflow problem: too many handoffs, not enough integration, the familiar diagnosis that produces a workflow redesign. Raymond had made that diagnosis twice in his career and been wrong both times. The real diagnosis was a care management problem: a cluster of high-utilization patients whose cases were being managed by referral rather than by a single coordinating provider. The workflow was not the problem. The patient stratification was.

He made this diagnosis in session two, looking at the center’s data with Julia. He did not tell Patricia immediately. He spent two sessions building the evidence that would make the diagnosis comprehensible to a director who had not seen this pattern before. That is what the pattern recognition looks like in practice: not a quick declaration but a structured presentation of why what it looks like is not what it is.

Julia built the data model that made his diagnosis visible in a form a board of directors could evaluate. She coded the patient stratification by utilization quartile, linked it to the cost data, and produced a visualization that showed the gap between what the standard care coordination model cost and what a targeted care management model for the top-utilization quartile would cost. The board saw the numbers. Raymond saw the pattern. Julia made the board see what Raymond saw.

Julia’s account begins with a presentation she did not plan to give.

In week eight, Raymond told her the board needed to see the care coordination analysis before they could authorize the scope expansion to include the workflow redesign. The board meeting was eleven days away. The analysis was complete. The presentation did not exist. Julia built it in four days, starting from the financial model Raymond had structured and working backward into a narrative that a board of directors who thought in operational terms rather than financial terms could follow.

Raymond reviewed it the night before the board meeting over a video call from Cincinnati. He made two changes: he moved the conclusion to the front of the presentation, which Julia had placed at the end because that was how her MPH program had taught her to present research, and he rewrote one paragraph of financial explanation that he said would lose two board members in the third sentence. She made the changes. She understood why he made them when she watched the board read the presentation the next day.

The board authorized the scope expansion. The workflow redesign became the most important thing they did in the deployment. The financial model was the foundation. The workflow redesign was what the health center could actually use in the next month, when a specific high-utilization patient whose case had been managed by five different providers in eighteen months was assigned a single coordinating provider and the care coordination cost for that patient dropped by 40 percent in ninety days.

Julia did not design the workflow from scratch. She built the structure Raymond outlined, using the health center’s existing staffing model as the constraint. The combination required both of them.

Patricia’s account is the one that includes the scope of what she did not know she needed.

She knew she needed a financial model she could defend to her board. She did not know until the deployment started that what she actually needed was a framework for understanding her own institution’s cost structure, not just a model someone else had built. She knew she needed a care coordination review. She did not know until session four that the problem was not the workflow she had been looking at but the patient stratification she had not thought to look at. She knew she was getting expertise for twelve weeks. She did not know that she would still be using it nine months later.

The morning she asked the AI about the financial model reasoning was the morning she understood what the knowledge library was for. She had thought of it as documentation: a record of what Raymond had recommended, available for reference if she forgot a specific recommendation. The AI returned something different. It returned Raymond’s reasoning process: the questions he asked in order to arrive at the recommendation, the alternatives he rejected and why, the assumptions embedded in the model that she would need to revisit if her patient volume changed significantly. She did not have this from any document Raymond had left her. She had it from the AI’s capture of his reasoning across twelve sessions.

She queried the library forty-three times in the nine months after the deployment ended. Forty-one of those queries returned a useful answer. The other two required a call to Raymond, which she made and which he answered. He answered from Cincinnati, where he is no longer managing a project but remains available for specific questions the knowledge library cannot reach.

The care coordination workflow redesign produced results that surprised the health center’s staff.

In the first quarter after implementation, the center’s care coordination costs for its top-utilization quartile dropped 23 percent. Patient-reported satisfaction with care continuity increased on the center’s biannual survey. Two of the five providers who had been involved in the highest-utilization patient’s care were reassigned to general care coordination, reducing their caseloads and their documentation burden. The executive director of the county health department called Patricia after the survey results came in and asked what she had changed.

Raymond’s pattern recognition made this possible. Julia’s analytical work made it visible. Patricia’s institutional knowledge made it implementable: she knew which staff members would adapt quickly to the new workflow, which ones would need more support, and where the implementation would face the most resistance. None of the three of them alone could have produced what all three of them together produced.

The AI’s knowledge library captured the what and the how. It captured the financial model. It captured the care coordination methodology. It captured Raymond’s diagnostic reasoning, documented through twelve sessions of session summaries that recorded not just conclusions but the questions Raymond asked on the way to them.

What it could not capture, as the piece must honestly name, is the calibration that Raymond carries in his presence: the judgment about which institutional problems deserve immediate attention and which ones need to wait until the institution is ready to receive the solution. Patricia’s staff has the financial model and the workflow redesign. They do not have Raymond’s thirty years of reading institutional readiness. That judgment did not transfer into structure. His email address is the partial substitute, and he answers it.

The fourth account is the AI’s, and it is the most complete.

The project timeline shows the full arc: week two’s pattern recognition, week four’s scope expansion discussion, week eight’s board preparation, week eleven’s workflow implementation planning, week twelve’s final deliverable review. The session summaries show what was said in each session and in what order. The expertise capture shows the reasoning behind the recommendations: the questions asked, the alternatives considered, the assumptions stated.

Patricia can move through this record. She cannot have Raymond’s judgment about a new situation the record did not anticipate. The library is what it is: a structured memory of what the deployment built. It is not Raymond. But Raymond is in Cincinnati and the library is in the AI, and the AI is available at 7 PM on a Sunday when the board meeting is Monday morning. Raymond’s availability is more limited than the library’s, and the library answers 41 of 43 questions. That ratio matters.

What Exists Now, What Is Coming, and What Requires Time

#

Executive service corps deployments in major cities provide structured placements with deliverable requirements but no AI preparation layer and no post-deployment knowledge library. SCORE mentoring provides relationship continuity and domain expertise without the project management infrastructure or knowledge capture. Both produce real value in their current form.

BGO pilot deployments are operational with early-stage AI preparation and capture infrastructure. The knowledge library capability described here is functional in pilot form. Post-deployment query access is available to receiving institutions.

Within one to two years, full BGO deployment infrastructure: AI session preparation, real-time expertise capture, knowledge library generation, and post-deployment query access as standard components of every deployment.

Within three to five years, the BGO knowledge library as a recognized institutional asset class: insurable, documentable in grant applications, and eligible for ongoing maintenance funding as a knowledge infrastructure investment. The structured reasoning library as a model for tacit knowledge preservation across institutional types.

Patricia’s board meeting is Monday. The AI is available. Raymond’s reasoning from session seven is still in the room.