CareLoop

CUSTOMER CARE • PRODUCT GO-LIVE • OSONWANNE GROUP LLC

CareLoop: a customer care loop for a product go-live

A working prototype built to accompany an unsolicited solution brief on customer care prototyping, prepared after the Defense Health Agency’s Customer Care Prototyping Commercial Solutions Opening (HT0038-25-S-C002) was canceled. It closes the loop from self-service help, to watching new users through a go-live, to turning unanswered questions into backlog items, to measuring whether any of it worked.

Public help content, a simulated rollout, and a falsifiable hypothesis

Answers are quoted from public U.S. Government help pages. The go-live data is synthetic and labeled that way on every screen.

100

Passages from 18 public U.S. Government help pages, each with its link and retrieval date

8

Fictional sites in a seeded 42-day go-live simulation

30%

Target cut in human-touch tickets per 100 new users, tested against a control cohort

0

Language models: answers are quoted passages, so nothing is invented

CareLoop self-service assistant answering a question with a quoted, cited passage and an event log
The self-service assistant: a quoted answer with its source, and every exchange logged as an event.

What it does

SCREEN 1 • TIER 0

Self-service assistant

Answers from public help content with quoted passages that cite the page, publisher, link and retrieval date. When the match is weak it hands off to a person with a context packet, so nobody repeats themselves.

SCREEN 2

Go-live radar

Watches every new user’s first weeks: adoption, time to first successful task, self-service containment, tickets per 100 new users, sentiment and stuck users. Stuck users trigger proactive nudges from editable rules.

SCREEN 3

Insights to backlog

Groups unresolved and low-confidence questions into themes with evidence counts and examples, and drafts a product backlog item for each.

SCREEN 4

Experiment and outcome ledger

One hypothesis, a control cohort, a 95% confidence interval, a sentiment guardrail, a weekly stop or adjust rule, and milestones tied to outcomes rather than hours.

Go-live radar with synthetic adoption, time to first task, containment, ticket, sentiment and stuck-user metrics
Go-live radar on synthetic telemetry for a fictional rollout.
Experiment ledger showing the hypothesis, treatment and control ticket rates and a simulated reduction with confidence interval
Experiment ledger. These results are simulated, not measured.

Who it helps

PATIENTS AND FAMILIES

Get help without a phone call

A quoted answer with its source, or a hand-off to a person who already sees the question and what was tried.

HELP DESK STAFF

Start with context

Each hand-off arrives with a context packet: segment, suggested routing queue, what was tried and the transcript.

PRODUCT TEAMS

Fix the real gaps

Unanswered questions become themed backlog items with evidence, instead of anecdotes.

PROGRAM LEADERS

Pay for outcomes

A pre-registered hypothesis and weekly stop or adjust rule show whether the work is moving the measure.

What is real, and what is simulated

The help content is real: a snapshot of 18 public U.S. Government pages from fehrm.gov, militaryonesource.mil and va.gov, stored with each page’s link and retrieval date. Answers are retrieved and quoted word for word; no language model writes them, so the assistant can still return the wrong passage.

Everything else is synthetic: the go-live telemetry for 8 fictional sites, the treatment effect, the seeded unresolved questions and the milestone figures. The go-live results shown on the ledger come from a seeded simulation; they are not measured results. There are no real users, sites or systems.

The prototype collects no personal or health information and nothing leaves the browser. A production version would need a HIPAA Business Associate Agreement and Government-authorized hosting; neither is claimed for this prototype. It is not past performance.

See it running

Try it with no login: ask a question, trigger a hand-off, edit a nudge rule on the radar, or read the experiment ledger.

Boniface Osonwanne, CEO • bon@osonwanne.com • 602-632-9589

CareLoop is an independent working prototype by Osonwanne Group LLC, built to accompany an unsolicited solution brief sent after the Defense Health Agency’s Customer Care Prototyping CSO HT0038-25-S-C002 was canceled. No award or selection has been made, and the Government has no obligation to review unsolicited submissions. It uses public U.S. Government help content and synthetic telemetry; the go-live results shown are simulated, not measured. It is not a Government system and not past performance. It is not affiliated with or endorsed by the Department of Defense, the Defense Health Agency or any government agency.