Home Use Cases Work Insights About Contact
Back to Our Work

How an Enterprise Healthcare Provider Built a Support Agent That Answers Members from Their Own Records

Operations1
Open the walkthrough

Six-Step AI Workflow with Human Oversight

Each step shows the workspace, what the AI completed, and what still needs human review.

Steps 4 and 6 need a person. Step 4 is waiting for approval right now. Click it to open the full walkthrough.

Estimates are directional and based on stated assumptions. All names, organizations, and identifying details have been anonymized in accordance with our confidentiality agreements.

The Transformation

How an Enterprise Healthcare Provider Built a Support Agent That Answers Members from Their Own Records

Before
A person used to log every inquiry and look the member up by hand.
A person used to read every message and decide where it went.
A person used to write every reply, and the queue set the wait.
A person used to close each ticket and type up what happened.
After
every message on every channel, one conversation per member
what the member wants, read before anything answers
every reply traced to an article or a record
the short queue only a person can settle
the ticket closed, confirmed and on the record
Executive Summary

Answers a member's question the moment it arrives, on whichever channel it arrives, and never sends a sentence it cannot trace to an article or to the member's own record. Written for an org where the queue was the worst part of the product: the repetitive questions get answered at once and around the clock, the hard ones reach a person carrying the whole story, and the one thing a person signs each week is what the bot is allowed to say. Six stations, run as a loop: the article station 6 approves is the article station 3 answers from.

What Was Broken

A person used to log every inquiry and look the member up by hand
A person used to read every message and decide where it went
A person used to write every reply, and the queue set the wait
The real cost
A person used to close each ticket and type up what happened.

What We Built

Six stations and 14 subagents. Each subagent carries its own tasks and its own refusal.

A4
Channel Watch
Poll chat, email and SMS, and stamp each message with its channel and time
A3
Identity Match
Match the sender to a member record by phone, email or member number
A3
Intent Reading
Classify each message as one of the questions the desk actually handles: claim status, refill, appointment, billing, card replacement, plan detail
A4
Safety Screen
Screen every message for symptom wording or anything urgent
A3
Knowledge Search
Search the article library and settled tickets for passages that answer this question
A2
Answer Writing
Compose the reply from the passages found, at the length the channel takes
A4
Sentiment Watch
Watch every live conversation for frustration building turn over turn
A3
Handoff Routing
Route each handoff to the person on shift who can settle that kind of question
A2
Handoff Brief
Summarize each handoff in one line a person can decide from
A3
Resolution Check
Confirm with the member that the answer settled the question
A4
Record Keeping
Write the full transcript, intent and outcome onto the ticket record
A2
Gap Finding
Group the week's handoffs by what the bot was missing when it handed over
A3
Answer Audit
Check every article the bot cited this week is still the current version
A4
Metrics Reporting
Assemble the week's counts: settled by the bot, handed to a person, wait at the handoff queue

How It Runs

1

Inquiry Intake

Nobody is asked anything here

Agents watch every channel members write on, match each sender to a member record, and stitch a member's email and chat about the same matter into one conversation. Nothing waits on a person here. An unverified sender reaches somebody only through station 4.

2

Intent Reading

Nobody is asked anything here

Agents read each message for what the member wants, pull out the claim numbers and dates the answer will need, and screen every turn for anything a bot must never answer. Nothing waits on a person here. What cannot be read goes to station 4 with its transcript, and anything clinical goes straight to the nurse line.

3

Answer Writing

Nobody is asked anything here

Agents search the article library, past settled tickets and the member's own claim and appointment records, then write a reply with a source behind every sentence and send it on the channel the member asked on. Nothing waits on a person here. A question with no answer on file is handed over, never improvised.

4

Human Handoff

A person answers here

Agents pull a conversation out of the bot's hands the moment it is stuck or the member is upset, and hand it to the right person as one line they can act on, with the transcript and the history already attached. This is the station where a person works. Everything else exists to keep this queue short.

5

Ticket Closing

Nobody is asked anything here

Agents confirm with the member that the answer settled it, write the full transcript, cause and outcome onto the ticket, and reopen the ticket when the member comes back on any channel. Nothing waits on a person here. A reopened ticket goes back through the loop on its own.

6

Knowledge Review

A person answers here

Agents group the week's handoffs by what the bot was missing, draft the article that would have answered the commonest gap, and trace complaints back to the answer that caused them. A person signs the week. The bot says only what somebody approved, and an answer proven wrong comes down before anything else goes up.

Where a Person Decides

Step 4, Human Handoff. Read the one-line handoff and answer, or take the conversation live. Take K. Vance first, the clock ranks the queue for you. the member is now waiting on you.
Step 6, Knowledge Review. Confirm the pulled answer stays down, and approve its corrected version. Accept or reject the drafted article, then sign the week. a wrong answer outranks a missing one.

Operating Model

This changes how work flows through the team.

Role
Responsibility
Human Handoff owner
Read the one-line handoff and answer, or take the conversation live. Take K. Vance first, the clock ranks the queue for you. the member is now waiting on you.
Knowledge Review owner
Confirm the pulled answer stays down, and approve its corrected version. Accept or reject the drafted article, then sign the week. a wrong answer outranks a missing one.

What Transfers, What Must Be True

What transfers
A person is in the loop wherever a member is waiting on a judgment or the bot's own script changes. Two of the six stations refuse to proceed on their own, Human Handoff and Knowledge Review, and those are the two a person carries. Everywhere else the agents run at volume and reach you only when they cannot settle something.
Human Handoff stops for a person, and the member is now waiting on you.
Knowledge Review stops for a person, and a wrong answer outranks a missing one.
Every subagent says what it will not do. 14 of them do.
What must be true in your environment
The agent can read the systems your records already live in. This one reads 17.
Somebody owns Human Handoff and has time for it.
Somebody owns Knowledge Review and has time for it.

Failure Modes

What breaks this pattern:

✗ Records opened on a guess

The bot matches a sender to a member record on a name that looks close. It reads one person's health history to answer another person's question, and nobody knows until the member does.

✗ The bot answers symptom messages

A member describes chest pain, and an article about chest pain matches it. The bot sends the article, and a machine is now the first responder to a medical complaint.

✗ Unsupported sentences reach members

The bot writes what sounds right instead of what a passage says. A member acts on a coverage claim that no policy backs, and the company owns that sentence.

✗ Angry members stay with the bot

The bot keeps answering a member who has already stopped trusting it. Every correct reply after that point makes the conversation worse, because the member wanted a person three messages ago.

Directional Outcomes

What the agent counts, and the station that counts it.

These counts are the tallies from one monitored run of the agents. They are not monthly or annual totals.

Conversations matched to a member
Counted at Inquiry Intake
6,412
Matched to the wrong member
Counted at Inquiry Intake
0
Held for a person to verify
Counted at Inquiry Intake
6
Conversations read and routed
Counted at Intent Reading
6,412
Sent to the bot when they belonged with the nurse line
Counted at Intent Reading
0
Held as unclear and handed to a person
Counted at Intent Reading
121
Replies sent
Counted at Answer Writing
11,730
Sentences sent without a source
Counted at Answer Writing
0
Questions it declined and handed over
Counted at Answer Writing
388
Tickets closed with a full record
Counted at Ticket Closing
6,380
Closed while the member was still asking
Counted at Ticket Closing
0
Reopened when the member came back
Counted at Ticket Closing
57
Our measurement policy: We do not publish precise ROI without baseline methodology. Every figure above carries its basis.

What Runs Where

Every step names the subagent that does the work, the record it writes, the thing that raises a question for a person, and what it is allowed to touch. This is drawn from the source, not from a diagram somebody kept in sync by hand.

1Inquiry Intake
subagentchannel-watch
writesqueue/inbound/<conv-id>.json
raisesidentity-unverified
may touchqueue/**, members/** read-only
2Intent Reading
subagentread-intent
writesqueue/understood/<conv-id>.json
raisesintent-unclear
may touchqueue/**, history/** read-only
3Answer Writing
subagentanswer-from-kb
writesqueue/answered/<conv-id>.json
raisesno-source-found
may touchqueue/**, kb/** read-only
4Human Handoff
GATE
subagenthandoff-brief
writesqueue/handoffs/<conv-id>.json
raisesawait-human-answer
may touchqueue/handoffs/**, everything else read-only
5Ticket Closing
subagentclose-ticket
writestickets/closed/<conv-id>.json
raisesmember-still-waiting
may touchtickets/**, queue/** read-only
6Knowledge Review
GATE
subagentweekly-review
writeskb/proposed/<article-id>.json, reports/week-<n>.json
raisessign-kb-change
may touchkb/**, reports/**, tickets/** read-only

Stack

Every system this agent reads or writes.

System
Read at
Stations
past settled tickets
Answer Writing
1 of 6
the SMS number
Inquiry Intake
1 of 6
the chat widget on the member portal
Inquiry Intake
1 of 6
the claims and appointment systems
Answer Writing
1 of 6
the help article library
Answer Writing
2 of 6
the intent model
Intent Reading
1 of 6
the member directory
Inquiry Intake
1 of 6
the member's ticket history
Intent Reading
1 of 6
the metrics dashboard
Knowledge Review
1 of 6
the nurse line
Intent Reading
2 of 6
the on-shift roster
Human Handoff
1 of 6
the satisfaction survey
Ticket Closing
1 of 6
the support mailbox
Inquiry Intake
1 of 6
the support team's queue
Human Handoff
1 of 6
the ticket archive
Ticket Closing
1 of 6
the ticketing system
Ticket Closing
1 of 6
the weekly review
Knowledge Review
1 of 6
Next Step

Want to see if this pattern fits your customer support bots?

No build commitment·Real samples, not a demo·Estimate in writing