Home Use Cases Work Insights About Contact
Back to Our Work

How a Growth-Stage Technology Company Built a Request Loop That Keeps a Record Anyone Can Ask

Administration1
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 3 and 6 need a person. Step 3 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 a Growth-Stage Technology Company Built a Request Loop That Keeps a Record Anyone Can Ask

Before
An office administrator used to read every request and work out what it was.
A person used to chase each fact across five systems by hand.
A person used to key each change into every system by hand.
A person used to ask around until somebody remembered.
After
everything raised, named, with the unclear ones held
every fact the request needs, pulled with its source
money and access, waiting on your word
carried out system by system, read back before done
every finished request, findable, quotable
Executive Summary

Every request an employee raises, from the moment it is raised to the moment it shows up in the Friday numbers. The agents name each request, gather what it needs from the systems that hold it, carry it out once it is cleared, and keep a record anyone can ask. A person signs wherever money or access moves, and decides how the process itself changes. Six stations, run as a loop: what Friday's report finds about a snag becomes the rule intake runs on Monday.

What Was Broken

An office administrator used to read every request and work out what it was
A person used to chase each fact across five systems by hand
A person used to key each change into every system by hand
The real cost
A person used to ask around until somebody remembered.

What We Built

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

A4
Request Sweep
Collect every request off the chat channel, the email alias and the form
A3
Type Match
Match each request to one of the request types the company already runs
A4
System Pull
Pull the fields each request type needs from the systems that hold them
A3
Package Check
Assemble each request's facts into one package a signer can read in a minute
A3
Approval Router
Route each assembled package to the sign-off steps its request type carries
A1
Brief Writer
Draft the one-line case above each waiting request: what is asked, what it costs
A3
Decision Record
Record each decision, who made it, and what was on screen when they did
A3
Task Runner
Run each cleared request's steps across the systems its type touches
A4
Progress Keeper
Track every open request against the timeline its type promises
A4
Record Writer
Close each finished request with its whole trail: who asked, who signed, what was done, in which systems
A4
Record Search
Search every closed request and filed document behind a question
A1
Answer Builder
Write the answer in three lines, each carrying the request or document it came from
A4
Report Builder
Compile the week's report off the record alone: raised, cleared, done, how long each type took
A2
Process Proposal
Propose the process change that would have stopped each repeated snag

How It Runs

1

Request Intake

Nobody is asked anything here

Agents collect every request off every route it can arrive by, attach it to the person who raised it, match it to a request type the company already runs, and hold anything that names no ask. Nothing waits on a person here. A held request is never dropped and never guessed at.

2

Fact Gathering

Nobody is asked anything here

Agents pull the facts each request type needs from the systems that hold them, name where every fact came from, show the ones that disagree, and say plainly when a fact exists nowhere. Nothing waits on a person here. A gap in a package is written down, never papered over.

3

Sign-Off

A person answers here

Agents clear the requests policy already answers, and put the rest in front of a person with the ask in one line, the cost beside it, and every conflicting or missing fact marked above the button. This is where a person works. Everything else exists so that three requests, not thirty, reach this screen.

4

Carrying Out

Nobody is asked anything here

Agents run each cleared request's steps across the systems it touches, read every change back before calling it done, tell the raiser where their request stands, and stop any run a system did not take. Nothing waits on a person here. A change that does not read back did not happen, whatever the connector said.

5

The Record

Nobody is asked anything here

Agents close each finished request onto the company record with its whole trail, answer questions out of that record with the line each claim came from, and say not found where the record does not answer. Nothing waits on a person here. An answer the record cannot support is refused, not softened.

6

Reporting

A person answers here

Agents build the week's report off the record alone, mark the number that broke its own pattern, and propose the process changes that would have stopped the week's repeated snags, with what each change would slow written beside what it saves. A person accepts or rejects each change. How the company runs a request is not something an agent changes quietly.

Where a Person Decides

Step 3, Sign-Off. Sign the offsite venue, the quote and the budget line are above it. Decide Lucia's finance access, and pick the cost center that is right. money and access move on your word.
Step 6, Reporting. Accept or reject the two proposed changes. Say which onboarding, the 9 steps or the 6, is the one everyone runs. the process is yours to change.

Operating Model

This changes how work flows through the team.

Role
Responsibility
Sign-Off owner
Sign the offsite venue, the quote and the budget line are above it. Decide Lucia's finance access, and pick the cost center that is right. money and access move on your word.
Reporting owner
Accept or reject the two proposed changes. Say which onboarding, the 9 steps or the 6, is the one everyone runs. the process is yours to change.
Request Intake, when it goes wrong
Look here only if somebody says their request vanished.
Fact Gathering, when it goes wrong
Look here only if a fact reads differently than the system you trust.
Carrying Out, when it goes wrong
Look here only if a system shows something the record does not.

What Transfers, What Must Be True

What transfers
A person is in the loop wherever money moves, access is granted, or the process itself changes. Two stations refuse to proceed on their own, Sign-Off and Reporting, and those are the two a person carries. Everywhere else the agents name, gather, carry out, read back and record, and reach you only when the decision is yours to make.
Sign-Off stops for a person, and money and access move on your word.
Reporting stops for a person, and the process is yours to change.
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 21.
Somebody owns Sign-Off and has time for it.
Somebody owns Reporting and has time for it.

Failure Modes

What breaks this pattern:

✗ A guessed field reaches sign-off

The package looks complete. One field was filled with a guess, and the approver signs a number nobody actually pulled.

✗ Money moves on a conflicted fact

Two systems disagree on the amount. The payment goes out anyway, and the cleanup starts after the money is gone.

✗ The run reports a failed write

The agent sends the change and moves on without reading it back. The target system dropped it, and the request closes on work that did not happen.

✗ Search returns the nearest thing

Someone asks what happened on a request the record does not hold. The search hands back a similar one, and the wrong answer carries a real citation.

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.

Requests typed
Counted at Request Intake
6,200
Requests typed wrongly
Counted at Request Intake
0
Requests held for a person
Counted at Request Intake
58
Fields pulled
Counted at Fact Gathering
41,300
Facts invented to fill a gap
Counted at Fact Gathering
0
Conflicts shown to a person
Counted at Fact Gathering
122
Changes written and read back
Counted at Carrying Out
18,700
Changes marked done that a system did not take
Counted at Carrying Out
0
Stopped runs handed to a person
Counted at Carrying Out
31
Questions answered from the record
Counted at The Record
3,400
Answers given beyond the record
Counted at The Record
0
Questions sent to the person who knows
Counted at The Record
210
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.

1Request Intake
subagentrequest-intake
writesrequests/open/<req-id>.json
raiseshold-unmatched-request
may touchrequests/open/**, requests/held/**, the channels read-only
2Fact Gathering
subagentfact-gathering
writesrequests/assembled/<req-id>.json
raisesfield-no-system-holds
may touchrequests/assembled/**, the HR system, the finance system, the CRM and the drive read-only
3Sign-Off
GATE
subagentsign-off-desk
writesrequests/cleared/<req-id>.json
raisesawait-your-signoff
may touchrequests/cleared/**, requests/assembled/** read-only
4Carrying Out
subagentcarry-out
writesrequests/done/<req-id>.json
raiseswrite-did-not-read-back
may touchrequests/done/**, the finance system, the directory, the ordering tool and the CRM through their connectors
5The Record
subagentrecord-keeping
writesrecord/closed/<req-id>.json, answers/<q-id>.json
raisesanswer-not-in-the-record
may touchrecord/**, answers/**, everything else read-only
6Reporting
GATE
subagentreport-and-tune
writesreports/week-<n>.json, process/proposed/<change-id>.json
raisesaccept-process-change
may touchreports/**, process/proposed/**, the record read-only

Stack

Every system this agent reads or writes.

System
Read at
Stations
the CRM
Fact Gathering
2 of 6
the HR system
Fact Gathering
1 of 6
the access directory
Carrying Out
1 of 6
the access policy
Sign-Off
1 of 6
the assembled packages
Sign-Off
1 of 6
the budget lines
Sign-Off
1 of 6
the chat tool
Request Intake
1 of 6
the closed requests
The Record
2 of 6
the company wiki
Fact Gathering
2 of 6
the email alias
Request Intake
1 of 6
the employee directory
Request Intake
1 of 6
the filed documents
The Record
1 of 6
the finance system
Fact Gathering
2 of 6
the last eight weeks of reports
Reporting
1 of 6
the last two years of trails
The Record
1 of 6
the ordering tool
Carrying Out
1 of 6
the request form
Request Intake
1 of 6
the request types
Reporting
1 of 6
the shared drive
Fact Gathering
1 of 6
the spending limits
Sign-Off
2 of 6
their connectors
Carrying Out
1 of 6
Next Step

Want to see if this pattern fits your internal tools?

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