Use cases

Examples of work agentic AI can automate 6 of 7

Completing expense applications before approval

A fictional employee submits a ¥8,640 travel expense. The receipt is attached, but the visit destination and project code are missing. AI asks for the missing information and prepares the packet for an approver.

All companies, people, amounts, times, records and screens here are fictional illustrations, not customer evidence or a claim of deployed features.

01 / Trigger

The information the team first sees

Expense request / EX-208Missing details · awaiting reply

Illustrative workflow screen

Business travel · ¥8,640

Applicant
Mori / Sales (fictional)
Attachment
One receipt · date and amount checked
Missing
Visit destination / project code

Name, team, amount, request number and receipt are fictional.

02 / Manual workflow

Explain the gap and track the applicant's reply

Before approval, staff compare the form, receipt and expense policy. They tell the applicant exactly what is missing, follow up and recheck the application after the reply.

Coordinator's work boardCheck → Decide → Notify
01

Compare request and evidence

Check the ¥8,640 amount and date against the receipt and identify required policy fields.

02

Ask for missing fields

Ask for destination and project code; do not guess and track the pending reply.

03

Prepare the approval packet

Bundle the reply, receipt and policy comparison for the approver.

03 / AI workflow

AI gathers missing details; a person decides approval

On submission, AI checks required fields and evidence. If details are missing, it asks the applicant and waits, then resumes after the reply. It may explain policy fit but does not approve or deny.

Operations agent progressExpense application / EX-208
01

Find gaps

Extract missing fields and inconsistencies from form, receipt and policy.

02

Ask and resume

Ask for destination and project code; attach the reply and recheck.

03

Present to approver

Present evidence, policy, applicant's answer and any unresolved issues together.

Expense application / EX-208Sample conversation
AI
Operations agent

Mori replied that the destination is Aoba Industries and the project code is PJ-42. I rechecked the ¥8,640 receipt, date and required policy fields. Please assess the purpose and reasonableness, then approve or deny.

P
Approver · Takahashi

I checked the visit purpose and project. I approve this request. The receipt and application amounts also match.

AI
Operations agent

Recorded the approver's decision, read back the approved status and amount, and notified the applicant.

Here the applicant supplies the destination and project code, completing the required fields. AI passes the checks and receipt to the approver, who considers purpose and reasonableness and then approves or denies.

04 / Information flow

From application to completed approval packet

Processing pauses while waiting for the applicant and resumes on their reply.

Sources

Expense form

Receipt

Expense policy

Check · decide · act

1Check fields and evidence

2Ask applicant about gaps

3Recheck after reply

Records and recipients

Application record

Human approval or denial

05 / Checks and exceptions

Keep completeness separate from approval

Verify the result

After the reply, AI reads back required fields, amount and date against the receipt. It records the applicant's answer and policy reference for the approver.

The approver judges whether the expense is appropriate. A complete packet does not constitute approval.

When to pause and hand off

If the receipt conflicts, policy needs interpretation or no reply arrives, AI pauses and shows the discrepancy or wait state to staff.

Give approvers a decision-ready request

Gathering missing fields and tracking replies lets approvers review evidence, policy and purpose together. Measure returns and wait time to assess the effect.

Start with one travel-expense category and explicit required fields. Define when ambiguous policy or exceptions go to a person.

Talk about your workflow