AI-supported workflow scope coordinating a global payment chase.
Two Mach Lilies AI Helpers support agent-payment documentation and follow-up through a shared Outlook workflow and custom dashboard; people control financial validation, approval and release.
inDinero
- 47% less Human time spent Across the measured workflow
- 82% shorter End-to-end payment cycle Across the full payment loop
- 1,780 Reported global agent network Network population · not completed cases
Documentation and follow-up · human-controlled decisions · shared Outlook workflow · custom dashboard · Not replacement or automation scope.
Adjacent live deployment · client-reported · unaudited · source detail not heldClient-reported and not independently audited; measurement period, sample detail, source detail not held, and calculation formula not supplied. This adjacent live deployment is not Quiet Quarter, not a UK-practice result, and not MTD proof. The 1,780 figure is reported network population, not cases; the 21-person team and two Helpers are wider-process scope, not replacement scope, and financial validation, approval and release remain human-controlled.
The payment was only the final step.
Before anything could be paid, the information behind each agent payment had to be complete, consistent and validated. In practice that is one coherent exception system: any of six recurring difficulties can pause a case, and each pause needs another round of follow-up before the wider process can continue.
- Account-format differences
- Details can arrive in different structures and formats.
- Identity and address mismatches
- Account names or international address information may not align.
- IBAN and bank-detail inconsistencies
- Exceptions require controlled validation by people.
- Missing supporting information
- Payment preparation can pause while additional data is requested.
- Language and communication barriers
- Global agents and representatives can require clearer or repeated follow-up.
- Multiple internal stages
- Work moves through several team-controlled steps before payment.
Any of these can recur on the same case, and none is resolved by the AI alone: final validation of financial details always sits with people.
How the operation fits together.
One shared surface sits between a reported global agent network and the people who control payment. The Helpers work inside that surface — never around it.
- Network context Reported global network 1,780 agents The reported operating population — not a caseload and not a completion count.
- Shared surface Shared Outlook workflow The agreed workflow where agent-payment queries and follow-up live.
- AI-supported 2 Mach Lilies AI Helpers Support documentation follow-up and keep missing information moving.
- Shared surface Custom operating dashboard Keeps replies, status and exceptions visible to the team.
- Team context 21-person team across the wider process People work across several internal stages around the measured workflow.
- Human-controlled Human-controlled payment stages Validation, approval and release stay with people.
In plain terms: agent-payment queries connected to a reported global network of 1,780 agents arrive in a shared Outlook workflow. 2 Mach Lilies AI Helpers work in both directions between that shared workflow and a custom operating dashboard, keeping follow-up, replies and status visible. A 21-person accounting team works across the wider process, and every payment stage — validation, approval and release — is controlled by people. The 1,780 figure is network context: the reported operating population, not the number of completed payment cases.
Client-reported and not independently audited; measurement period, sample detail, source detail not held, and calculation formula not supplied. This adjacent live deployment is not Quiet Quarter, not a UK-practice result, and not MTD proof. The 1,780 figure is reported network population, not cases; the 21-person team and two Helpers are wider-process scope, not replacement scope, and financial validation, approval and release remain human-controlled.
The same chase, restructured.
The loop did not disappear — repeated follow-up is the nature of international payment preparation. What changed is where the coordination work sits and how visible each case stays.
Before
- Agent or representative submits information
- Team identifies what is missing or inconsistent
- Team prepares follow-up
- Reply arrives
- Team rechecks the information
- Case moves onward or requires another chase
With Mach Lilies
- Agent-payment query enters the agreed shared workflow
- AI-supported process identifies missing-information context
- Follow-up is coordinated through the shared inbox
- Replies and status remain visible
- Exceptions return to people People decide
- The wider payment process continues
Financial validation, exceptions and payment decisions remain with people.
Follow one query around the loop.
Six stages, moved only when you choose — nothing advances on its own. The loop can repeat: a case that comes back incomplete returns to follow-up, which is exactly why coordination is the expensive part of payment preparation.
Stage 1 of 6
-
Stage 1 of 6 Agent-payment query
- What entered the stage
- An agent-payment query connected to the reported global network arrives in the shared Outlook workflow.
- What the AI-supported workflow did
- The AI-supported process registers the query on the agreed shared surface, so the case is visible from its first minute.
- What remained human-owned
- The team owns the workflow and can pick up any case directly at any point.
- What visible state changed
- A new case appears in the shared queue.
-
Stage 2 of 6 Missing or inconsistent information
- What entered the stage
- Details in differing account formats, international addresses, account-name mismatches or IBAN inconsistencies.
- What the AI-supported workflow did
- The AI-supported process identifies the missing-information context — what has arrived, and what is still needed before the case can progress.
- What remained human-owned
- People decide what counts as acceptable information; the Helpers never validate sensitive financial details.
- What visible state changed
- The case is marked as waiting on additional information.
-
Stage 3 of 6 Follow-up
- What entered the stage
- A case that cannot progress until the agent or representative supplies more.
- What the AI-supported workflow did
- Follow-up is coordinated through the shared inbox — and repeated where language or communication barriers mean a request needs restating.
- What remained human-owned
- Sensitive or unusual communication is handled by people.
- What visible state changed
- The follow-up and its status stay visible in the shared workflow and on the dashboard.
-
Stage 4 of 6 Reply or supporting data
- What entered the stage
- A reply, document or corrected detail from the agent or representative.
- What the AI-supported workflow did
- The AI-supported process connects what came back to its case, keeping replies and status visible.
- What remained human-owned
- People judge whether the information is acceptable — or whether the loop needs another pass.
- What visible state changed
- The case moves from waiting to reply received; incomplete cases return to follow-up.
-
Stage 5 of 6 Validation or exception route
- What entered the stage
- A case with its information assembled — or an exception such as a bank-detail mismatch.
- What the AI-supported workflow did
- Exceptional or unresolved work is returned to people, with its context attached.
- What remained human-owned
- Confirming identities, resolving name, address and IBAN exceptions, and every payment-related decision.
- What visible state changed
- The case is marked awaiting human validation, or routed as an exception.
-
Stage 6 of 6 Wider payment process
- What entered the stage
- A case validated by a person.
- What the AI-supported workflow did
- The Helpers’ part is done; the dashboard keeps progress visible while the case moves on.
- What remained human-owned
- The 21-person team carries the case through several internal stages; releasing or executing payments stays with people.
- What visible state changed
- The case leaves the follow-up loop and continues through team-controlled payment stages.
Client-reported and not independently audited; measurement period, sample detail, source detail not held, and calculation formula not supplied. This adjacent live deployment is not Quiet Quarter, not a UK-practice result, and not MTD proof. The 1,780 figure is reported network population, not cases; the 21-person team and two Helpers are wider-process scope, not replacement scope, and financial validation, approval and release remain human-controlled.
What the team sees.
No private Outlook or dashboard screenshots are published. This is a sanitised reconstruction of the two shared surfaces, using only synthetic identifiers — it shows the shape of the workflow, not customer data.
Shared inbox · agent-payment follow-up
- Agent 0147 Bank-detail mismatch Additional information requested
- Agent 0293 Missing supporting information Follow-up sent
- Agent 0415 Account-name mismatch Reply received
- Agent 0562 IBAN inconsistency Awaiting human validation
- Agent 0688 Account-format difference Exception — returned to people
Custom dashboard · case progress
Agent 0147 Bank-detail mismatch
- Missing information identified
- Additional information requested
- Reply received
- Awaiting human validation
- Human review
Sanitised workflow reconstruction and illustrative interface with no customer data; it is not a real screenshot, event log or additional performance proof.
Two improvements. Two clearly defined measures.
The two results are separate measures of separate things and are never combined into one number. Both render on a full 0–100 index scale, derived from the client-reported comparison.
Human time spent · index
Index scale 0–100, uncropped.
47% lower · less human time spent across the measured workflow.
Indexed visual derived from the client-reported reduction in time spent across the measured workflow. It does not represent invented hours or minutes.
| Measure | Index |
|---|---|
| Reported baseline human-time index | 100 |
| Reported human-time index during deployment | 53 |
| Difference | 47% lower |
End-to-end payment cycle · index
Index scale 0–100, uncropped.
82% shorter · a shorter end-to-end payment cycle across the full payment loop.
Indexed visual derived from the reported change across the full payment loop. It does not represent invented business-day values.
| Measure | Index |
|---|---|
| Reported baseline payment-cycle index | 100 |
| Reported cycle index during deployment | 18 |
| Difference | 82% shorter |
Client-reported and not independently audited; measurement period, sample detail, source detail not held, and calculation formula not supplied. This adjacent live deployment is not Quiet Quarter, not a UK-practice result, and not MTD proof. The 1,780 figure is reported network population, not cases; the 21-person team and two Helpers are wider-process scope, not replacement scope, and financial validation, approval and release remain human-controlled.
What the numbers describe.
- 2 AI Helpers
- The reported configuration — not measured payment cases.
- Shared Outlook workflow
- The agreed surface where follow-up happens.
- Custom Operating dashboard
- Where replies, status and exceptions stay visible.
- 1,780 Reported global agent network
- Network context — the reported operating population, not completed payment cases.
- 21 People across the wider process
- Team context — the results are workflow measures, not a change in team size.
None of these figures implies that every agent was active in the measured period, that every team role was automated, or that the reported time reduction describes a smaller team. The percentages describe the measured workflow, not headcount.
Adjacent live deployment · client-reported · unaudited · source detail not heldClient-reported and not independently audited; measurement period, sample detail, source detail not held, and calculation formula not supplied. This adjacent live deployment is not Quiet Quarter, not a UK-practice result, and not MTD proof. The 1,780 figure is reported network population, not cases; the 21-person team and two Helpers are wider-process scope, not replacement scope, and financial validation, approval and release remain human-controlled.
AI-supported. Human-owned.
The boundary is the point of the design: coordination is supported, judgement is not transferred.
AI-supported operating scope
- Shared-inbox agent-payment follow-up.
- Missing-information communication.
- Maintaining visibility through the chase loop.
- Supporting progress through the agreed workflow.
- Returning exceptional or unresolved work to people.
Human-owned scope
- Confirming identities and sensitive financial details.
- Deciding whether information is acceptable.
- Resolving name, address and IBAN exceptions.
- Approving payment-related decisions.
- Releasing or executing payments.
- Handling sensitive or unusual communication.
Financial validation, exceptions and payment decisions remain with people.
Read the result and how it was measured.
- Relationship
- Ongoing live proof of concept
- Deployment start
- Configuration
- 2 Mach Lilies AI Helpers
- Operating surfaces
- Shared Outlook workflow and custom dashboard
- Reported operating population
- The reported operating population is a global network of 1,780 agents, not completed payment cases.
- Team context
- A 21-person team supports the wider process; this is not replacement scope.
- Result 1
- The client reports 47% less human time spent across the measured workflow.
- Result 2
- The client reports an 82% shorter end-to-end payment cycle across the full payment loop.
- Result source
- Internal client comparison
- Audit status
- Not independently audited
- Financial outcome
- Not published
inDinero reports a global operating network of 1,780 agents. This is network-population context, not a caseload or a count of agents contacted, processed, completed or paid. The results are specific to this workflow and internal comparison and do not guarantee the same outcome in another organisation.
Adjacent live deployment · client-reported · unaudited · source detail not heldClient-reported and not independently audited; measurement period, sample detail, source detail not held, and calculation formula not supplied. This adjacent live deployment is not Quiet Quarter, not a UK-practice result, and not MTD proof. The 1,780 figure is reported network population, not cases; the 21-person team and two Helpers are wider-process scope, not replacement scope, and financial validation, approval and release remain human-controlled.
A different workflow that demonstrates an adjacent operating pattern.
This deployment concerns international agent-payment documentation and follow-up, not UK MTD client-record chasing. The adjacent operating pattern is narrower: a repeatable chase loop, shared work surfaces, visible exceptions and a named human owner.
The inDinero percentages describe this deployment and its internal comparison; they are not presented as expected Quiet Quarter results.
Read how Quiet Quarter works, see how data safety is scoped, take the readiness scorecard, or book an ai workflow assessment.
Bring us one repeatable chase loop.
We will help identify what can move safely, what evidence should be measured and what must remain human-owned.