Use cases by function. Where the hours actually go.
Industry changes the vocabulary; the function changes the work. Pick the department, read the "before," and see what each maturity level would change.
Finance and accounting. month-end close, AP/AR, reconciliations, expense review.
What it looks like before: Close takes days because numbers are reconciled by hand across the ERP, bank feeds and spreadsheets. Approvals chase people by email. Exceptions land on one controller.
What each level changes
Write down the close as it actually runs, with owners per task.
Rules match invoices to POs and post the routine entries; anything outside halts loudly.
An agent reads exceptions, drafts the explanation and recommends the fix; the controller approves.
Recurring exception causes surface monthly and the rules change under review.
KPIs that move
| Metric | How to measure it |
|---|---|
| Days to close | Working days from period end to sign-off |
| Rework rate | Corrections as a share of entries |
| Touchless invoices | % of invoices posted with no manual step |
Procurement and supplier operations. supplier onboarding, PO matching, contract compliance, spend review.
What it looks like before: Suppliers are onboarded by email thread, POs are matched by eye, and contract terms live in a PDF nobody checks until a dispute.
What each level changes
One onboarding checklist and one PO-matching rulebook.
Three-way match runs on rules in the ERP you already own.
An agent reads supplier documents and invoices against contract terms and flags deviations with reasons.
Supplier performance patterns drive threshold changes monthly.
KPIs that move
| Metric | How to measure it |
|---|---|
| Straight-through match rate | % of invoices matched without touch |
| Onboarding cycle time | Days from request to approved supplier |
| Contract leakage | Overbilling caught as a share of spend |
Customer service and support. ticket triage, returns, order exceptions, first responses.
What it looks like before: Every message is read and classified by a person. Routine cases wait behind hard ones, and the same question is answered a hundred different ways.
What each level changes
Categories, owners and response standards written down.
Rules route by category and answer the fully-scripted cases.
An agent drafts the first response under policy and a person approves; routine cases resolve same day.
Contact reasons are reviewed monthly, and the product or process changes along with the reply.
KPIs that move
| Metric | How to measure it |
|---|---|
| Same-day resolution | % of cases closed the day they arrive |
| First-response time | Median minutes to first reply |
| Cases per agent | Resolved cases per person per week |
HR and people operations. onboarding, offboarding, payroll changes, policy questions.
What it looks like before: A new hire's first week depends on a dozen manual steps across HR, IT and the manager. Leaver access lingers, and the same policy question comes up in every channel.
What each level changes
One joiner-mover-leaver checklist with owners.
The HR record drives accounts, equipment and access, with a tested de-provisioning path.
An agent answers policy questions from the handbook and routes the edge cases to HR.
Recurring questions rewrite the handbook; recurring delays rewrite the checklist.
KPIs that move
| Metric | How to measure it |
|---|---|
| Time to productive | Days from start date to full access and tools |
| Leaver access closed | % of leavers de-provisioned on day one |
| HR ticket volume | Questions reaching HR per 100 employees |
Sales and revenue operations. quote-to-order, CRM hygiene, pipeline reporting, handoffs.
What it looks like before: Quotes are built in spreadsheets, the CRM is updated after the fact, and the weekly pipeline review runs on numbers assembled the night before.
What each level changes
A quote process and CRM standards everyone follows.
Quote-to-order flows from CRM to ERP without rekeying; stale records are flagged by rule.
An agent drafts follow-ups and proposals from CRM context for a rep to send.
Win/loss and cycle-time patterns feed the playbook monthly.
KPIs that move
| Metric | How to measure it |
|---|---|
| Quote turnaround | Hours from request to quote sent |
| CRM completeness | % of opportunities with required fields |
| Sales cycle | Median days from first meeting to close |
IT and operations. service desk, access requests, incident response, change management.
What it looks like before: Tickets are triaged by hand, access is copied from 'someone similar', and the same incident recurs after every change.
What each level changes
An inventory of automations with owners, and a runbook per service.
Access provisioning by role; routine tickets resolved by rule.
An agent triages tickets, drafts responses and proposes fixes; engineers approve.
Incident patterns against the change log drive runbook updates.
KPIs that move
| Metric | How to measure it |
|---|---|
| Mean time to detect | Minutes from failure to alert |
| Provisioning time | Hours from request to access |
| Runbook coverage | % of automations with a current runbook |
Questions people ask about automating a function.
Which function should we automate first?
The one where the same people rekey the same data every week and nobody measures it. Usually that's finance or customer service. The assessment tells you which of your processes is highest on effort and lowest on risk to change.
Do we need an AI agent, or just automation?
If the inputs are structured and the answer should be the same every time, a rule is cheaper, faster and easier to audit. Agents earn their keep only where variance is real: messy documents, open-ended messages, drafting from context.
Can we do this on the systems we already have?
Almost always. The first step of every Survey is finding what your current platforms can do that nobody switched on. We take no margin on any tool we recommend, so the advice is never self-interested.
Thirty minutes. Worst case, you lose half an hour.
Bring one process. You leave knowing which level it's on, what to fix first, and whether it's worth doing anything at all.