The Library · Guide

What AI automation actually looks like, department by department.

Updated July 2026

Eight worked examples across sales, operations, finance, and customer support. Each one shows the workflow before, the workflow after, and what it concretely takes to get there. Every example keeps a person at the point where judgment matters.

How to read these

Borrow the shape, not the specifics.

Every example below is the same pattern: AI takes over the repeatable step, and a person keeps the approval step. Your tools will differ, and that is fine; the pattern is what transfers. When an example feels close to a workflow you run, shortlist it, then score it before you build anything.

Sales

Where speed and consistency beat craft.

Sales automation pays off first on the work where being fast and consistent matters more than being artisanal: first replies and first drafts.

01
Inbound lead triage and first reply
Before: new inquiries wait in the inbox until someone checks, and the fastest competitor answers first.
After: AI reads each inquiry, drafts a tailored first reply, and files the lead in the CRM with what they asked for; a person approves the send in seconds. What it takes: an inbox AI can reach, a CRM, and a written definition of a qualified lead.
02
Proposal and quote drafting
Before: every proposal starts from last month's document, and the hours go to boilerplate instead of the parts that win the deal.
After: AI assembles a first draft from your call notes, price list, and an approved template; the owner edits the judgment parts only. What it takes: call notes stored somewhere reachable, a current price list, and one good proposal to serve as the template.
Operations

The reports and re-typing nobody was hired to do.

Ops wins are unglamorous and immediate: the recurring report and the manual data entry that quietly eat a day every week.

03
Weekly status reporting
Before: someone spends Friday afternoon chasing numbers across tools and pasting them into the same document shape as last week.
After: AI pulls the numbers, drafts the report in your format, and flags what changed enough to deserve a sentence; a person reviews and sends it. What it takes: metrics that live in systems rather than in heads, and a written report format.
04
Documents into systems
Before: orders, delivery notes, and confirmations arrive as PDFs and emails, someone re-types them into the tracking sheet, and the typos surface weeks later.
After: AI extracts the fields into the system and routes anything ambiguous to a person, so review happens at the exceptions instead of at every entry. What it takes: reasonably consistent documents, one agreed place the data lands, and a rule for what counts as ambiguous.
Finance

AI proposes. A person always approves.

Money workflows are where errors cost the most, so the pattern here is strict: AI prepares, flags, and drafts; a person signs off on anything that moves money or touches a relationship.

05
Transaction coding
Before: month-end means a pile of receipts and card lines nobody remembers, and an evening or two of categorizing them.
After: AI codes each transaction against your chart of accounts and marks the ones it is unsure about; the bookkeeper reviews the marked ones and closes faster. What it takes: bank and card feeds flowing into your accounting tool, a chart of accounts, and a named owner for exceptions.
06
Receivables follow-up
Before: overdue invoices get chased when cash feels tight, in whatever tone the day produced, and some never get chased at all.
After: AI drafts reminders matched to how overdue the invoice is and what was agreed, on a schedule; a person approves anything large or attached to a relationship. What it takes: invoice statuses in a system, and an agreed reminder ladder.
Customer support

The clearest before and after on this page.

Support is where most businesses feel the difference first, because most of the volume is the same handful of questions on repeat.

07
Ticket triage and suggested replies
Before: everything lands in one queue, most of it is the same handful of questions, and response time depends on who is on shift.
After: AI tags and prioritizes each ticket and drafts answers to the routine ones from your help docs; people approve those and spend the recovered time on the hard cases. What it takes: a help center that is actually current, and a support inbox AI can reach.
08
Voice-of-customer summaries
Before: what customers keep complaining about lives across a thousand tickets, so decisions run on the loudest recent anecdote.
After: AI summarizes the week's themes with counts and example quotes; the team reads it in five minutes and argues from the same page. What it takes: tickets in one tool, and a standing slot to actually read the summary.
Picking yours

From "that one sounds like us" to a first build.

Step 01
Shortlist the lookalikes
Mark the two or three examples closest to workflows you actually run. Resemblance is the cheapest useful signal you will get.
Step 02
Score before you build
Rate each candidate against the ten criteria with the AI Use Case Scorecard. Impact and frequency count double; the totals pick for you.
Step 03
Check the foundations
If you're not sure the process and data underneath will hold, the free readiness check tells you in about 90 seconds.
Take the AI Readiness Check → Use the AI Use Case Scorecard →
FAQ

Common questions

Which department should automate with AI first? ›

The one with a frequent, documented, low-risk workflow, not the one with the loudest pain. Score your candidates on impact, frequency, clarity, data, feasibility, and risk; the totals usually point at operations or support before anything customer-critical.

Do these AI automation examples need custom software? ›

Usually not at the start. Most first versions combine the tools you already run with a capable AI model and a small amount of glue. The pattern matters more than the stack: AI does the repeatable step, and a person approves before anything ships.

What stays human in an AI-automated workflow? ›

The approval points, the exceptions, and the relationships. Every example on this page keeps a person at the step where an error would be expensive: sending, paying, promising, or apologizing.

How do I know if my business is ready for these? ›

Take the free AI Readiness Check. It scores your processes, data, tools, and capacity in about 90 seconds and recommends one of five first moves, including when the honest answer is to clean up a workflow before automating anything.

Score your own candidates next.

Join the list and the AI Use Case Scorecard spreadsheet lands in your inbox in the next minute: the weights wired in, one column per workflow. New worked examples ship to the list first.

One useful email at a time. Unsubscribe whenever.