How do I measure AI ROI for my business?

What you will be able to do

You will build a lightweight ROI scorecard for one AI workflow so you can keep, fix, or cancel it with evidence.

What ROI means here

For a small service business, AI ROI is usually:

(Value from better/faster work − full cost of the tool and the people around it)

Value often shows up as:

• Hours returned to owners or staff
• Faster response on quotes or messages
• Fewer missed steps on checklists
• Slightly higher conversion when follow-up becomes consistent

It rarely shows up as a dramatic overnight revenue spike from “having AI.”

Step 1: Name one workflow

Write:

• Trigger
• Output
• Owner
• Where quality fails today

Example: “After we email a quote, draft a 48-hour follow-up; coordinator edits and sends.”

No workflow name, no ROI.

Step 2: Baseline for two weeks without changing tools

Record:

• Volume (how many times)
• Average minutes per item
• Error or redo rate (wrong price, wrong tone, missed send)
• Business result if relevant (replies, bookings)

Rough numbers beat precise fantasy.

Step 3: Add AI and keep the same definitions

Run the same workflow with AI drafts for two to four weeks. Measure the same fields plus:

• Minutes to edit AI output into something sendable
• Times you rejected the draft and started over

If edit time erases the gain, the prompt kit or the job choice is wrong—not “AI doesn’t work.”

Step 4: Count full cost

Include:

• Subscription for the period
• Setup and prompt-building time (once)
• Training time
• Ongoing review time
• Any new mistakes that cost goodwill or rework

Do not pretend review is free. Review is part of the system.

Step 5: Convert carefully

Minutes saved matter when they become:

• More jobs completed
• Less overtime
• Less owner nights on admin
• A delayed hire you no longer need for that slice of work

If saved minutes turn into scrolling, your financial ROI is near zero even if the tool “feels faster.” Assign the saved block to a real use for the test period (call backlog, quote follow-ups, training).

A one-page scorecard template

Copy this:

• Workflow:
• Period:
• Volume before / after:
• Minutes per item before / after:
• Edit minutes with AI:
• Defects before / after:
• Tool + training cost:
• Owner judgment (keep / fix / kill):
• Next review date:

Fill it in a single sitting at day 30. Schedule day 90 if you kept it.

Quality is part of ROI

A faster bad email is negative ROI. Watch for:

• Invented policies or prices
• Tone that does not match your shop
• Customer confusion in replies
• Staff skipping review because “it sounded fine”

Put a hard rule: no send without a human skim for customer-facing text.

What not to count as proof

• Number of prompts run
• Number of tools purchased
• Impressive demo screenshots
• One lucky viral post

Activity is not return.

When ROI is “not yet”

Sometimes the first month is training cost. Decide in advance:

• Soft expectation at day 14 (are drafts usable?)
• Hard keep/kill at day 30 (did net time improve?)

If day 30 fails, change the job or the recipe before you buy more seats.

Example: quote follow-up scorecard (illustrative pattern)

Suppose your office sends 20 quote follow-ups a week. Baseline: 8 minutes each to write from scratch, with occasional missed sends. After AI: 3 minutes to draft + 3 minutes to edit, and fewer missed sends because the coordinator runs a fixed afternoon batch.

You would record volume, minutes, defects, and tool cost on the same scorecard—not a story about “AI transformed sales.” If replies rise, note it as a secondary metric. If replies stay flat but time dropped and quality held, that can still be a keep if the hours go back into calling warm leads.

Short close

Baseline, full cost, quality, keep-or-kill date—that is enough ROI math for most shops.