The AI Cost Audit — What You're Paying For And What You're Not Using
Six steps, about half a day, to find ghost seats, duplicate tools and runaway usage bills, and to match what you pay to what each tool is worth.
Ask a small-business owner what they spend on AI and you will usually get a guess. Ask for the list and you find a pattern: three overlapping chat subscriptions, a writing tool bought during a promotion, an image generator one person used twice, an automation add-on nobody remembers approving, and a usage-based bill that crept up quietly. None of it looks expensive alone. Together it is often a few hundred dollars a month of waste. This workflow breakdown shows how to audit what you pay for, what you actually use and what each tool is worth, in about half a day. Per the Day 50 rule, example figures here are illustrative, not measured, and your own numbers will differ.
Why AI Spend Drifts
AI tools are easy to buy and easy to forget. Most are monthly, cheap enough to expense without approval and sold on free trials that quietly convert. Individual staff sign up with a card, with a personal account or through a department budget, so no single person sees the whole picture. Features also migrate: your email, design and project tools may now bundle AI assistants, so you may be paying twice for the same capability.
There is a second drift, in usage-based costs. Tools billed by volume, such as API access, automation runs or generated images, can double in a month if a workflow loops or a team starts using it more. A flat subscription fails slowly. A usage bill fails fast.
The goal of an audit is not austerity. Cutting a tool that saves ten hours a month to save thirty dollars would be absurd. The goal is to match spending to value and to stop paying for things nobody uses.
Step One: Build The Full Inventory (60 minutes)
Make a spreadsheet with one row per tool. Columns: tool name, what it is for, who pays, who uses it, billing type (monthly, annual or usage), cost per month, renewal date and data it touches. Gather the facts from three places: your accounting software or card statements for the last three months, an email search for 'receipt', 'invoice' and 'subscription', and a short question to every team member: which AI or AI-adjacent tools do you use or pay for?
Ask the question without blame, as with the Day 53 audit. You will find personal accounts expensed quietly, free-tier tools holding business data and trials that never ended. The inventory is also a security exercise, because every row is a place your data may live.
You cannot judge what a tool is worth until you know it exists, who uses it and what it costs.
Step Two: Measure Real Usage (45 minutes)
For each tool, find out how often it is used. Most platforms show login history, seat activity or usage dashboards in the admin area. If they do not, ask users: in the last 30 days, roughly how many times did you use this, and for what? Record sessions per person per week.
Sort the tools into three groups. Daily use: people rely on it. Occasional use: a few uses a month. Ghost seats: paid for and unused for 30 days or more. Ghost seats are the easiest saving, because cancelling them removes cost without removing value.
An illustrative example: a team of eight pays for ten seats on a team plan at $30 each. Activity shows six seats used weekly, two used twice, two never. Dropping the two unused seats saves $60 a month, or $720 a year, in five minutes.
Step Three: Find The Overlaps (30 minutes)
List what each tool does in one line, then look for duplicates. Common overlaps: two or three general chat assistants, a writing assistant plus a chat assistant plus an AI built into your word processor, separate image generators, a meeting-notes tool plus an assistant inside your video-call software, and AI features bundled in design or project tools you already pay for.
For each overlap, ask which tool the team actually prefers and why. Run a short comparison on a real task if needed, using the fair-test method from Day 47. Then standardise on one primary tool per job, with the others cancelled or downgraded. Fewer tools also means simpler training, simpler data rules and simpler security.
Step Four: Attach A Value To Each Tool (60 minutes)
For every tool you keep, write one sentence: this tool saves roughly X hours a month on task Y. Then do the arithmetic. Hours saved per month multiplied by the hourly cost of the people involved, minus the monthly price, gives net value.
An illustrative example: a meeting-notes tool costs $20 a month. Three people each save about 40 minutes a week on write-ups, around 8 hours a month in total. At an internal cost of $40 an hour, that is $320 of time against $20 of cost. Keep it. Compare a design add-on at $25 a month that two people used twice last month: perhaps 30 minutes saved in total, worth about $20 of time. Cancel it, or downgrade.
Be sceptical of your own numbers. People tend to overstate time saved, so use measured estimates where you can: time a task before and after, as in earlier Practice cards. If you cannot name a task and a time saving, the tool has not earned its place.
Step Five: Watch The Usage-Based Bills (30 minutes)
For anything billed by volume, set limits. Most platforms let you cap monthly spend, set alerts at 50 and 80 percent of budget and restrict who can create keys or run workflows. Switch these on. Review which automations run and how often. A loop that fires on every incoming email can quietly burn a month's budget in a day.
Check, too, that no one is using a personal account for business work on a paid tier you are not aware of, and that old keys and integrations from abandoned experiments are revoked. Dormant access is both a cost and a risk.
Step Six: Negotiate And Time Your Renewals (30 minutes)
Put every renewal date in a calendar with a reminder two weeks ahead. Annual plans are often cheaper per month, but only commit to one for a tool you have used for at least three months and plan to keep. For a tool you are unsure about, stay monthly even at a premium; the flexibility is worth paying for.
Ask vendors about team or non-profit pricing, volume discounts and what happens to your data on cancellation. And before cancelling, export what you need: saved prompts, documents, transcripts and settings. Getting your data back after you have lost access is far harder.
A Worked Example: Before And After
Take an illustrative eight-person agency. Before the audit: three chat subscriptions at $30, ten seats on a writing tool at $18, a meeting-notes tool at $20, an image generator at $40, an automation platform billed at about $90 on average, and two design-tool add-ons at $25. Monthly total: roughly $430.
After the audit: two chat subscriptions consolidated into one team plan, writing tool cut to six seats, meeting notes kept, the image generator dropped in favour of the design tool's built-in feature, the automation platform capped with alerts, and one add-on cancelled. Monthly total: roughly $300. Saving: about $130 a month, or around $1,560 a year, for four to five hours of work. The real win was clarity, because the team now knew what it had and why.
Treat those figures as a template for the arithmetic, not a benchmark. Your savings could be higher or lower, and some audits will tell you to spend more, on a tool that is clearly worth it but under-licensed.
Questions To Ask Before Buying Anything New
Stop the drift at the source with five questions, asked before any new subscription. What specific task will this do, and how long does that task take today? Do we already pay for a tool that does it? Who will own it, and who will use it? What data will it touch, and is that allowed under our policy? What does it cost at the usage we expect, and what is the exit route if we stop?
If the requester cannot answer in two minutes, the purchase waits. Most impulse subscriptions die at question one. Estimated, not measured: this habit alone tends to cut new, unused subscriptions sharply, because the discipline of naming a task and a number filters out the curiosity buys.
Bundled Features: The Quiet Duplicate
One overlooked saving hides in tools you already own. Email, document, spreadsheet, design, CRM and video-call platforms increasingly bundle AI assistants, sometimes free and sometimes as a paid add-on. Before paying for a standalone tool, test the bundled feature on a real task. It is often good enough for routine work, and it comes with data terms you have already accepted.
The reverse is also true. A bundled assistant that is mediocre and expensive may be worth replacing with a standalone one that your team actually prefers. Judge by output on your tasks, not by the brand.
Where The Audit Breaks
Three honest limits. First, seat counts and logins measure presence, not value: a person may log in daily and get little from the tool, or use it once a month for something critical. Pair the data with conversation. Second, time-saved estimates are soft, and they flatter the tool. Third, consolidating onto one vendor can create lock-in and a single point of failure, so keep exports current and know your fallback.
Also remember the human cost. If you cancel a tool people enjoy using, explain why and what replaces it, or they will quietly sign up for it again on a personal card.
Make It A Habit
Run the full audit twice a year and a 20-minute check every quarter: scan the list, check usage, review renewals. Assign one owner for the inventory and require that any new tool is added to it before purchase. Put a simple rule in the Day 53 policy: no new AI tool without being on the list and having a named owner and a stated purpose. That one rule prevents most of the drift.
Build The Inventory In 60 Minutes
Open your last three months of card and bank statements and list every AI or AI-adjacent subscription. Ask your team what they pay for personally. Mark each tool daily, occasional or ghost, and cancel one ghost seat today.
