The AI Tool Audit Checklist: How to Find What Your Team Actually Uses
You think your team uses 5 or 6 AI tools. Maybe 7 if someone added a new one recently. The reality: the average knowledge worker at a growing startup uses 12-15 AI-powered tools, and finance only knows about half of them.
This isn't negligence. It's a structural blind spot. AI tools are easy to adopt — free tiers, credit card signups, browser extensions — and hard to track. Nobody tells finance when a product manager signs up for Jasper on a personal card. Nobody flags the $400/month Claude Pro seat that's been running on a company card nobody reconciled.
Result: shadow AI subscriptions bleeding budget, overlap between tools nobody's connecting, and no visibility when it's time to cut costs.
This checklist fixes that. 5 steps, 20-30 minutes total, and you'll know exactly what your team is paying for.
Step 1: Check Expense Reports (5 minutes)
Start where money actually moves. Pull credit card statements from the last 3 months and scan for anything AI-adjacent.
What to look for:
- Vendor names: OpenAI, Anthropic, Jasper, Copy.ai, Writesonic, Midjourney, Runway, ElevenLabs, Cohere, Perplexity, Claude, Notion AI, GitHub Copilot, Cursor, Replit, Codeium — plus any tool with "AI" or "GPT" in the name
- Line items you don't recognize: Anything billed in USD or GBP that isn't obviously infrastructure (AWS, Vercel, Figma)
- Duplicate charges: Multiple charges from the same vendor in the same month (usually different plans, different people)
Quick Win
If you're doing this manually, export 3 months of statements to a spreadsheet and search for "AI", "OpenAI", "Anthropic", "Jasper", "Midjourney", and "Copilot". Flag anything you can't explain. That's your first audit list.
Step 2: Survey Your Teams (5 minutes)
Finance found half the picture. Now ask the humans. Send a Slack message or a simple form to every team — engineering, marketing, sales, ops:
"Quick question: List every AI tool you've used in the last month — anything that helped you write, code, design, research, or automate. No wrong answers. We just want to know what tools are in use."
You'd be surprised what comes back. Most people will list 3-5 tools they use regularly but didn't think to mention because "everyone probably knows about it."
Pro tips for the survey:
- Don't ask who pays for it — that adds friction and shame. Just ask what they use.
- Include "others I can't remember" — people genuinely forget.
- Do this before you share findings — you want clean data, not people conforming to what they think you already know.
Step 3: Scan SSO and Identity Logs (5 minutes)
Your SSO provider (Okta, Google Workspace, Azure AD) logs every app your team authenticates to. This is often the single fastest path to discovering tools nobody mentioned.
How to run this:
- Okta: Reports → System Log → Filter by "Application access" in the last 90 days
- Google Workspace: Admin Console → Apps → Additional Google services → Review access
- Azure AD: Enterprise applications → All applications → Sort by "Last sign-in"
Look for apps with "AI", "GPT", "Copilot", "Assistant", or similar keywords in the name. Also flag anything with a large number of active users that wasn't on your original list.
Browser Extensions and Personal Accounts
SSO only catches enterprise-authenticated apps. AI tools added as browser extensions (Monica, Monica AI, Sider) or personal subscriptions paid outside company systems won't show up here. Survey data fills this gap.
Step 4: Review API Keys and Integrations (5 minutes)
Engineering teams often spin up AI API keys for internal projects that never make it to finance. Check:
- OpenAI API dashboard: Who has active API keys? What's the spend?
- Anthropic API dashboard: Same check
- GitHub: Any Copilot Business/Enterprise licenses beyond what IT provisioned?
- Internal tools: Any homegrown apps that call AI APIs on behalf of employees?
API spend is particularly easy to miss because it often hits a corporate card as "platform services" rather than "AI tool subscription."
Step 5: Calculate the Total (5 minutes)
You now have three data sources. Cross-reference them and build your audit list.
| Tool Name | Where Found | Est. Monthly Cost | Known Seats | Notes |
|---|---|---|---|---|
| ChatGPT Team | Expense report + Survey | $240/mo | 20 | All confirmed users |
| Claude API | API logs | $180/mo | 5 devs | 3 internal tools consuming this |
| Jasper | Expense report | $125/mo | Unknown | 2 survey responses mention it |
| Monica AI (extension) | Survey | $0 (free tier) | 3 | Flag for review |
Add up the monthly total. Compare to what finance was tracking. The gap is your hidden AI spend.
What a Typical Audit Finds
- 2-4 tools nobody knew were running on company cards
- 1-2 personal subscriptions still expensed as "software"
- 3-5 users on plans where team plans existed (paying for individual + team)
- 1-3 free-tier tools that have active usage worth upgrading — or consolidating
- $300-2,000/month in tools that can be eliminated or consolidated
What to Do With the Results
The audit is the hard part. Now you have options:
- Eliminate duplicates: If two tools do the same job, pick one and cancel the other. Average overlap costs $200-500/month.
- Migrate to team plans: If three people are on individual plans, one team plan is usually cheaper.
- Cancel unused seats: Tools with low engagement that someone renewed out of habit.
- Consolidate billing: Move everything to one corporate card or platform so the next audit is easier.
Set a recurring audit cadence. Do this once a quarter. AI tool sprawl grows fast — a tool adopted in January that's paid annually and forgotten by March can run for 9 months before anyone notices.
Track Every AI Tool Automatically
StackPilot discovers AI tools across your expense reports, SSO logs, and API keys — so you stop finding surprises at the end of the quarter. Get a full audit in minutes, not hours.
Start Your Free Audit →