Start with one piece of work.
An AI inventory becomes useful when you can connect the subscription to something the business needs. Start with a repeated workflow: following up a sales call, handling a customer question or preparing an internal report.
Write down its trigger, the systems it touches, the person responsible and what “done” means. Follow an actual example from beginning to end. You may discover that the expensive part is the handoff between otherwise useful tools.
“Produces a summary” describes an output.
“The right person completes the follow-up” describes an outcome.
Build an inventory people can maintain.
Use billing records, your software directory and conversations with department owners. Include AI features inside existing products as well as standalone subscriptions. Ask employees what they actually use; an invoice alone does not establish adoption.
Record the tool, owner, business purpose, paid seats, active use, billing period, renewal date, connected systems and data involved. Mark unknowns as unknown rather than assuming a tool is unused.
Your AI tool inventory.
A blank CSV worksheet you can open in Excel or Google Sheets. Fill it in internally; no upload is needed.
Download the worksheetUse actual invoices and currency. Separate committed annual charges from cancellable monthly spend. Keep client-paid costs and usage charges visible so your totals do not imply savings you cannot realize.
Ask the people closest to the tool.
For each tool, ask its owner three questions:
- What real task did this help you complete recently?
- Where did its output go, and what happened next?
- What would break or take longer if we removed it?
A tool used only once a month can still be important. A tool opened every day can still create work that someone else has to redo. Capture both frequency and usefulness.
Do not collect passwords or API keys in the inventory. Record who manages access and where the approved access process lives.
Look for overlap and missing connections.
Group tools by the job they do rather than their marketing category. Two meeting assistants may overlap; a meeting assistant and a CRM may be complementary. Compare the same task, with comparable inputs and quality requirements.
Check differences that matter: access permissions, data handling, required integrations, export options, accessibility and a team’s specialist needs. Avoid equating feature similarity with interchangeability.
Also look for repeated manual steps. If a team downloads a report, pastes it into another tool and sends the same update by email, the first improvement may be a connection rather than a cancellation.
Turn the inventory into a short decision list.
Give each item a next action: keep, investigate a connection, evaluate a replacement or review for retirement. Add an owner and a review date. Treat that as a decision queue, not permission to cancel anything.
Start with one low-risk workflow whose baseline you can measure. Record its current completion time, manual touches and common errors. Change it, test it, then compare the result over a representative period.
The free Sprawl Audit can help frame the conversation. It uses the information you enter; it cannot see your usage, contracts or dependencies. Bring those facts into the decision before making changes.
A practical framework from the team behind AI Sprawl Cleanup. Examples are illustrative, not customer results.
Meet the team →