How-to guideFree templateAdded 6 Oct 2026

See which AI tools you pay for and what each one is for

List every AI tool the team uses, the job it does, who looks after it and what data it sees, then decide what to keep, merge or stop.

Use it when People have signed up for AI apps on their own, and nobody can say for sure which ones are still used, who pays for them or what they can see.

Download template

It is a simple text file. Open it in any text editor, or paste it into ChatGPT, Claude or Gemini and ask for help filling it in.

How to use it

Gather every AI tool from the bills and by asking the team. Include free tools people use for work and AI features switched on inside apps you already pay for.
Fill one row per tool: its job, its owner, what data it can see and where its output ends up. Leave a blank wherever you do not know, rather than guessing.
Mark the flags, then give each tool one decision. Cancelling, changing access or moving data stays with the named owner and the person who pays.

What you will need before you start

Access to the bills, card statements or admin pages where AI tools are paid for
A few minutes with each person who uses an AI tool for work, to ask what they use it for
Someone who can decide what to keep, change or stop, and who knows who pays for each tool

The template

Fill in the blanks for your business. This is what you get when you download or copy it.

Audit header

Organisation or team:

Audit owner:

Date:

Where the list came from (bills, card statements, admin pages, asking the team):

Tool list

Fill one block per tool. Copy costs and renewal dates from the bill. If nobody knows, leave it blank.

Tool name (or a plain description, such as "meeting note-taker"):
Who signed up and who pays:
Plan and renewal date:
Who uses it, and roughly how often (ask them; do not guess):
The job it does, in one line (the piece of work it helps with):
Owner (the person who answers for it):
What data it can see (client files, customer messages, money, staff or personal details):
Where its output ends up (saved where the team can find it, or left in someone’s chat history):

Flags

Mark every flag that applies to each tool.

No clear job: nobody can name the work it helps with
Doubled up: another tool already does the same job
No owner: nobody answers for it
Sees client or personal data with no agreed rule
Output is not saved anywhere the team can find it
Paid for, but nobody asked says they use it

Decision for each tool

Give each tool exactly one decision.

Keep: it has a job, an owner and a clear data rule
Keep with a change: name the change (an owner, a data rule, where output is saved)
Merge: move its job into another tool you already use; name which one
Stop at renewal: note what needs exporting first and who will do it
Need more information: say what is missing and set a date to decide

Before you add the next tool

Answer these five questions in writing first. If any answer is blank, treat the new tool as a trial, not a decision.

Which piece of work will it help with?
What will it be allowed to do, and what will it never do?
Who decides whether its output is good enough?
What context has to travel with its output so a teammate can pick it up?
Where will accepted output be saved so nobody has to redo the same thinking?

Stop rules

Do not cancel, downgrade or change anyone’s access until the owner and the person who pays have agreed. Export anything you need first.
Do not invent costs, usage or savings. Copy figures from bills or leave them blank.
This list does not show that a tool is secure or that its terms suit your business. Where a tool sees client data, personal details or contracts, the named owner checks its terms before anything changes.
If a tool sees client or personal data and nobody can name the rule for it, raise it with the owner the same day.

See an example

A made-up example, so you can see what a filled-in version looks like.

This is an illustrative scenario, not a client case study or a measured result. Do not reuse its details as facts about your work.

A sample seven-person recruitment consultancy in Bengaluru fills mid-level roles for startups. Over the past year, consultants signed up for AI tools on their own. The founder sees several AI charges on the company card each month and cannot say which ones are still used or what they can see.

Sample inputs

Demonstration inputs only. Tools are described by type, not by brand, and amounts are left out on purpose. From the card statement: two separate paid accounts for the same general chat assistant, one meeting note-taker, one CV-screening add-on. From asking the team: the founder also uses a free writing extension in the browser for job adverts. The note-taker joins candidate calls and keeps the recordings. The CV-screening add-on started as a trial and moved to a paid plan; one consultant says they tried it once.

Example output

Audit header

[Fact within this demonstration] Audit owner: [Name the founder]. List built from the company card statement and short chats with all seven people.

Tool list

[Fact within this demonstration] General chat assistant: two consultants pay separately; both use it to draft candidate summaries for clients. Owner: none named. Meeting note-taker: used by three consultants on candidate calls; recordings and notes stay inside the tool. Owner: none named. CV-screening add-on: one consultant tried it once. Renewal date: [copy from the bill]. Writing extension: free; used by the founder for job adverts; it can read whatever is open in the browser.

Flags

Chat assistant: doubled up; no owner; sees candidate details with no agreed rule. Note-taker: no owner; sees personal data with no agreed rule; output not saved where the team can find it. CV-screening add-on: paid for, but nobody asked says they use it. Writing extension: sees client or personal data with no agreed rule, since it can read the open screen. [Inference] Candidate personal details are the common thread across most flags. How much each tool costs has not been checked in this demonstration.

Decisions

[Recommendation] Chat assistant: merge into one team account with a named owner and a written rule on which candidate details may go in. Note-taker: keep with a change; name an owner, agree whether candidates are told about recordings, and save accepted notes to the shared candidate file. CV-screening add-on: stop at renewal, after the consultant confirms nothing needs exporting. Writing extension: need more information; the founder checks what it can read before using it near candidate or client files. These recommendations are not the consultancy’s decision.

Stop rules applied

Nothing was cancelled or changed during the audit. Candidate data questions stay with the named founder, and the note-taker’s recording terms are checked before any change.

How to tell it is done well

Every tool has a named job or a No clear job flag, an owner, the data it can see and one decision; costs, renewal dates and usage come from bills or the people asked, with blanks left visible; nothing was cancelled or changed before the owner agreed.

Where a person should decide

The audit shows what you use and why; it does not prove that a tool is secure, compliant or good value, and it does not check vendor terms. A filled list is not approval to cancel, buy or change anyone’s access. Keep consequential actions with the named human owner.

Edit your own copy on this site

About this version

New in Kosh at version 1.1. Provider-neutral Markdown instructions; no installation or platform compatibility is implied. Runtime behaviour has not been verified across AI providers. Reuse terms have not yet been published. Background reading: Before the Next AI Subscription, Trace the Operating Thread. Checking which tools support which work, before you add more, is part of Kramaniti’s Foundation Strategy audit.

How we write Kosh templates