The Audit: How to Inventory What You Actually Use
Before you can constrain your tools to six, you need to know what you actually have. Not what you think you have — what you actually have. The gap between those two things is where the audit gets uncomfortable.
Most people, when asked to list their AI tools, will name five or six from memory. The ones they used this week, or the ones they're paying the most for, or the ones they talk about when someone asks "what AI tools do you use." This is not an inventory. This is a highlight reel. The real number is hiding in your browser's saved passwords, your credit card statement, your bookmarks bar, and the six tabs you haven't closed in three weeks because closing them feels like giving up.
How to Run the Audit
The audit takes 90 minutes if you're honest. Longer if you keep stopping to justify things. Here's the process.
Start with your credit card and bank statements from the last three months. Pull every recurring charge that's an AI tool, a productivity tool, or a SaaS subscription you use in conjunction with AI tools. This includes the obvious ones — Claude, ChatGPT, Midjourney — and the less obvious ones. The Notion subscription you use as a "knowledge base." The Zapier plan you set up for one automation that ran twice. The image editing tool you bought because the AI generation output needed cleanup. Write them all down. Every one.
Next, open your browser and check saved passwords. Sort by site. You will find tools you forgot you signed up for. I did this exercise last year and found active accounts on three AI writing tools, two image generators, and an audio tool I'd tested for one afternoon and never returned to. The free tiers count too — if it's installed, configured, or bookmarked, it's occupying a slot in your attention even if it's not on your credit card.
Then check your phone. AI apps accumulate on phones the way condiment packets accumulate in kitchen drawers. You downloaded them to try, used them once, and they're still there, sending you notifications you've trained yourself to ignore. Every app on your phone with an AI component goes on the list.
Finally — and this is the one people skip — check your desktop, your Downloads folder, and your terminal history if you use one. Local AI tools hide here. Ollama instances you forgot were running. Stable Diffusion checkpoints eating disk space. ComfyUI configs from that weekend you decided to learn image generation pipelines. If it's installed, it's on the list.
What the List Usually Looks Like
When I walk people through this process — through the Done-With-You sessions or just in conversation — the average number of AI-adjacent tools on the list is fourteen. The median is twelve. The range is six to twenty-three. The person who came in saying "I use about five tools" usually has eleven or twelve when the audit is done.
The list typically breaks into three tiers when you look at it honestly.
Tier one is the tools you actually use. These are the tools you opened in the last seven days to produce something. Not to test, not to browse, not to "check on" — to produce output that went somewhere. A draft you sent. An image you published. An automation that ran. For most people, this tier has two to four tools. Sometimes five. Rarely six.
Tier two is the tools you intend to use. These are the ones you're "getting set up" or "planning to integrate" or "waiting for the right project." You might have logged in this month. You might have watched a tutorial. But they haven't produced output yet, and honestly, the timeline for when they will is vague. This tier usually has three to five tools. These are the dangerous ones because they feel productive — you're investing in future capability. But the future keeps not arriving.
Tier three is the tools you forgot about. Active subscriptions you're not using. Free accounts you set up and abandoned. Browser extensions you installed. Mobile apps sitting in a folder. This tier is usually two to six tools, and it's the easiest to cut because the decision was already made — you stopped using them. The audit just makes the quiet abandonment official.
How to Score What You Find
Numbers don't lie, but they also don't interpret themselves. For each tool on your list, answer three questions.
When did I last use this to produce real output? Not "when did I last log in" or "when did I last test it" — when did it produce something that left my computer and went to another human or system. If the answer is more than two weeks ago, the tool is not earning its slot. Mark it accordingly.
What would break if I cancelled this today? This question separates the load-bearing tools from the aspirational ones. If cancelling Claude would mean you can't draft articles, that's load-bearing. If cancelling your second automation platform would mean... nothing changes because you haven't built anything in it... that's aspirational. Be ruthless here. "I might need it someday" is not a load-bearing answer.
Does this tool connect to anything else in my stack? A tool that works in isolation — where you manually copy output from it into another tool — is a friction point, not an asset. A tool that connects to your other tools through MCP, API, or automation is part of a system. Isolated tools are the first candidates for elimination because their value is limited by the manual effort required to use their output. If you're copying and pasting between tools, you don't have a stack. You have a collection.
The Emotional Part
The audit will make you feel something, and what it makes you feel tells you something useful.
If you feel relief — "oh good, I can finally cancel these" — your tool sprawl was a source of low-grade anxiety you'd been ignoring. The subscriptions you weren't using were a persistent reminder of projects you weren't finishing, and cutting them feels like clearing a debt.
If you feel resistance — "but I just haven't had time to learn it properly" — that's the sunk cost talking. The time you spent setting up the tool is gone. The subscription fee you've paid for three months is gone. Keeping the tool doesn't recover those costs. It just adds future costs to past ones. The tool's value is entirely forward-looking: will it produce output in the next 30 days? If you're not sure, the answer is no.
If you feel defensive — "you don't understand my workflow, I actually need all of these" — that's worth examining. You might be right. Some work genuinely requires a wide set of tools. But the defense usually comes from identity, not function. Your tool list is a proxy for who you are in the AI space. Cutting it down feels like shrinking. This is the hardest part of the audit, and it's where most people stall out. They'll do the inventory, categorize everything, score it honestly, and then find reasons to keep eleven of the fourteen tools on the list.
What to Do With the Results
Don't cut anything yet. The audit is step one of a three-step process — audit, eliminate, test. This article covers the audit. The next covers elimination. Cutting tools without a method leads to either cutting the wrong ones or not cutting enough.
What you should have after the audit is a complete, honest list of every AI tool in your life, scored by usage recency, load-bearing status, and integration capability. You should know, looking at the list, which tools are doing real work and which tools are doing nothing but draining attention and money.
You should also have a number. The total count. Write it down. You'll want to remember it in 30 days when your hex is running and you're producing more with six tools than you were with fourteen. The number is the "before" picture, and it's useful not for shame but for clarity. You weren't spending too much money on AI tools — you were spending too much attention. The money is easy to recover. The attention is the real cost, and you can't get it back.
One more thing. During the audit, you will be tempted to sign up for something new. You'll read about a tool while checking your list. You'll think "that would solve the gap between tool X and tool Y." Do not do this. The audit is a closed process. Nothing gets added until the elimination is done and the hex is set. The urge to add during an audit is the exact impulse the hex is designed to control.
A Note on Free Tiers
Free tools count. The "it's free so it doesn't matter" rationalization is how tool sprawl starts. A tool that costs zero dollars still costs attention. It still takes up space in your browser. It still generates notifications. It still represents an option you have to evaluate every time you sit down to do a task. Free tiers are not free — they're paid for in cognitive overhead, and that overhead compounds with every tool on the list.
Some of the hardest audit conversations I've had are with people who have eighteen free-tier AI tools and no paid ones. They're spending zero dollars and losing hours every week to tool-switching, option paralysis, and the background hum of tools they're not using but haven't officially abandoned. The hex doesn't distinguish between paid and free. If it's in your stack, it needs to earn its slot.
This article is part of The Hex System series at CustomClanker.
Previous: The Hex Constraint Explained | Next: The Elimination Method