Best AI Tools For Business Owners: The Stack I Actually Run

---
name: tools-i-run
description: Audits a business's software stack the way an operator does: groups every paid tool by the job it does, finds duplicates and empty jobs, asks for the human owner and the one metric behind each tool, runs the four part keep or kill test, and returns a keep, merge or kill table in install order. Trigger on "audit my stack", "what tools should I keep", "am I paying for too many tools", "should I buy this tool".
---

# The Tools I Run And Why

A tool stays in a business only if it replaces a role or returns time you can measure.
Everything else is a subscription with a dashboard. You audit the list by job, not by
category, and you never recommend a tool the owner did not mention unless a job is
empty.

## Inputs

Every tool the owner pays for, with monthly cost and last login if known. A card
statement or billing export is fine. If the owner cannot produce the list, the first
finding is that the list does not exist.

## Step 1: Group by job

Put every tool under exactly one job:

1. **Catch the lead.** CRM, comment to DM, ads, forms. Where every lead lands.
2. **Convert the lead.** Setter, nurture, booking, conversion video.
3. **Make the content.** Record, dictate, ideate, edit, generate, publish.
4. **Think and remember.** Models, notebooks, the vault the agents read, the source of
   truth for client status.
5. **Research.** Scrapers, clip finders, outlier tools.
6. **Other.** Anything that does not fit. Ask what job it does. If the owner cannot say,
   that is the answer.

Name every job with two or more tools. Name every job with none. The CRM is the center
of gravity; if there is no CRM, install order starts there.

## Step 2: Owner and metric, one tool at a time

For each tool ask:
- Who is the human that owns it: checks quality, catches failures, decides what happens
  when it misses?
- What is the one number that proves it is working?

"Unknown" is an accepted answer. Mark it. Do not fill it in for them.

## Step 3: The test

For each tool, with a number where one exists:
- Does it make money?
- Does it return meaningful time?
- Does it protect quality?

If none of the three, it does not get to stay. "Cheaper than the old way" is not a
result.

For a tool the owner is considering buying, refuse to continue until they write four
things and each is specific:
1. The baseline: what it costs now in money, time, rework and quality.
2. The metric that proves it worked.
3. The human owner.
4. The review dates, 14 and 30 days, with keep, fix or kill.
Then name the existing tool it replaces, because it must replace one.

## Step 4: The output

A table in install order (catch, convert, content, think, research):

| Tool | Job | Owner | Metric | Verdict | Reason |

Verdict is keep, merge or kill. Reason is one sentence. Below the table:
- Monthly saving from the kills.
- The one tool to add next if a job has nobody doing it, and only then.
- Any fact that lives in two tools, with which one should be the home. One home per
  fact.

## Step 5: The stack file

Return the table as a file the owner saves next to their company operating file, so
every agent reads the same list. Columns: tool, job, owner, metric, baseline, review
date, status. Anything past its review date without a number in the metric column goes
on the next review list automatically.

## Rules

- One tool per job.
- Install order: CRM first, content last.
- A human owns every tool. AI is not accountable to itself.
- Never recommend a tool the owner did not mention unless a job is empty.
- Never invent a number. Unknown stays unknown until the owner measures it.