Pick the Four Tools You Keep and Retire the Rest

Nobody set out to have fifteen tools

Every piece of software in your business got there for a good reason. Somebody had a problem on a Tuesday, found something that fixed it by Thursday, and got back to work. Do that a few dozen times over four years and you end up with a stack nobody sat down and picked.

I want to say plainly that this is not a discipline problem. It is what happens when people are trying to serve customers and the software is not the point of the job. The person who signed up for that scheduling app was solving a real thing that week.

The cost is not really the subscriptions. The cost is that the answer to a customer question now lives in three places and a person has to go check all three before they can reply. That is the part that slows you down, and that is the part AI cannot fix on top of.

The four buckets that cover most of it

When Tommy and I look at a small team's stack, we are usually sorting everything into four buckets. Most businesses with 5 to 50 people need one clear winner in each and not much else.

1. Where the work lives. Jobs, projects, tickets, whatever you call the thing that moves from open to done.
2. Where the customer lives. Name, contact info, history, what you sold them. A CRM, or the closest thing you have to one.
3. Where the files live. Contracts, photos, drawings, invoices, the stuff you have to find again in eight months.
4. Where the conversation lives. Internal chat and email, and the record of what you told the customer.

Pick one winner per bucket. Not the one with the best features. The one your team already opens without being reminded, because adoption is the whole ballgame and you cannot argue people into a tool they do not like.

How to run the count on a Friday afternoon

Start with the card statement, not with memory. Pull the last two months and write down every recurring software charge. Then add the free tools, because those are usually where the shadow processes are hiding. A free form builder, a personal notes app, a spreadsheet somebody keeps on their desktop.

Next to each one, write two things. Which of the four buckets it serves, and the name of the last person who opened it. If you do not know who opened it last, that is your answer already.

Now imagine what that list looks like for a hypothetical 14 person services company. Say they land on fifteen tools total. Four are the real ones people use every day. Three are duplicates of those four that somebody signed up for during a busy season. Two do something the main tools already do, just slightly differently. Three are one person's personal habit that never spread. Three nobody has logged into since last year and the card keeps getting charged. I am making those numbers up to show the shape of it, but that shape is very common.

So you are usually not deciding between fifteen tools. You are confirming four and figuring out what to do with the rest.

Do not delete anything on Friday

Here is where I would slow you down, because the satisfying move is to cancel eleven subscriptions in one sitting and feel great about it. What we do instead is freeze rather than delete.

Give it thirty days. Turn off new work going into the tool, tell the team where the new home is, and leave the data where it sits. If nobody misses it in thirty days, export what matters and cancel. If somebody does miss it, you just learned that the tool was carrying a real process you did not know about, and now you can move that process on purpose instead of breaking it.

The one thing worth doing immediately is stopping the charges on the tools nobody has logged into in a year. Those are not a decision. Those are just cleanup.

Where AI comes in, and it is later than you think

The reason we push tool consolidation before anything else is that most of the AI ideas people bring us are really data location problems in a costume. You want a draft of a customer update, but the job status is in one system, the last three emails are in another, and the photos are in someone's phone. No prompt fixes that. The workflow around the tool is the work.

Once you are down to four systems, the useful AI step gets small and boring in a good way. For example, a weekly draft that pulls open jobs from one place and recent customer notes from another and writes a plain summary for you to send. You read it, you fix the two things that are wrong, you send it. It is a draft, not a decision, and a person signs off every time.

That step is buildable when the information has four homes. It is guesswork when it has fifteen.

What I would do this week

Pull the statement. Sort into the four buckets. Name your winner in each one. That is it. Do not migrate anything yet, do not buy anything, and do not let anyone sell you an assistant that sits on top of a stack you cannot describe.

If you get through the sorting and two buckets have a tie you cannot break, that is usually a sign the two tools are serving two different teams who never agreed on a process. That is worth a conversation before it is worth a purchase, and it is the kind of thing Tommy and I are happy to walk through with you.