You probably don't need another AI tool
By Sam Kim · · 6 min read

Should my business buy a separate AI tool, or use what we already have?
Check what you already pay for first. Email, accounting, practice management and office suites increasingly carry AI features, and many sit unused because nobody switched them on, they are on a plan tier you haven't looked at, or nobody knows they exist. Try the built-in option on one real task. Buy something separate only when the job crosses systems nothing you own covers, the built-in version fails on your actual work, or it can't meet your data-handling requirements.
Why does buying a new tool feel like progress?
Because it's the one step that looks finished. You sign up, you get a login, there's a dashboard, and for a week it feels like something has changed. Then the tool sits next to the old way of doing things instead of replacing it, and the team keeps working the way they always have.
This is the same mistake as believing you need the newest tool. The tool is rarely what's missing. What's missing is a clear task, a written process and someone responsible for it. A new subscription supplies none of those.
There's a cost on the other side as well. Every extra tool is another login to manage, another place your clients' information goes, another bill and another thing to switch off when someone leaves.
How do you check what you already have?
Work from your bank statement, not your memory. An afternoon is usually enough.
- List every software subscription. Pull twelve months of card and account statements and write down every recurring software charge, including the small ones somebody put on a personal card and claims back.
- Note the plan each one is on. Then compare that against the vendor's current plan page. AI features are often tied to a particular tier, and the tier you signed up on may have changed shape since.
- Open the admin settings. Look for anything labelled assistant, AI, drafting, summarise or suggestions. Many business accounts ship with these switched off until an administrator turns them on.
- Read the vendor's release notes for the past year. This is the most reliable record of what was added, and it's written by the vendor rather than a reviewer.
- Ask their support. One email: "What AI features does our current plan include, and what are the data-handling terms for them?" The second half of that question matters as much as the first.
Keep the result on one page: system, plan, AI features available, switched on or off, terms read or not. It's a small document, and most owners have never seen one for their own business.
How do you tell whether a built-in feature is good enough?
Try it on one real task for two weeks. Pick something that passes the three tests from what AI won't do: it happens at least weekly, you could explain it to a new hire on one page, and a wrong answer is cheap to catch before it leaves the building.
Then judge it on your work, not on a demo. Did the drafts need light edits or a rewrite? Did it save time after the checking, or only before it? Would the person doing the task keep using it if you stopped asking? If you want to be more rigorous than a gut feel, take a baseline first and measure whether it worked.
Good enough is a real answer. A built-in feature that does eighty per cent of the job, inside a system your team already opens every day, will usually beat a better standalone tool that nobody remembers to open.
When is a separate tool justified?
When you can name the gap it fills. These are the cases where I'd say yes:
- The job crosses systems and none of them owns it. An enquiry that arrives by email, needs checking against the booking system and ends up as a quote in accounting sits between three tools. None of their built-in features will see the whole flow.
- The built-in version failed on your actual work. Not in theory. You ran the two-week test and it couldn't do it.
- It can't meet your data-handling requirements. If a feature processes or stores information somewhere your client agreements or obligations don't allow, that's a reason to look elsewhere, or to not use AI for that task at all.
- It replaces something rather than adding to it. A new tool that lets you cancel two others is a different decision from one that sits on top.
- The arithmetic works. Sometimes the feature you want is only on a higher tier of a product priced per seat, and a separate tool for the two people who need it costs less. Do the sum on current prices rather than assuming either way.
If none of those applies, the honest answer is that you don't need it yet. The build, buy or hire decision goes through the same logic in more depth.
What does an extra tool really cost?
More than the subscription. Add these up for any tool you're considering:
- Fees. Monthly price times seats times twelve, at the vendor's current rates.
- Setup time. The hours to configure it, connect it and show the team, priced at what those hours are worth to you.
- Ongoing ownership. Someone has to manage users, updates and the occasional thing that breaks. If you can't name that person, it's you.
- Another place data goes. Every tool that touches personal information may need to appear in your privacy policy, including where the provider stores it.
- The exit. How you get your data out if you stop paying, and what happens to anything built on top of it.
The last two are the ones owners skip. They're also the ones that get expensive later.
Where does a person stay in the loop?
In exactly the same place as with any other AI. Built-in features draft, summarise and suggest in the same confident tone as a standalone tool, and they know nothing about your business you haven't told them. A person still reads anything that goes to a client, and still makes any decision about money, dates or obligations. Buying the feature from a vendor you trust doesn't change that.
This is general information rather than advice for your circumstances. Before you switch on an AI feature that will handle client or staff information, check its terms and confirm your own privacy and confidentiality obligations.
What to do next
If the gap is real, the free AI subscription chooser suggests where to start, beginning with what you already pay for.
- Pull twelve months of statements and list every software subscription, with the plan each one is on.
- Check the admin settings and release notes of your three most-used systems for AI features you aren't using.
- Pick one weekly task and run the two-week test with whatever you already have.
- Only then write down the gap, if there is one, and look for a tool that fills that gap and nothing else.
If you'd rather someone else did the inventory and the testing, that's what the AI Tools Assessment covers.
General information, not legal, tax or financial advice · Drafted with AI tools to my written rules · Terms