Twelve AI tools, three that pay back: the SME stack audit

If your last attempt at AI was disappointing, there is a fair chance you are paying for that disappointment monthly.

The typical UK small business with five to fifteen staff now carries somewhere between eight and fourteen AI-related subscriptions. ChatGPT Plus on the owner's card. A team Slack with AI features nobody asked for. Notion AI on the operations side. A meeting transcription tool that was brilliant for the first two weeks. A "free trial" CRM that auto-renewed. Two image generation tools. A grammar checker.

Each feels affordable on its own. Combined, the bill sits between £200 and £600 a month - between £2,400 and £7,200 a year - on tools where, if you asked the team to name what they paid back this quarter, the room would go quiet.

This post does what most owners have never sat down to do: a structured audit of the stack, with a simple test for each tool and a clear view of which categories pay back and which do not.

The three questions that decide every tool's fate

Question 1: Does anyone use it every working week?

Not "has anyone used it in the last month". Every working week, as a matter of routine. If the answer is no for more than three consecutive weeks, the tool is shelfware. Something that is hard to remember to use is not saving anyone time.

Question 2: What does it replace?

"It makes us more productive" is not an answer. "It replaces the 45 minutes I used to spend writing meeting notes" is. If you cannot name the specific task the tool replaces, it is probably filling a gap that did not exist rather than solving a problem that does.

Question 3: Keep, cut, or trial first?

Keep: passes questions 1 and 2, cost is justified. Cut: fails either question. Trial first: potential but nobody has properly tested it - give it one month with a named person responsible, then decide.

A 7-person consultancy that went through this exercise with us had 11 AI-related subscriptions at £427 a month. After the audit, they cut 6, kept 5, and saved £223 a month - £2,676 a year. None of the six cut tools were ones anybody protested about. They had all been quietly abandoned.

Run your own version with the free AI tool stack audit. Four minutes, per-tool verdict, net annual position.

The three categories that almost always pay back

One general-purpose chat tool - and only one

ChatGPT (Plus or Team) or Claude (Pro or Team) - pick one and use it consistently. These tools pay back on writing tasks above everything else: client proposals, email replies, onboarding documents, training notes. The output needs editing, but the time saving on a repetitive writing task is 30 to 60 per cent.

The failure mode: paying for both ChatGPT Plus and Claude Pro simultaneously while using neither habitually. They are close enough in everyday quality that one well-used subscription outperforms two neglected ones. If you are trying to choose, the Claude Projects vs Custom GPTs comparison covers the practical differences.

One meeting transcription tool

If you take five or more client calls a week, a transcription tool is a fast two-hour weekly saving. Otter.ai, Fireflies.ai, and Zoom Pro (transcripts included in the paid subscription) all do the same core job: record, transcribe, summarise. A 60-minute call no longer requires 20 minutes of note-writing.

If you already have Zoom, upgrading to Zoom Pro is the simplest route - transcripts are included. If you are on a free Zoom plan, compare Otter.ai and Fireflies.ai on their free tiers for a month before paying for either.

One workflow automation tool

Make or Zapier. The value is in connecting tools you already use so one action triggers another without manual copying. New enquiry creates a CRM contact. Invoice paid triggers a project setup. These workflows are not impressive to describe but they remove work permanently once they are set up.

One caveat: automation is only worth setting up on a process that already works. Automating a broken workflow makes the breakage happen faster. Fix the process first, then automate it.

The categories that usually do not pay back at this size

Custom image generators. Midjourney and similar tools are genuinely impressive. For a business producing visual content at volume, the subscription is justified. For most service businesses, the occasional social post does not cover the monthly cost or the learning curve.

AI grammar checkers as standalone subscriptions. Grammarly Premium produces marginal improvements over the grammar checking already in Google Docs, Word, or the general-purpose chat tool you are already paying for.

Dedicated AI sales assistants. At £150-£400 a month, these are rarely justified at five to fifteen staff. The quality of AI-generated outreach at this tier is not reliably high enough to represent the business well.

Specialised research tools. The output requires so much human correction that the time saving is minimal. A well-constructed prompt in a general-purpose chat tool handles most research needs adequately.

Audit before you add

The most expensive move in AI adoption is adding tools before auditing what you already have. The businesses that get the best return from AI are not the ones with the most tools. They are the ones with three well-chosen tools that are used every week, configured properly, and connected to the processes they already run.

Before adding anything new, run the three questions above on everything you are already paying for. Or use the free AI tool stack audit for a structured verdict with a clear net annual position.

If AI still feels like a disappointment, the AI fatigue post covers what to focus on for the next six months and what to safely ignore. And if you want a complete view of which tools in your specific business are worth investing in properly, a HoursBack Assessment reviews your current stack, identifies the two or three tools worth configuring well, and gives you the setup instructions to make them part of your weekly workflow.

Book your HoursBack Assessment - £799