The Small Business AI Stack — 5 Tools, Zero Hype, Real Use Cases
The BCG 10–20–70 rule applied to a real sub-ten-person operation. Five tools, what each one actually does, and why the tools were never the hard part.
Boston Consulting Group's research on AI implementation keeps landing on the same ratio: 10% of the value comes from the algorithm itself, 20% from the technology and data around it, and a full 70% from the people and process changes that make the tool actually stick. Most small operators build their AI stack backwards — obsessing over which tool, ignoring the 70% that determines whether any tool survives past week two.
Here's what the 10–20–70 rule looks like applied to a real sub-ten-person operation, tool by tool.
The 10–20–70, Translated
The 10% — the model itself — barely matters as much as marketing suggests. Most mainstream tools are good enough for most small-business tasks. The 20% — the tech and data layer — is where you make sure the tool actually connects to your real workflow instead of living in a browser tab you forget to open. The 70% — people and process — is whether someone on your team actually knows when to use it, when not to, and who's responsible for checking the output. Skip the 70% and the best tool in the world becomes an unused subscription by month two.
The Five Tools
1. A general assistant (ChatGPT or Claude). Drafting, editing, brainstorming, first-pass research. This is the one tool that touches nearly every role in a small operation, which makes it the highest-leverage subscription in the stack.
2. A scheduling assistant (Calendly + AI features, or a tool like Motion). Removes the back-and-forth of finding meeting times and, in the better tools, actively reshuffles your calendar around priority work instead of just booking slots.
3. A transcription and meeting-notes tool (Otter or Fireflies). Every client call, every internal meeting, captured and summarized without someone manually taking notes. The real value isn't the transcript — it's the searchable record you can query later.
4. An automation layer (Zapier or Make). Connects the other tools so information moves without someone manually copying it between apps. This is the "20% tech" layer doing its actual job — making the stack behave like one system instead of four disconnected subscriptions.
5. A finance/admin assistant (QuickBooks AI features or a dedicated bookkeeping AI). Categorization, anomaly flagging, invoice chasing. Unglamorous, and consistently one of the highest-ROI uses of AI in a small operation because the task volume is high and the judgment required is low.
The stack isn't five tools. It's five tools plus one person who owns making sure they're actually used.
Where The 70% Actually Lives
Every tool on this list fails the same way in a small operation: someone signs up during a burst of enthusiasm, uses it hard for two weeks, and quietly stops when a task comes up that doesn't fit the tool's obvious use case. The fix isn't a better tool. It's a five-minute weekly check-in where someone asks, out loud, "what did we actually use this for this week, and what got skipped?" That habit, not the subscription, is the 70%.
Assign one person per tool to own that question. Not to be the only user — to be the one who notices when usage quietly drops to zero and asks why before the renewal charge shows up for a tool nobody opened in six weeks.
Zero Hype, Real Use Cases
None of these five tools is exotic. None requires an engineering hire or a six-figure budget. What they require is the unglamorous 70% — someone deciding when they get used, checking that they're actually saving the time they're supposed to, and killing the ones that quietly aren't. That discipline, more than any tool on the list, is the actual stack.
Audit Your Own Stack
List every AI or automation tool your business currently pays for. Next to each, write down what it was actually used for this week — not what it's supposed to do. Anything with a blank next to it is a candidate to cut or to assign a real owner to before the next renewal.
