Menu
Everything AI

Everything a small business needs to know about AI. In one place.

No hype, no jargon, no vendor pitch. This page is the source of truth we wish existed when our clients started asking about AI — so we wrote it.

01The numbers

AI adoption isn't coming. It's here.

Summary

Most small businesses now use AI somewhere, the ones using it well save meaningful hours every week, and the cost of starting has never been lower.

~80%of SMBs use AI in at least one workflow

The question has shifted from "should we?" to "where next?" Businesses not using AI now compete against a lower cost base.

5–10 hrssaved per person per week, typical

Teams using AI for drafting, admin, and triage consistently report hours back in the week — the gains compound as workflows connect.

<$100/moto start meaningfully

One assistant licence plus one automation tool covers a real pilot. The expensive part isn't software — it's picking the wrong workflow.

02Where AI works

AI is a judgment machine, not a maths machine.

Summary

Use AI where the work involves reading, writing, or deciding at volume. Skip it where there's one right answer a normal program can compute — that's cheaper, faster, and never hallucinates.

AI earns its keep on

  • Unstructured input at volume— reading emails, invoices, contracts, reviews, and doing something sensible with each one.
  • First drafts of everything— proposals, job ads, product descriptions, replies. Humans edit; AI starts.
  • After-hours coverage— phones answered, tickets triaged, appointments booked while you sleep.
  • Finding patterns in your data— which customers are about to churn, which quotes convert, what your reviews actually say.

Skip AI for

  • Deterministic logic— billing, tax, inventory counts. There's one right answer; use a formula.
  • High-stakes, low-volume decisions— a wrong answer costs more than the time saved.
  • Problems you haven't defined— AI amplifies a process. If the process is broken, it amplifies the mess.
  • Replacing accountability— AI can recommend; a named human should still decide anything a customer would want to appeal.

This is the same framework we use on paid engagements — it's why about 30% of our discovery calls end with “don't build this yet.”

03What it costs

Six categories cover 90% of small-business AI spend.

Summary

Start with an assistant, add automation when tools need connecting, and only consider custom builds when off-the-shelf demonstrably can't do the job.

AI assistants

$0 – $30 /user/mo

Claude, ChatGPT, Gemini. The universal starting point: writing, analysis, research. Buy one before anything else.

Workflow automation

$20 – $100+ /mo

Zapier, Make, n8n. Connects the software you already own and puts AI steps in between.

Marketing & content

$15 – $50 /user/mo

Canva, writing tools, SEO assistants. Useful, crowded, increasingly bundled into platforms you already pay for.

Voice & phone AI

$50 – $500 /mo

AI receptionists and call routing. One of the highest-ROI categories for service businesses.

Vertical tools

Varies widely

Industry-specific AI — legal drafting, dental scheduling, restaurant forecasting. Judge on workflow fit, not AI claims.

Custom systems

$15k+ engagements

Agents and integrations built for your exact operation. Only worth it when volume is high and off-the-shelf fails. This is what we do.

04Plain-English glossary

Eight terms that cover most conversations.

Summary

You don't need to know how AI works to use it well — but these eight terms will get you through any vendor meeting.

LLM
Large language model — the engine behind modern AI. Reads and writes text; the thing Claude, GPT, and Gemini all are underneath.
Agent
AI that takes actions, not just answers — it can look things up, use your software, and complete multi-step tasks with a goal.
Hallucination
When AI states something false with total confidence. Why every AI workflow needs either low stakes or a human check.
RAG
Retrieval-augmented generation — giving AI your documents to answer from, so it cites your price list instead of guessing.
Fine-tuning
Retraining a model on your data. Expensive and usually unnecessary — RAG or good prompting covers most business needs.
Token
The unit AI pricing is measured in — roughly three-quarters of a word. API costs are quoted per million of them.
Prompt
The instruction you give an AI. The difference between mediocre and excellent output is usually the prompt, not the model.
Human-in-the-loop
A workflow where AI does the work and a person approves it before it counts. The right default for anything customer-facing.
05Go deeper

Keep this page bookmarked. Then go deeper.

Summary

This page is the map. These are the roads: step-by-step instructions, current news, and honest tool reviews — all updated as the landscape moves.

Want the shortcut? Ask the people who build this stuff.

A thirty-minute call with a named partner beats a month of research. We'll tell you if you don't need us.

Get a quoteBook a call