Where small businesses actually are with AI
Ignore the survey headlines claiming everyone's already done it. Hard transaction data from the JPMorgan Chase Institute shows 17.7% of small businesses were paying for AI by the end of 2025 — up from 1.7% in 2019, with median spend around $30/month [JPMorganChase Institute, "Understanding the use of AI among small businesses," April 2026]. The U.S. Census Bureau finds overall business AI use at 17–20% in 2025–26, rising for firms with 20+ employees and flat for the smallest firms [Census BTOS, May 2026]. Translation: your competitors have opened a chat window; very few have working agents. The gap is wide open — for now.
The order we coach agents in — and why
This is the sequence we use in our own builds and coach clients through, because trust has to be earned by the system before you give it bigger jobs:
- Lead follow-up — highest payoff, easiest to verify, and the failure mode (a missed inquiry) is already happening today.
- Reporting — the Monday numbers assembled from your systems before you sit down. Zero customer risk while you learn to direct agents.
- Quoting and admin — drafts from your rules; you approve. This is where hours come back in bulk.
- Content and marketing — powerful, but do it after the operational wins; it needs the strongest approval gate. (See the 260-page site build for what's possible.)
The two ideas that keep agents safe
The approval gate: every agent that touches customers or money gets a human sign-off step, enforced by the system, not by good intentions. Ownership: agents run in accounts you control, with prompts and documentation you hold, so no vendor — including any coach — can lock up your operations. Every diagram in the workshop shows both.
What about "AI employees"?
The phrase sells a fantasy of unsupervised replacement — and unsupervised customer contact is the classic failure mode of small-business AI. The truthful frame: an agent is a power tool, not a hire. It does the recurring 80% and routes the judgment calls to you.