The Big Story: The AI Gap Between Small Businesses Is Widening, Not Closing

New research from Pax8, a marketplace that sells cloud and AI tools to small businesses, found that 61% of small businesses are now actively using AI, and 71% of those AI users say the tools let them compete with much bigger companies (Pax8 Q2 2026 SMB AI Pulse Report). But the same report found nearly one in three AI-using businesses are stuck in “experimentation” — they’ve started, but can’t move from testing into everyday use. What separates the ones pulling ahead? Not the tools themselves. It’s leadership alignment (91% of active AI users say their leadership agrees on AI’s role, versus just 32% among non-users) and having some kind of written AI policy or guideline (two-thirds of active users have one, versus 23% of all small businesses). Meanwhile, 13% of small businesses have slowed or paused their tech spending altogether — even though half of those same businesses admit they won’t stay competitive in three years without AI. The tools aren’t the hard part anymore. Deciding to actually use them, on purpose, with some structure, is.

Warren’s Take: This tracks with what I see with clients. The businesses that get stuck aren’t missing a better chatbot — they’re missing a plan. If your team can’t answer “who owns this and what’s it for,” you’ll stay in test mode indefinitely, no matter how good the tool is.

Story #2: Google Limits Meta’s Access to Its AI Model Over a Compute Shortage

Google reportedly capped how much of its Gemini AI model Meta could use this month, after demand for AI computing power outpaced what’s actually available (AI World Today). This isn’t really a story about Google and Meta’s business relationship — it’s a reminder that the computing power behind AI tools is genuinely scarce, and scarcity tends to mean rationing, price hikes, and outages, not less of them.

Warren’s Take: If your organization built a workflow that only works with one AI provider, this is your nudge to know your backup plan. It doesn’t have to be a second subscription today — just know what you’d switch to if your main tool got slower, pricier, or unavailable for a stretch.

Story #3: AI Grant-Writing Tools Are Getting Real, But None of Them Do It All

A new comparison of AI grant-writing tools found that no single tool covers the whole job (Grantable). Purpose-built tools like Grantable pair AI drafting with funder research pulled from nonprofits’ own tax filings. General chatbots like ChatGPT and Claude work fine for drafting once you’ve already done your own research. Dedicated funder-search tools like Instrumentl and Foundation Directory Online are built only for finding the right grants, not writing them.

Warren’s Take: Don’t pay for a specialized tool to replace a job a free chatbot can already do for you. Figure out which step of grant writing actually eats your time — drafting, or finding the right funders — before you pick a tool to fix it.

Practical Tip of the Week

Before you add a new AI tool to your nonprofit or small business, write down one sentence: “This tool is for ___, owned by ___, and we’ll know it’s working if ___.” That’s it. Per this week’s Pax8 data, having any written guideline — even a one-line one — puts you ahead of most small businesses still without one.

By The Numbers

61% of small businesses are actively using AI this quarter, roughly flat with last quarter’s 62% — but the gap between users and non-users is widening underneath that flat number (Pax8).

$0.46 vs. $4.18 — the average cost of an AI-resolved customer support ticket compared to a human-handled one, roughly a 9x difference (Taylance Tech, citing PwC AI Agent Survey data).

~5 months — the median time it takes a small business to earn back what it spent deploying an AI agent, according to 2026 industry data (Taylance Tech).