OC-206 / Small business guide

Q3 2026 — current edition

The Small Business Guide to AI Education and Adoption.

Everyone in your company is already touching AI. Almost no one is using it strategically. This quarter's edition: where small businesses actually stand, what changed in the last ninety days, and the first moves that pay for themselves.

OC-206 — The small business guide · First edition

Free · Updated quarterly · Published by OceanCrest AI

Corrected through 2026-08-18

How to read this chart

OC-206 / Sheet 02

Start anywhere. We did the ordering for you.

You don't need another article about robots. You run a business — payroll is real, customers are real, and your week is already full. So this guide is built the way a good working session runs: broad before specific, plain language before technical language, and every claim tied to a source you can check on the last page.

If you read only three pages, read the quarter at a glance, the three lenses, and your next ninety days. The technical material sits near the back on purpose — it's there when you're ready, not in your way when you're not.

A new edition ships every quarter. The spine stays the same; the pages marked Q3 below are rewritten each quarter as the facts change.

Contents — seventeen sheets, one course

03The quarter at a glanceFour numbers that set the pictureQ3
04Where businesses actually areAdoption is broad; maturity is rare
05The gap — and the spectrumWherever you are, that's the starting point
06Money · Energy · TimeThe only three reasons to do any of this
07–09Where it landsFive departments, concretely
10Adoption → Integration → AmplificationThe course we chart for every client
11Why education comes firstYour people are already using it
12Guardrails for small firmsA one-page policy beats no policyQ3
13–14This quarter in AIWhat changed, and what it means for youQ3
15For the technical readerPrompts, routines, agents — demystifiedQ3
16Your next 90 daysAnd a free diagnostic to start from
17About OceanCrest · SourcesEvery figure, checked or markedQ3

Legend — PA, position approximate: a nautical chart marks a hazard it believes but hasn't confirmed. We do the same. A figure carrying ^PA is our best current reading, not yet verified at its source — we mark it rather than pretend.

The quarter at a glance

OC-206 / Sheet 03 · Q3 2026

Four readings that set the picture.

Taken this quarter, from sources listed on sheet 17. From the next edition on, each carries its change since the previous quarter.

01

58%

of U.S. small businesses now use generative AI — up from 40% a year earlier, and more than double two years ago. If you're reading this, you're probably one of them.

U.S. Chamber of Commerce, Empowering Small Business, 2025 · S01

02

1%

of leaders call their company's AI use mature. Adoption is everywhere; AI woven into real workflows is still rare — that gap is this guide's whole subject, and your open lane.

McKinsey, Superagency in the Workplace, 2025 · S02

03

½ the price

Near-frontier AI roughly halved in price this quarter — and AI seats are now bundled into the office suites you already pay for, from $14 a seat. The cost excuse retired this summer.

Anthropic · Microsoft · Google, official releases, Jun–Aug 2026 · S19

04

80%

of AI users at small and mid-sized companies bring their own AI tools to work — while only 22% of employees say their organization has communicated a clear AI plan. Your people are ahead of your policy.

Microsoft/LinkedIn Work Trend Index 2024 · Gallup 2025 · S13, S14

Soundings

OC-206 / Sheet 04 · Soundings

Adoption is everywhere. Maturity is rare.

Two things are true at once: most businesses now use AI somewhere, and almost none would call their use of it mature. That gap — between touching the tool and running on it — is the whole story of this guide.

Organizations using AI somewhere

88%

Regular use in at least one business function.

Qualify as AI high performers

~6%

Attribute 5%+ of earnings impact to AI.

Call their companies "mature" on AI

1%

Leaders' own description of where they are.

Large-organization surveys — we're honest about that; the small-business read is below. S02, S03 · sheet 17

Small business specifically.

Among U.S. small-business leaders, 30% call themselves early adopters, 54% careful evaluators — experimenting selectively and measuring before expanding — and 21% prefer wait-and-see. (Self-chosen categories that overlap, so they sum past 100.) The picture is unmistakable either way: the middle of the market is evaluating, not ignoring.

Business.com, 2026 Small Business AI Outlook, n=1,009 · S05, sheet 17

95%

of enterprise AI pilots showed no measurable P&L return in MIT's preliminary 2025 research — overwhelmingly people-and-process failures, not technology failures.

82%

of AI-using small businesses grew their workforce over the past year. Adoption and hiring are moving together, not against each other.

The spectrum

OC-206 / Sheet 05

You're already using AI. At about one percent power.

Someone on your team planned a vacation with Claude this month. Someone else drafted an awkward email with ChatGPT. That's real — but it's the toy version of the tool. The strategic version looks different: the same technology, pointed at the workflows that cost you money, energy, and time, with your rules attached.

So this is not a guide for people who've never heard of AI. It's a guide for businesses standing somewhere on this line — and the honest answer is that every station is a fine place to board from. We continue from where you've begun.

Paper-and-pencil isn't behind; it's unencumbered. Spreadsheet shops already think in systems. If you've dipped a toe and it went sideways, you've bought a lesson most of your competitors haven't. And if you're building, the question becomes governance and compounding — sheets 12 and 10.

Five terms, spelled out once — used plainly from here on

AI: Artificial intelligence — software that can read, write, and reason over your business's words and numbers, not just store them.
LLM: Large language model — the engine behind tools like Claude and ChatGPT. Think of it as a very fast reader-writer you can give instructions to.
Prompt: The written instruction you give an LLM. Better instructions, better output. That's most of the skill.
Agent: An LLM that's been given tools and permission to take steps on its own — look something up, draft the reply, file the record — with a person checking the work that matters.
LMS: Learning management system — the software your team uses to take a course and show they finished it. It's how AI education becomes checkable, not hopeful.

Three lenses

OC-206 / Sheet 06

There are only three reasons to do any of this: money, energy, and time.

Every AI decision in this guide passes through the same three lenses. If a use case doesn't clearly save one of them, it's a hobby — skip it. Most good ones save two.

Money

What you pay, and what tasks cost.

Software seats you barely use, work you pay twice for — once to do it, once to fix it — and hours of skilled people spent on unskilled steps.

A worked example: the office that types every invoice into the accounting system by hand stops typing. The AI reads the invoice, fills the entry, and a person approves it in seconds instead of minutes — hundreds of times a month.

Energy

The attention your week leaks.

Not electricity — morale. The drag of repetitive work is real: it's why good people go flat, and why the important project never starts.

A worked example: every meeting produces its own notes, decisions, and follow-ups automatically. Nobody spends Friday reconstructing what was said on Tuesday, and nothing agreed to gets lost.

Time

Hours back, every single week.

The inbox, the scheduling ping-pong, the status chasing. None of it is your product, and all of it is automatable in part.

A worked example: appointment requests book themselves against your calendar, reminders go out on their own, and the inbox arrives pre-sorted with drafts waiting for the ones that need you.

The next three sheets put these lenses on five departments every small business recognizes. Each row says what the AI does — and where the human stays. That second column matters more. The businesses that get this right don't remove people from the work; they remove the work that was beneath their people.

Where it lands · I

OC-206 / Sheet 07

Customer service, sales & scheduling.

The front of the house: where slow answers cost you customers you already paid to win.

Customer service

Answers, around the clock What the AI doesAnswers the common questions — hours, pricing, status, "how do I" — on your website and by text, at 2 PM and 2 AM, from your own materials. Where the human staysAny upset customer, any refund, anything the AI isn't sure of — handed to a person, with the full conversation attached. MET

Ticket triage What the AI doesReads every incoming request, tags it, routes it to the right person, and drafts a first reply in your voice. Where the human staysApproves or rewrites the draft. The send button belongs to a person. MET

The evidence: in the largest peer-reviewed field study to date, support agents given an AI assistant resolved 15% more issues per hour — and the newest agents gained ~30%, reaching six-month performance in two months. Quarterly Journal of Economics, 2025 · sheet 17

Sales & scheduling

Speed to lead What the AI doesAnswers every inquiry within minutes with a useful, specific reply, and follows up on the ones that go quiet — politely, more than once. Where the human staysThe actual sales conversation. The AI gets the meeting; a person wins the customer. MET

Self-serve booking What the AI doesBooks against your real calendar, sends confirmations and reminders, reschedules without the four-email dance, and writes the visit notes into your records. Where the human staysSets the rules — buffer times, priority customers, what's never bookable. MET

The evidence: across 21 clinical studies, automated reminders cut no-shows from 21% to 15% — a quarter fewer empty slots, in any appointment business. BMJ Open meta-analysis, 2016 · sheet 17

Where it lands · II

OC-206 / Sheet 08

Marketing, reviews & reputation.

The voice of the house: showing up consistently without hiring an agency or losing your weekends.

Marketing

One sitting, a month of content What the AI doesTurns one good conversation about your work into a month of drafts — social posts, the newsletter, ad copy — all in your established voice. Where the human staysThe voice itself, and the final yes. Nothing publishes without an owner's eye. MET

Repurposing What the AI doesOne case story becomes a web page, three posts, an email, and a printable one-pager — each shaped for its format, not pasted. Where the human staysChooses which stories are worth telling, and checks the claims. MET

The evidence: in a randomized study published in Science, professionals using AI finished business writing 40% faster, with 18% higher-rated quality. Noy & Zhang, 2023 · sheet 17

Reviews & reputation

Every review answered What the AI doesDrafts a specific, non-canned response to every review within hours, and keeps your tone — grateful, not groveling. Where the human staysThe flagged ones: the angry review, the legal-adjacent one, the one from a name you recognize. Those get a person, first. MET

Listening What the AI doesSummarizes what reviews and messages are actually saying month over month — the compliment you should double down on, the complaint that's trending. Where the human staysActs on it. The AI finds the pattern; the owner changes the business. MET

The evidence: 89% of consumers expect owners to respond to reviews (industry survey, 2026) — and a Marketing Science study found businesses that start responding earn 12% more reviews and higher ratings. Sheet 17

Where it lands · III

OC-206 / Sheet 09

Admin & the back office.

The quiet department everything gets dumped on — and where the fastest wins usually hide.

Paper in, records out What the AI doesReads invoices, receipts, and forms; extracts the numbers; files them where they belong — flagging anything that doesn't add up. Where the human staysReviews the flags and approves payments. Money moves only on human say-so. MET

Contracts, summarized What the AI doesProduces a one-page plain-language summary of any contract — dates, obligations, renewals, the clause that's unusual — before you sign or renew. Where the human staysJudgment and signature. For anything material, your attorney is still your attorney. MET

Meetings that file themselves What the AI doesNotes, decisions, and action items from every meeting, written up and routed to the people who own them — searchable forever. Where the human staysDecides. The meeting is still the humans'; only the paperwork left the room. MET

The evidence: workers who use generative AI report saving ~5.4% of weekly hours — about two hours a week, before any process redesign. Federal Reserve Bank of St. Louis, 2025 · sheet 17

The same week, before and after.

Illustrative shape of an office manager's week — the shaded share is admin mechanics: re-keying, chasing, filing, reconstructing.

Illustrative, not measured — drawn to the scale of engagements we've seen, not published client data.

The course

OC-206 / Sheet 10

Adoption → Integration → Amplification.

Every business that gets AI right passes the same three marks, in the same order. Skipping ahead is how pilots die: integration without adoption is software nobody uses; amplification without integration is expensive theater.

WPT 01 · Adoption

People first, tools second.

Leadership aligns on why. The whole team — not just the eager few — learns what AI is, what your policy allows, and gets quick wins in the tools they already use. Fear becomes fluency.

Boards here: Paper & pencil · Spreadsheets

WPT 02 · Integration

Wired into real workflows.

One or two departments get AI built into how the work actually flows — with baselines measured before and results measured after. Not a pilot on the side; the actual process, improved.

Boards here: Dipped a toe

WPT 03 · Amplification — berth

Compounding, owned by you.

Systems start talking to each other. Your team writes its own routines. The gains from one department fund the next. The infrastructure is yours — our job was to hand you the helm.

Boards here: Building

The spectrum on sheet 5 maps straight onto this course: wherever you stand is simply your boarding point. Nobody starts at amplification, and — done right — nobody stays in adoption forever.

The crew

OC-206 / Sheet 11

Why education comes first — before any build.

Here's the uncomfortable fact: your people are already using AI at work, with or without your blessing — four in five AI users at small and mid-sized companies bring their own tools in. Ungoverned, that's customer data pasted into personal accounts, advice taken from a tool nobody vetted, and work products no one can explain. The risk isn't that AI arrives at your company. It already has. The risk is that it arrived without you.

Education is how you take command of it. When the whole team learns together — what these tools are, what your rules allow, where the wins are — three things happen at once:

Your team gets excited instead of scared. The question in the break room changes from "will this take my job?" to "could this take the worst part of my job?" That shift is worth more than any single tool. Gartner's researchers put it plainly this spring: "AI adoption is a culture issue, not just a training issue" — standard software training alone doesn't build trust.

The savings ideas come to you. The people closest to the work know where the hours leak. Once they're fluent, they'll surface use cases no consultant would find — and they'll actually want the solutions to work.

Your people stay. Two-thirds of leaders now say they wouldn't hire someone without AI skills — which means your best people are building those skills somewhere. Gartner predicts that by 2027, half of organizations without a people-centric AI strategy will lose their top AI talent. Education is a retention strategy wearing a training badge.

This is also the cheapest phase of the whole journey — and the one with the highest failure cost when skipped. Most stalled AI efforts didn't fail on the technology; they failed on the people and process around it.

52%

of U.S. workers say they're worried about the future impact of AI use in the workplace.

80%

of AI users at small and mid-sized companies bring their own AI tools to work — usually without guidance or clearance from the top.

22%

of employees say their organization has communicated a clear AI plan or strategy.

1:3:5

McKinsey's 2026 rule for AI agents: for every dollar on the technology, three go to process redesign and five to capability building and adoption.

Pew Research Center 2025 · Microsoft/LinkedIn Work Trend Index 2024 · Gallup 2025 · Gartner 2026 · McKinsey 2026 — full citations, sheet 17

Guardrails

OC-206 / Sheet 12

A one-page policy beats no policy.

You don't need a legal department to use AI responsibly. You need six decisions, written down, told to everyone. Here they are.

Name the approved tools.

Which AI tools are in-bounds for company work, on company accounts. Personal accounts for company work are out — that's where data walks away.

Draw the data line.

What may never be pasted into an AI tool: customer personal information, health and financial records, anything under NDA. One sentence, zero exceptions.

Keep a human on external sends.

Anything that leaves the building — quotes, contracts, customer replies, published content — gets human review before it goes. Drafting is the AI's job; sending is yours.

Decide your disclosure manners.

When customers are talking to an AI, they can tell — say so plainly. Trust survives automation; it doesn't survive pretending.

Verify before you rely.

AI can be confidently wrong. Numbers, legal language, and anything you'd stake the business on get checked against the source. Train the habit, not just the rule.

Name an owner.

One person owns the policy, approves new tools, and hears about problems first. Without a name attached, every rule above is a suggestion.

The rules around you — status this quarter.

Connecticut Act by Oct 1

Connecticut passed a comprehensive AI law in June — the AI Responsibility and Transparency Act (Public Act 26-15). What it means first: if AI tools help make hiring, promotion, or firing decisions, you must disclose that to applicants and employees starting October 1, 2026 — the tool, its purpose, and what data it uses. There is no small-business exemption. Customer-facing chatbot rules follow in January 2027.

Separately, the state's privacy law reaches far more small firms as of July 1, 2026: it now applies at 35,000 Connecticut consumers — or any amount of sensitive data (health, biometric, precise location). Worth a coverage check even if it never applied to you before.

Florida No new duties

Florida's "AI Bill of Rights" passed the Senate twice this year — and died in the House twice. No new AI statute binds a typical small Florida business today. The state's privacy law starts at $1B in revenue; that's not you.

What applies everywhere, including Florida: ordinary truth-in-advertising law. The FTC is actively suing companies that oversell AI (its "Operation AI Comply" settled two more cases this year, both involving products sold to small businesses). Every "AI-powered" claim in your marketing must be true and provable — and be skeptical of vendors promising AI-driven passive income.

A plain-language status read, not legal advice. This panel is rewritten every edition; statutes, dates, and case records on sheet 17.

Notices to mariners

OC-206 / Sheet 13 · Q3 2026

This quarter in AI — what actually changed.

Chart corrections, in plain language. Everything below happened since the spring; each entry names its source on sheet 17.

Apr 23ChatGPT's engine steps up. OpenAI's GPT-5.5 arrives in ChatGPT — markedly better at documents and spreadsheets, the everyday work most owners already touch it for. OpenAI

May 6Payroll gets an agent. QuickBooks unveils Workforce: payroll, hiring, and onboarding run by AI agents inside the subscription many small firms already pay for. Intuit

May 28Copilot joins the standard Microsoft plans. Microsoft 365 Business Standard and Premium now ship with Copilot built in (permanent from July 1 at $23.50 and $32 per seat). AI stops being an add-on decision. Microsoft

May 19–20Google folds AI into Workspace. Gemini is included in all Workspace Business plans, and Google's newest everyday model is billed at less than half the cost of comparable models. Google

Jun 9A flagship ships — then vanishes for three weeks. Anthropic releases Claude Fable 5; a security incident triggers U.S. export controls and the model goes offline until July 1. The business lesson: never build a critical process on a single AI vendor. Anthropic

Jun 17Your storefront enters the assistants. Shopify switches on agentic commerce: merchants' products become findable and buyable inside ChatGPT, Copilot, and Gemini — automatically, with per-channel controls. Shopify

Jun 30Near-frontier AI at commodity prices. Claude Sonnet 5 launches at $2/$10 per million tokens and becomes the free-tier default; Opus 5 follows July 24 at half the flagship's price. Anthropic

Jul 1Restaurants, opted in. Square connects eligible food sellers to ordering inside ChatGPT and Claude — a new sales channel with no setup, no fees, no added commissions. Square

Jul 8Voice AI crosses the line. OpenAI's GPT-Live models listen and speak simultaneously — natural, interruption-tolerant voice, now the ChatGPT default. The bar for phone bots just moved. OpenAI

Jul 20The largest copyright settlement in history. A federal court approves a $1.5B class settlement over AI training data — a caution flag on how any business sources and uses AI-generated content. Court record / Authors Guild

Aug 2Europe's disclosure rules go live. The EU AI Act's transparency provisions take effect: serving EU customers with a chatbot or AI-generated content now carries disclosure duties there. European Commission

Notices to mariners · II

OC-206 / Sheet 14 · Q3 2026

…and what it means for a business your size.

The same developments, translated. Three verdicts: worth a look this quarter, watch for next quarter, or can wait without cost.

The AI already in your office suiteIf you pay for Microsoft 365 or Google Workspace, you now own AI seats — or can add them for less than a phone bill. Audit what you already have before buying anything new.Worth a look

Connecticut's new AI lawCT employers whose tools help screen, promote, or fire must disclose that use starting October 1. If software ranks your applicants, that's you — see sheet 12.Worth a look

Your products inside AI assistantsShopify and Square switched this on automatically. Check what the assistants show for your business before your customers do — and claim the channel settings.Worth a look

Voice AI on the phone lineThe technology crossed the "doesn't feel robotic" line this quarter. Appointment-heavy businesses should plan a pilot; everyone else can watch a quarter.Watch

Agents that operate softwareComputer-use agents shipped in mainstream models — real, but early. Pilot only behind human review, on work you can check.Watch

The model price warNear-frontier AI fell to commodity prices this quarter. You'll feel it through your software vendors — on renewal, ask what AI you're now paying for and what got cheaper.Watch

EU rules & the copyright settlementUnless you serve EU customers or build products on AI-generated content, note both and move on. If you do either, get advice now rather than later.Can wait

The engine room

OC-206 / Sheet 15

For the technical reader.

Skipped by most, loved by some. If sheets 7–9 made you ask "but how, exactly" — this page is yours.

A prompt is a work order.

The skill is specificity: who it's for, what good looks like, what to avoid. Compare a real one against "write a post about our company":

Prompt pattern — reviews (sheet 08)

Draft a reply to this customer review. Our voice: warm, specific,
no corporate phrases, never defensive. Mention the detail they praised.
Two sentences, then sign off as the owner. If the review raises a
complaint or anything legal, do not draft — flag it for me instead.

A routine is a prompt you've saved and run on schedule — the Monday-morning pipeline summary, the month-end draft. An agent is a routine with tools and permissions: it can look things up and take its approved steps, escalating to a person by rule, not by chance.

What the mainstream tools cost — as of August 2026.

ToolPer seat / monthNotes

Claude Team$20–25Anthropic's team workspace; $20 billed annually, $25 monthly. Premium seats with 5× usage: $100–125.

ChatGPT Business$20–25^PA$20 billed annually, $25 monthly, two-seat minimum — after an April price cut. Corroborated but not page-verified; OpenAI blocks automated checks.

Microsoft 365 Copilot Business$21Add-on to a Microsoft 365 business plan ($18 promo through Sept 2026); or take Business Standard with Copilot built in at $23.50 all-in.

Google Workspace Business Standard$14Gemini AI now included in every Workspace Business plan — no separate AI seat to buy.

Zoom AI Companion$0Meeting summaries and Q&A included with paid Zoom plans.

Verified at official pricing pages on Aug 18, 2026, except where PA-marked. Prices move often — check the vendor's page before buying.

Sailing orders

OC-206 / Sheet 16

Your next ninety days.

No purchase required for any of this. Just decisions, in order.

Days 1–30 · Educate & align

Get the crew aboard.

Hold one leadership conversation: what are we trying to save — money, energy, or time — and where does it hurt most?
Write the one-page policy (sheet 12). Tell everyone.
Put the whole team through a shared AI-fundamentals course — everyone, not just the eager few.
Ask every department for its three most repetitive tasks. Keep the list.

Days 31–60 · Pick & measure

Choose one workflow.

From the list: pick the one workflow that scores highest on money-energy-time.
Measure its baseline honestly — hours per week, error rate, cost. You can't celebrate what you didn't measure.
Set the human line: what stays with a person, no matter what.

Days 61–90 · Pilot & decide

Run it for real.

Put AI into that one workflow — the real one, not a sandbox.
Review against the baseline at day 90 with the team that runs it.
Decide: expand it, fix it, or kill it. All three are wins — you now decide from evidence.

Want the readings taken for you?

Start with the free AI readiness assessment — three minutes, scored 0–100, with prioritized next steps for your leadership team. It's the same instrument we open every engagement with.

ocv-ai-assessment.netlify.app

Scan · 3 min

The next edition of this guide ships at the start of Q4 2026 — free, at the same address: oceancrest-ai.netlify.app/guide/

The legend

OC-206 / Sheet 17

About OceanCrest · Sources & corrections.

OceanCrest AI helps small and mid-sized businesses adopt AI the durable way: education for the whole team, an honest diagnostic of where the money, energy, and time are leaking, and — when the evidence supports it — systems built inside your walls that you own. Everything we sell is designed to be finished.

Chris Kuselias

Founder — three decades of building and leading companies; founder mentorship and growth strategy.

Cheryl Rykowski

People & education — leadership alignment, team training, and the culture side of adoption.

Eliyahu Cohen

Systems & builds — architecture, diagnostics, and the software your team ends up owning.

David Taiwo

Engineering & delivery — enterprise systems background; department diagnostics and build delivery.

Engagements are contracted through Ocean Crest Consulting LLC.
oceancrest-ai.netlify.app · hello@oceancrestai.com^PA

Source ledger — Q3 2026.

Every figure in this edition, in order of first appearance. PA — not yet confirmed at the original publisher. S02–S04 survey large organizations, not small businesses; the text says so where they appear.

U.S. Chamber of Commerce · Empowering Small Business, 4th ed. · Aug 2025 · n=3,870 — 58% genAI use; 82% of AI users grew headcount; fewer than 1 in 3 well prepared for new AI laws. Sheets 3, 4, 12.
McKinsey & Co. · Superagency in the Workplace · Jan 2025 · n=3,613+238 — 1% of leaders call their companies mature on AI. Sheets 3–5.
McKinsey & Co. · The State of AI · Nov 2025 · n=1,993 — 88% use AI in ≥1 function; ~6% high performers. Sheet 4.
MIT Project NANDA · The GenAI Divide (preliminary findings) · Jul 2025 — 95% of enterprise genAI pilots, no measurable P&L return; small sample, self-labeled preliminary. Sheet 4.
Business.com · 2026 Small Business AI Outlook · Jan 2026 · n=1,009 — 30% early adopters / 54% careful evaluators / 21% wait-and-see; categories overlap, sum >100%. Sheet 4.
Brynjolfsson, Li & Raymond · Generative AI at Work · Quarterly Journal of Economics · 2025 · n=5,172 agents — +15% issues resolved/hour; ~+30% for newest agents. Sheet 7.
Robotham et al. · BMJ Open systematic review & meta-analysis · 2016 · 21 studies — reminders cut no-shows 21%→15%. Sheet 7.
Noy & Zhang · Science · 2023 · n=453 — professional writing 40% faster, quality +18%. Sheet 8.
BrightLocal · Local Consumer Review Survey · Feb 2026 · n=1,002 — 89% expect owners to respond to reviews. Industry survey. Sheet 8.
Proserpio & Zervas · Marketing Science · 2017 — businesses that respond to reviews receive 12% more reviews, higher ratings. Sheet 8.
Bick, Blandin & Deming · Federal Reserve Bank of St. Louis · Feb 2025 — genAI users report saving ~5.4% of weekly work hours. Sheet 9.
Pew Research Center · U.S. Workers & AI · Feb 2025 · n=5,273 — 52% worried about the future impact of AI use in the workplace. Sheet 11.
Microsoft + LinkedIn · Work Trend Index · May 2024 · n=31,000 — 80% of AI users at SMBs bring their own AI tools; 66% of leaders wouldn't hire without AI skills. Sheets 3, 11.
Gallup · AI Use at Work Has Nearly Doubled · Jun 2025 — only 22% say their organization has communicated a clear AI plan. Sheets 3, 11.
Gartner · press release · May 13, 2026 · n=12,004 — "AI adoption is a culture issue, not just a training issue"; by 2027, 50% of enterprises without a people-centric AI strategy will lose their top AI talent. Sheet 11.
McKinsey & Co. · Agentic AI Change Management · Aug 2026 — $1 technology : $3 process redesign : $5 capability building. Sheet 11.
State of Connecticut · Public Act 26-15 (CART Act, signed Jun 2026) and Public Act 25-113 (CTDPA amendments, in force Jul 2026) · cga.ct.gov. Sheet 12.
Florida Senate · SB 482 & SB 2-D (2026, died in House) · flsenate.gov — and FTC Operation AI Comply 2026 settlements (Growth Cave; Air AI) · ftc.gov. Sheet 12.
Vendor primary sources · Apr–Aug 2026 · Anthropic, OpenAI, Microsoft, Google, Intuit, Shopify, Square announcements + official pricing pages, checked Aug 18, 2026 (ChatGPT Business pricing PA — OpenAI blocks automated checks) · EU Commission AI Act timeline. Sheets 13–15.
U.S. District Court, N.D. Cal. · Bartz v. Anthropic settlement approval, Jul 2026 (via Authors Guild) — $1.5B. Sheet 13.

Edition record.

Q3 2026 — first edition · Corrected through 2026-08-18 · Next correction: late September 2026, for the Q4 2026 edition. Watch for that pass: U.S. Chamber 5th edition (~Sept) and McKinsey State of AI next wave (~Nov).