The AI Examiner — AI automation agencies, 2026-08-05

Headline

Referrals, not cold outreach, bring in most clients for AI automation agencies.

Summary

This is a baseline for AI automation agencies, a focus this publication hasn't covered before, so there's nothing yet to compare it against. Public directories confirm a real market in the US and Canada with real, published prices, though the three counts found don't add up to one number. Client acquisition again leans on referrals over cold outreach, matching what the AI-consulting issue found — though that finding still comes from advice articles written for agency owners, not from a survey of what actually happened to them. Pricing is real but the sources disagree with each other; those numbers are reported as found, not smoothed into a single range.

The evidence

Market size & growth

• top-ai-agencies.com's directory lists 40 agencies total, with a rough regional read of ~16 in North America (US + Canada combined, not split further), ~19 in Europe, ~5 elsewhere — (fetched, inferred — the split was reconstructed from individual listing locations, not a stated breakdown, and North America here is not separated into US vs. Canada).
• techreviewer.co states 397 verified companies in its AI automation agencies directory, with pagination showing 25 shown per page across 16 pages — (fetched, observed — a directly stated total). No geographic breakdown of that 397 is given anywhere on the page, so it cannot be compared apples-to-apples with the other two counts.
• bestaiagencies.ca — a Canada-specific directory — states 23 of 23 agencies shown, explicitly "Canada-wide," with representation from Alberta, Ontario, Quebec, and British Columbia — (fetched, observed — a directly stated, fully-shown count).
• These three counts don't add up to one market-size figure — the directories use different criteria and don't share a method. There's no earlier count to compare against yet, so today's numbers are just a starting point for whenever this focus gets checked again.

Pricing (hourly, project-based, and retainer kept separate; multiple sources shown separately where they conflict, per the "don't smooth over contradictions" rule)

Hourly (fetched, observed, from techreviewer.co's individual agency listings): a spread from $20-29/hr up to $100-149/hr depending on the agency, e.g. BeeWeb LLC $20-29/hr, Fingent $30-49/hr, Snotor and WeAreBrain and Technology Rivers $50-99/hr, Pell Software and Knoativ $100-149/hr. A separate aggregate guide (fetched, inferred) states a broader $150-350/hr range for discovery/audit/troubleshooting work specifically — notably higher than the per-agency listed rates, and not reconciled here.

Project-based — two sources, genuinely contradictory, not averaged:

• Source A (fetched, inferred, aggregated across multiple guide articles): $5,000-$75,000.
• Source B, digitalagencynetwork.com (fetched, observed — directly quoted from the article): "Setup projects typically range from $2,500 to $15,000+."
• techreviewer.co's buyer's guide (fetched, observed): "Discovery audits typically range $2,000-$15,000, while fixed-price projects span $10,000-$150,000+."
• These three ranges overlap at the low end but diverge sharply at the top ($15,000 vs. $75,000 vs. $150,000+) — treat as a real market-pricing spread or inconsistent source methodology, not resolved by this issue.

Retainer — also contradictory: Source A (fetched, inferred) states $3,000-$20,000/month; digitalagencynetwork.com (fetched, observed) states "$500 to $5,000+" for ongoing monitoring/maintenance specifically. These may be describing different service scopes (full retainer vs. maintenance-only), which would explain some of the gap, but neither source says so explicitly — noted as a gap, not assumed.

SaaS-style / basic tier (fetched, observed, digitalagencynetwork.com): basic automation packages starting at $99/month, scaling to "$1,000-$5,000+/mo for enterprise-level personalization, predictive analytics, and cross-channel orchestration."

How they find clients

• Referral partnerships are described as one of the fastest, cheapest ways to get clients: a formal referral agreement with a 10-15% referral fee is called standard, and warm introductions are reported to close at 30-50%, compared to a 5-10% close rate for cold outreach — (fetched, inferred — from several "how to get clients" guide articles written for agency owners, not a survey of real practitioners).
• A single strong referral partnership (e.g. with a marketing agency serving 30+ local businesses) is cited as capable of generating 2-4 referrals per month on an ongoing basis — (fetched, self-reported example within a guide article, not independently verified).
• Building a public reputation (YouTube, LinkedIn, speaking at industry events) and picking a niche are both repeatedly cited as helping indirectly, by making word-of-mouth and inbound referrals more likely — (fetched, inferred).
• This matches what the AI-consulting issue found, and answers a question raised afterward in learnings.md — "worth checking whether the same pattern holds in adjacent practitioner niches (automation agencies, fractional AI officers)" — on the evidence gathered here, yes, it does.

What they read

• n8n Discord: 84,834 members — (fetched, observed, directly stated). n8n is a workflow-automation platform many AI automation agencies build on, making its community a plausible, sizable distribution channel for this audience specifically — more specific to this niche than the general AI newsletters below. No ad-rate or sponsorship information was found for it.
• No newsletter or publication specific to "AI automation agencies" (as opposed to AI/tech generally) with a stated subscriber count or ad rate was found this run.
• The general AI newsletters already identified in issue 1 (The Rundown AI, Ben's Bites, Superhuman AI) probably also reach this audience, since automation-agency operators likely read general AI news too — but that's a carried-over assumption, not something checked again this run.
• General (not niche-specific) newsletter ad-rate benchmarks found this run: AI/tech newsletters have a median open rate of ~50.4%, the highest of any category (fetched, observed); a rough CPM benchmark of $10-30 for consumer niches was cited, with one worked example (5,000 subscribers, 40% open rate, $60 CPM → $120/slot) (fetched, inferred — general benchmark, not confirmed to apply to this audience specifically).

Methodology & sample

Everything in this issue came from public web searches and page fetches — five searches and four direct page pulls, all fetched-tier (top-ai-agencies.com, bestaiagencies.ca, digitalagencynetwork.com's pricing article, and techreviewer.co's automation-agency directory). Nobody had dropped anything into examiner/sources/ before this run, so there's no operator-supplied data here. Some numbers are stronger than others: the directory counts and the n8n Discord member count came straight from the source (observed). The client-acquisition pattern and the general newsletter benchmarks are inferred — read across several guide articles rather than confirmed by a survey of real practitioners. Pricing is the shakiest part of this issue: several observed and inferred figures are shown side by side because they don't agree, not blended into one number that would hide the disagreement.

Gaps / more research needed

• No single, reconciled total count of AI automation agencies in either market — three directories (40, 397, 23) use different criteria and don't share a common geographic breakdown, so they can't be summed or compared directly.
• techreviewer.co's 397-company count has no stated US/Canada/international split at all — the largest count found this run is also the least geographically specific.
• Project-based and retainer pricing both show real, unreconciled contradictions across sources (project-based upper bound ranges from $15,000 to $150,000+ depending on source). This may reflect genuinely different service scopes rather than inconsistent reporting, but no source made that distinction explicit — a real gap, not smoothed over.
• No automation-agency-specific publication with a public ad rate was found — the same gap issue 1 had for AI consulting. The Rundown AI's pattern (large audience, pricing hidden behind a contact form) may well repeat here, but wasn't checked again this run.
• The referral finding matches what issue 1 found for AI consulting, but it's still based on advice articles, not a survey of what automation-agency operators actually experience. Seeing the same pattern twice makes it a bit more convincing, but doesn't confirm it.
• No trend data yet — there's only one issue on this focus so far, so there's nothing to measure change against until the next one.
• Nobody supplied any gated data for this run — no rate-card PDF, no paywalled report on this niche. If someone drops files into examiner/sources/ before the next run on this focus, the pricing section is where it would help most; it's the weakest part of what's here.

Framing note

AI is moving fast, and this research is still early. Later issues will fill in the gaps and sharpen these numbers — and if something here turns out wrong, we'll say so in public, not quietly fix it.