The AI Examiner — AI consulting, 2026-07-31

Headline

Referrals dominate how AI consultants actually get clients — cold outreach and lead platforms rank lowest. If real, that's the single most important fact for anyone building a product to sell this audience leads, and it should be read with the caution below: it's the pattern with the least primary verification in this issue.

Summary

This is the first issue — a baseline, not a trend. Public data confirms a real, price-transparent AI consulting market in both the US and Canada, with hourly and project-based rates independently documented in each geo. The client-acquisition finding above is the most consequential result but rests on aggregator/guide-article sources, not primary practitioner surveys — flagged accordingly, not smoothed over. No week-over-week growth can be stated yet; that requires a second issue to compare against.

The evidence

Market size & growth

• Multiple directories list AI consultants in both markets, but none published a definitive total count. innovatingwithai.com's "Certified AI Consultant Directory" shows roughly 70 visible listings with a rough geographic split of ~70% US / ~15% Canada / ~15% other international — (fetched, inferred — the split is the source's own rough estimate, not a precise stated breakdown).
• Verifywise.ai's Canada-specific AI governance consultants directory lists 13 providers — (fetched, observed — a directly stated count, though scoped to governance specifically, not AI consulting broadly).
• Clutch.co's Canada AI consulting page lists 10 named firms (Spiria, Devfortress, Denologix, Wizard Labs, Software Secured, WiserBrand, Sigma Software Group, Softmax Data Inc., SaaSberry Innovation Laboratories, Techworks) — (fetched, observed count of firms shown on that page, not a stated total for the category).
• No baseline exists yet to compute real growth. This issue can only record today's snapshot counts above as the starting point for a future comparison.

Pricing (hourly and project kept separate, per prior lesson not to blend incompatible bases)

US, hourly (fetched, inferred — compiled from multiple pricing-guide articles, not a single primary source): solo/independent $80-350/hr; boutique $150-650/hr; mid-tier firms (Accenture, IBM Consulting, Capgemini) $400-700/hr; Big 4 $400-1,200/hr; MBB $500-1,000+/hr.

US, project-based (fetched, inferred, except where noted): strategy sprint / readiness assessment $25K-75K; proof of concept $50K-250K; single use-case build $50K-500K; pilot implementation $10K-25K; full done-for-you implementation $25K-60K. One named firm's discovery engagement starts at $3,500 (fetched, observed — a single firm's own published starting price, Layer3).

US, retainer: $2K-50K/month; advisory-only (office hours) $2K-5K/month (fetched, inferred).

Canada, hourly (fetched, inferred): independent CAD $175-400/hr; boutique CAD $200-450/hr; large firms CAD $400-900+/hr; Big Four partners CAD $600-1,200/hr; contract ML engineers CAD $125-250/hr.

Canada, project-based (fetched, inferred): SME fixed-price (2-4 week) CAD $5K-15K; automation starter CAD $2.5K-7.5K; AI agents CAD $5K-15K; SaaS replacement CAD $10K-35K.

Canada, retainer: CAD $2K-10K/month (fetched, inferred).

One source claims Canadian rates run "generally 10-20% lower than US rates" — (fetched, inferred, single-source — not cross-verified against a second source, treat with real skepticism).

How they find clients

• Referrals are described as the highest-converting client source, with 2-3x the close rate of cold outreach (fetched, inferred — pattern compiled across guide articles).
• One cited statistic: solopreneurs earning $100K+/year get 84% of work through word-of-mouth, attributed to an HBR-featured survey (fetched, observed as stated by the secondary source citing it — the primary HBR study itself was not fetched directly, so provenance is one layer removed).
• Cold outreach, online platforms (Upwork, etc.), and formal referral programs rank lowest among effective business-development strategies in the aggregated survey data found (fetched, inferred).
• One source characterizes Upwork specifically as "a race to the bottom" due to competition from lower-cost-of-living freelancers (fetched, self-reported — a single source's stated opinion, not a broad finding).
• First clients typically come from warm/existing relationships (former coworkers, personal network) before any cold channel activates (fetched, inferred).
• Only ~25% of consultants market their business daily, despite consistency correlating with better client acquisition (fetched, self-reported/inferred).

What they read

• The Rundown AI: 1.7M subscribers on its main AI newsletter, 51.2% open rate, 2.5% click rate — (fetched, observed, directly from their own advertise page). No public ad rate card exists — pricing is gated behind a contact form with a promised 24-hour custom quote. Per this skill's inbound-only rule, that form was not filled out; this is recorded as a gap, not worked around.
• Ben's Bites: 120,000+ subscribers (fetched, observed, cited). No ad-rate data found.
• Superhuman AI: 1.25M+ readers (fetched, observed, cited). No ad-rate data found.
• General (not niche-specific) newsletter ad-rate benchmarks: mid-sized newsletters typically charge $500-3,000/placement, most around $1,000-1,500; one example rate card showed Main Sponsor Spot $750/issue, 4-issue package $2,500, annual (12 issues) $6,000, footer link $200/issue — (fetched, inferred — general B2B/AI-newsletter-industry benchmarks, not confirmed to apply to any specific publication this audience reads, and the exact source of the example rate card is unclear from the search result).
• Real practitioner communities identified as distribution channels beyond newsletters: DataTalks.Club, MLOps Community, TWIML Community, Build Club (AI product community), AI Tinkerers (global Slack network with a confirmed Toronto chapter — real Canada presence) — (fetched, observed to exist). No member counts or ad-rate data found for any of them.
• No LinkedIn group specifically for "AI consultants" with a stated member count was found.

Methodology & sample

This issue rests entirely on fetched-tier data — roughly a dozen WebSearch queries and two direct WebFetch pulls (The Rundown AI's advertise page; the innovatingwithai.com directory) gathered this run. research/supply/sources/ was empty, so there is no operator-supplied data in this issue at all. Confidence is uneven by section: the "where they read" subscriber figures and the directory listing counts are directly observed from primary pages. Almost everything in "pricing" and "how they find clients" is inferred — compiled from aggregator/guide-article content that itself may be compiling from other guides rather than independent primary research. Treat the pricing ranges and the referral-dominance finding as directional, not confirmed.

Gaps / more research needed

• No definitive total count of AI consultants in either market — only partial directory-listing counts, one of them scoped to a narrower niche (AI governance) than the overall focus.
• No growth/trend data — this is a baseline; a second issue is needed before "up/down from last issue" is possible.
• The Rundown AI, the single largest publication identified reaching this audience, does not publish ad rates — real gap, not worked around.
• No ad-rate data for Ben's Bites or Superhuman AI despite finding their subscriber counts.
• No member counts or ad rates for any of the five practitioner communities identified.
• Pricing data leans heavily on aggregator guides rather than primary practitioner-published rate cards — real risk of guides citing each other rather than independently verifying rates.
• The single "Canada rates run 10-20% lower" claim was not cross-verified against a second source.
• No operator-supplied gated data (Upwork rate distributions, a newsletter media kit PDF, etc.) was available for this first run — dropping files into research/supply/sources/ before the next run would materially strengthen the pricing and ad-rate sections specifically.

Framing note

This is an early issue in a fast-moving industry. Numbers here will sharpen, and gaps will close, as more issues accumulate — the Examiner corrects itself in public as that happens.