The AI Tools Assessment — A Productized AI-Audit Business (Corey Gannon)

Source: raw/The_1_000_hour_Solo_AI_business_Full_Course.md — Greg Isenberg’s Startup Ideas podcast, full free-course episode with Corey Gannon walking through his “AI Tools Assessment” productized-service business, its fulfillment pipeline, upsell menu, and client-acquisition playbook. All figures are Gannon’s self-reported business numbers, not independently audited.

Corey Gannon runs a $999 “AI Tools Assessment” — a doctor-style diagnostic where he interviews a small-business owner for 45 minutes, then prescribes 3-7 off-the-shelf AI/SaaS tools to reclaim 5-10 hours a week (guaranteed, or a full refund). No custom building or coding is required to deliver the base offer — the business model is explicitly “prescribing tools that already exist,” which he frames as replicable by anyone willing to learn the AI landscape well enough to make good recommendations. He gives away his exact report template and full upsell/client-acquisition playbook in the episode.

Key Takeaways

• The core offer: $999AIToolsAssessment.ICP:businesseswith2-20employees,$500K-5M annual revenue. Format: 45-minute structured discovery interview, then a prescribed list of 3-7 off-the-shelf tools guaranteed to reclaim at least 5 hours/week (average client reclaims ~7) — full refund if that bar isn’t hit. Gannon reports 15 assessments sold in 2026, with 50-60% of clients buying a follow-on implementation engagement.
• Four-phase fulfillment pipeline:
  1. Discovery call — recorded via an AI note-taker (Fathom, Otter, or Fireflies). Pure probing, no pitching: “walk me through your day yesterday,” “what do you dread doing,” “if you could delete any process with a magic wand, what would it be” (answer is very often “email”).
  2. Transcript analysis — feed the call transcript to a Claude Skill (a simple version works too: “Here’s the transcript, research off-the-shelf SaaS/AI tools that fix these pain points”) which researches and prescribes tools, then the operator does a QA pass substituting tools that are miscalibrated for the client’s size (his example: swapping out an over-scaled Salesforce prescription for a four-person landscaping business). Two directories he uses for tool research: futurepedia.io (since acquired by HubSpot) and thereisanaiforthat.com.
  3. Report generation — client-facing deliverable, built in Claude Design (previously built in Gamma). Structure: executive summary (main pain point + hours reclaimed/week + “primary focus” — effectiveness/efficiency/quality), an effort-vs-impact matrix (top-left “quick wins” quadrant = the report’s actual recommendations), a 4-day quick-start plan (~10 min/day to get most of the value), and a financial-impact slide (monthly net ROI = weekly hours reclaimed × hourly rate − monthly tool cost; average tool cost across clients is ~$60/month). The free downloadable template is at audittemplate.ai, built to plug directly into Claude Design (unverified — not independently checked at ingest time).
  4. Review call — 30-minute screen-share walkthrough of the report, closing with three questions: which recommendation is most urgent, do they want to self-implement or hire help, and what’s their timeline. This is where 50-60% convert to an upsell.
• Six-item upsell menu (all sold after the assessment, to an already-warm client):
  • Process redesign ($3,000-3,500) — fixing a broken process (cutting steps, e.g. 16→7) before automating it. Explicit rule: don’t throw AI at a broken process — a client asking to “just automate” a 10-hour/week Amazon ad-spend review got a process fix first, automation as a separate engagement.
  • Simple automation build (~$1,500) — Zapier/Make/n8n for cut-and-dry 1-3-step workflows where a full Claude Skill is overkill.
  • Knowledge system — e.g., a custom GPT trained on a business-broker’s listing materials, replacing 400-500 repetitive buyer-inquiry emails per listing with a link both buyer and seller can query directly.
  • Custom workflows — proprietary, business-specific processes built as Claude Skills, often with a recurring maintenance/update retainer layered on top.
  • “AI concierge” retainer (his favorite, currently in growth) — two 45-minute monthly calls teaching the client to use Claude Co-work and build Claude Skills directly, plus “unlimited” Voxer access (12-business-hour response — in practice almost never used: 3 of 5 clients messaged him once each in ~2.5 months, despite driving the sale on high perceived value). Priced per-client, ratcheting upward as demand proved out: $1,200\to$1,500 → $1,800\to$2,000/month for his first five clients. First call is pure onboarding (connect tools, build context files, set global instructions, show a scheduled task and a skill) so every subsequent call compounds. Ongoing session structure: AOA — Audit the current process, Optimize by cutting steps, Automate into a Claude Skill — repeated per workflow for the life of the engagement. Reported result: $8K MRR within 10 days of first pitching the offer (closed 5 of the first 6 people he told about it), 4 of 5 initial clients hit their stated 90-day win inside 60 days.
  • Full implementation — any combination of the above bundled (example: $8,000 package spanning two process redesigns + several Claude Skills + a Zapier workflow).
  • Pricing psychology note: crediting the $999assessmentfeetowardalaterupsell(e.g.,$5K → $4K)whilequietlymarkingtheupsellupanextra$1K nets the same revenue but reads to the client as a discount.
• Seven zero-capital, zero-audience client-acquisition methods he actively uses or has seen work:
  1. Local “AI for business” meetup — free co-working-space conference room (bartered for the space’s own exposure), 20-minute talk (usually “how to use Claude”), collect contacts, follow up within 24 hours. Framed as a long-game channel, like SEO — results compound by the 3rd-6th event, not the 1st.
  2. Door knocking — a listener who door-knocked 30 local service businesses converted 2 clients from 5 resulting meetings. Gannon: “the single best way to learn sales, period.”
  3. LinkedIn DMs — probing, not pitching, in message one; targeted at local business owners specifically (not cold spam-list outreach).
  4. Free mini-audits for personal network — a 15-minute free version of the assessment as a lead magnet, explicitly framed to the prospect as “worst case, you learn one tool; best case, you get a partner.”
  5. Agency-partner referrals — insurance agents, accountants, marketing consultants, coaches: offer to be their AI-question resource with a referral fee, requires following up every 2-3 weeks to actually generate referrals. Can be combined with method 1 as a co-branded workshop.
  6. “AI office hours” at co-working spaces — sit in a co-working space once a week as a free walk-up AI resource. One community member combined this with door-knocked local businesses for free gift cards as a draw, got 11 people to a first session, 2 booked follow-ups.
  7. Build in public — post small, real wins (even “I answered someone’s ChatGPT question well”) regularly; explicitly framed as manufacturing content in the early days before real client wins exist yet.
• Niching-down advice (from host Greg Isenberg): specialize by geography (“the AI guy in Charlotte”) or by vertical (10 years in financial services → “the AI assessment guy for financial services”) — reduces price sensitivity and differentiates against the “thousands of people starting this exact thing” competition Gannon himself acknowledges exists.

Try It

1. Copy the 4-phase structure directly: recorded discovery call → Claude-driven transcript analysis with a manual QA pass → a templated report (start from audittemplate.ai if it’s live, or replicate the four-slide structure: executive summary, effort/impact matrix, quick-start plan, financial-impact slide) → a review call ending in the three closing questions.
2. Build the base offer as a Claude Skill, not a one-off prompt — feed it corrected transcripts and finished reports from your first assessments so it improves per Gannon’s own account (60-70% right on the first few, “copy-paste” close by the 4th-6th).
3. Price the guarantee, not just the deliverable — “at least 5 hours/week back or full refund” is what makes a cold $999 ask land as “irresistible” (host’s word) rather than risky.
4. Pick one of the 7 acquisition methods that fits your personality and actually run it for 7+ days — Gannon’s repeated point is that the competition on these specific channels is low precisely because most people quit after 2-3 attempts.
5. If considering the retainer upsell, model the Voxer-access line item carefully — it drives perceived value in the sales conversation far more than it costs in actual response-time load, per his own usage data (3 of 5 clients, once each, in 2.5 months).

Open Questions

• audittemplate.ai was not independently verified as live or as the actual template — referenced in the transcript, not fetched or confirmed during this ingest.
• All revenue and conversion figures are Gannon’s self-reported numbers (15 assessments, 50-60% upsell rate, $8K MRR in 10 days, 4-of-5 clients hitting a 90-day win) — no independent verification exists; treat as anecdote from a single practitioner, not a benchmarked result.
• Geographic/market sensitivity of the $999 price point is not addressed in the source — Gannon operates out of Charlotte, NC; whether this price holds in other markets or against different ICP revenue bands is unstated.
• How much of Phase 2 (tool research) actually runs unattended vs. requires the operator’s judgment calls is somewhat in tension in the source — Gannon says it becomes “copy-paste” by the 4th-6th assessment, but also frames tool substitution (e.g., swapping an oversized CRM prescription) as an ongoing necessary QA step even at scale.

Related

• Clay + Claude Code — Natural-Language Lead Generation — sibling operator-workflow article in the same “AI-powered service business” cluster, same emphasis on a single-prompt/skill pipeline replacing manual work.
• Hermes User Stories — documents a directly comparable business model (an IT/automation consultant deploying Hermes into client SMEs with per-client isolation), the closest existing parallel to Gannon’s AI-concierge retainer.
• Claude Design — the tool Gannon’s client-facing report template is built in.
• Agent Skills Overview — the underlying primitive behind both the transcript-analysis skill and the “AI concierge” retainer’s Claude Skills coaching.
• The Marketing Prompt Stack — where productized-service prompt patterns like this one fit in the broader marketing/ops prompt landscape.
• Karpathy AutoResearch Cold Outbound — a heavier, more automated cold-outbound system, useful contrast against Gannon’s deliberately manual/high-touch, zero-capital acquisition methods.