AI consulting is advising businesses on where AI makes them money, then getting paid to implement it. At the enterprise tier that means data scientists and 7-figure transformation programs. At the operator tier — the open opportunity — it means a non-engineer selling audits, tool setups, and automation retainers to small businesses at $1,500-$5,000 per engagement plus $500-$1,500/mo to keep it running.
This post defines the operator-level version: what clients actually buy, what it pays, and whether it's a real business (it is — here's the evidence).
The two tiers of AI consulting
Enterprise AI consulting is Accenture and McKinsey territory: custom models, data pipelines, compliance programs. Credentialed, slow-moving, and irrelevant to you.
Operator-level AI consulting serves the $500K-$10M business — the law firm, the HVAC company, the coaching practice. These companies will never hire Deloitte. What they need is someone to look at their operation, point at the 3 places AI pays for itself, and set it up with tools that exist off the shelf: Claude at $20/mo for Pro, Zapier from $19/mo billed annually, a voice agent platform billed by the minute.
No code is required at this tier. The work is diagnosis, configuration, and accountability — operator skills, not engineering. That's the entire thesis of how to become an AI consultant.
What clients actually buy
Five engagement types cover most of the market:
- AI audits and readiness assessments — $1,500-$5,000. A scored diagnosis of where AI pays, with a 90-day roadmap. The natural first sale; the AI readiness assessment is the best-selling format.
- The AI operating system install — $2,500-$5,000. Set up Claude for the client's team: Projects per department, a custom Skills library that encodes their SOPs, training for staff. Then $500-$1,500/mo to maintain it. The same stack consultants run on their own client work — sell what you use.
- Automation builds — $1,500-$4,000. Lead routing, intake-to-CRM, invoicing flows on Zapier or Make, plus a monthly retainer to monitor and extend.
- Custom assistants — $1,000-$2,500. A branded internal GPT or Claude Project trained on the client's documents — sales FAQ bot, proposal writer. ChatGPT custom GPT build-outs run $1,000-$2,500 including workspace setup.
- Team training — $1,000-$3,000 per workshop. Live sessions teaching the client's staff to use what you installed. Often bundled with the install.
Notice what's not on the list: building models, writing software, anything requiring a CS degree.
What AI consultants charge
The honest default rates at the operator tier:
| Engagement | Price | |---|---| | AI audit / readiness assessment | $1,500-$5,000 flat | | AI operating system install | $2,500-$5,000 + $500-$1,500/mo | | Automation build | $1,500-$4,000 + $300-$750/mo | | Custom assistant | $1,000-$2,500 | | Training workshop | $1,000-$3,000 |
If a client insists on hourly, $100-$250/hr is the range — but packaging beats hourly every time. A flat-fee audit with a defined deliverable closes faster and caps scope. Anchor the offer around a clear outcome, not effort.
A typical first engagement, end to end
A 12-person bookkeeping firm, $1.8M revenue. The owner books a $2,500 readiness assessment after meeting you at a local business event.
Discovery surfaces the two real problems: proposals take 5 days because the owner writes each one personally, and client meeting context lives in 3 people's heads. The report scores the firm section by section, and the roadmap leads with a Claude install — a Project per client, a proposal Skill trained on past wins — plus free call capture with Fathom feeding a weekly follow-up digest.
The firm approves a $3,500 install and $750/mo to maintain the Skills library and run monthly training. First-year value of the relationship: about $13,000, from one local conversation. Nothing in it required code. It required knowing what to look for and owning the outcome.
The solo practice math
A conservative year 1, selling part-time until it isn't: 8 assessments at an average $2,200 ($17,600), 4 installs at $3,000 ($12,000), and ending the year with 4 retainers averaging $700/mo (roughly $14,000 collected across the ramp). Call it $40,000-$45,000 in year 1, on tool costs under $100/mo — and the retainer base alone enters year 2 at $33,600 annualized before a single new sale.
The leverage is structural: every assessment feeds installs, every install feeds a retainer, and fulfillment gets faster as your Skills library grows.
Is this a real business?
The skeptic's question, so here's the operator's evidence. SMB owners are actively buying diagnosis: they know AI matters, they don't know where to start, and they'd rather pay a trusted person $2,000 than burn 6 months experimenting. The demand shows up in our inbox weekly from the other side too — businesses asking "how do we use AI in a way that actually helps us get work done, not just write random one-off things."
The risk isn't demand. It's that solo operators under-package — selling vague "AI help" hourly instead of named deliverables at flat prices. The model works when the offer is concrete. AI Operator Academy ($999/yr) exists for exactly that gap: packaging, pricing, and fulfillment systems from operators already selling, traded peer-to-peer rather than theorized.
How to start
Sell one audit. Everything else — installs, retainers, training — falls out of the audit roadmap. Pick your vertical using who to target when selling AI services, run the audit playbook, and treat your first 3 engagements as paid R&D: underpriced, imperfect, and worth more than any course because they produce sample reports, testimonials, and a delivery process you've actually run. And if you're worried AI itself makes this career obsolete, read will consulting be replaced by AI — short answer: the deliverables get replaced, the consultant who runs AI doesn't.
FAQ
Will clients actually pay for this?
Yes — when the offer is concrete. "AI consulting" doesn't sell; "$1,500 audit, scored report, 90-day roadmap in 2 weeks" does. The buyers are established service businesses with real revenue and manual operations, which is why targeting matters more than credentials.
Do I need to be technical to be an AI consultant?
No. The operator tier runs on no-code tools and prompt fluency. Being a former salesperson or service-business operator is the advantage, not the handicap — you speak the client's language and can scope what actually matters. Engineers routinely over-build; operators ship.
Isn't the market already saturated?
The content market is saturated — endless YouTube videos about AI consulting. The delivery market isn't: most businesses in the $500K-$10M range have never been pitched a concrete AI offer by a human they trust. Show up in person in a local market and you'll find you're usually the first.
What does an AI consultant actually do day to day?
In a normal week: half the time on pipeline (conversations, local events, follow-ups), the rest on delivery — discovery calls, analysis in Claude, building reports and Skills, client readouts. The split surprises people: the consulting is the easy part. The selling is the job, which is why operators with sales backgrounds outperform engineers in this market almost every time.
How is AI consulting different from an AI automation agency?
Consulting leads with diagnosis and advice; an automation agency leads with builds and monthly retainers. In practice they blend — most consultants end up with retainer revenue. See what is an AI automation agency for that model's economics.