For a small or mid-sized business: $2,500–$5,000 for a scoped operations audit, $8,000–$35,000 for a first working build, or $30–$500 an hour if you buy time instead of outcomes. Here is how each number is built — with our own prices in the table.
Most AI consultancies won’t put a number on their website. We will. For a business with roughly 5–200 employees, a scoped operations audit costs $2,500–$5,000 and takes 1–2 weeks; a first working build costs $8,000–$35,000 and takes 3–6 weeks; and if you buy hours instead of outcomes, published rates run from about $30 an hour for marketplace engineers to $500+ for big-firm partners. The table below shows what each tier of the market actually publishes, and where our prices sit inside it.
| What you’re buying | Typical 2026 price | Where the number comes from |
|---|---|---|
| Freelance AI engineer (marketplace) | $30–$50/hr entry · $50–$75/hr intermediate · $75–$100+/hr expert; median $35–$60/hr | Upwork |
| AI development agency (reviewed) | Most charge $25–$49/hr; typical minimum project $10,000–$49,999; average project $120,594; average timeline 10 months | Clutch |
| Independent / boutique / big-firm consultants | $150–$350/hr independent · $250–$450/hr boutique · $350–$500+/hr Big 4 & strategy firms | Leanware guide (vendor-published) |
| Strategy assessment or roadmap (fixed price) | $5,000–$25,000 for a small assessment; $25,000–$100,000 mid-size | Leanware guide |
| Monthly advisory retainer | $2,000–$5,000/mo for 5–10 hrs; $5,000–$15,000/mo for 10–25 hrs | Leanware guide |
| Salary benchmark if you hired instead | Management analyst median $48.65/hr ($101,190/yr); computer & information research scientist median $67.74/hr ($140,910/yr), 2024 | BLS · BLS |
| Cerebrum — Phase 1 Operations Audit | $2,500–$5,000, 1–2 weeks, 2–3 working sessions, written roadmap you keep | cerebrumone.ai/services |
| Cerebrum — Phase 2 Solution Build | $8,000–$35,000, 3–6 weeks, fixed price scoped after the audit, 30-day check-in | cerebrumone.ai/services |
The spread is real. On Upwork the median AI engineer bills $35–$60 an hour, with experts “well over $100.” Agencies reviewed on Clutch most commonly bill $25–$49 an hour, but their typical minimum project is $10,000–$49,999 and the average project comes in at $120,594 over ten months — enterprise work, mostly. Published guides for independents and boutiques cluster at $150–$450 an hour, and Big 4 or strategy firms at $350–$500 or more. For context, the U.S. median wage for a management analyst is $48.65 an hour, and for a computer and information research scientist $67.74. An hourly rate is a cost of input. It says nothing about how many hours the job needs or whether the person has done it before in a business like yours — which is why we don’t sell hours at all.
When you buy an outcome instead of time, the range tightens. Small strategy assessments in the published guides run $5,000–$25,000; ours is $2,500–$5,000 because it is deliberately narrow — one to two weeks, inside your real workflows, ending in a prioritised roadmap rather than a deck. A first build then runs $8,000–$35,000 depending on how many systems it touches and how much judgment sits at the edges. Compare that with Gartner’s finding that enterprise GenAI deployments cost $5 million to $20 million: the tools have got cheap enough that a 15-person firm can afford real automation, but only if the scope stays surgical. Fixed price also fixes the incentive. If we misjudge the effort, that is our problem, not your invoice.
Every quote we write is built from the same five variables. Systems touched — one system (email to your CRM) is cheap; four (email, PDF, ERP, spreadsheet, owner report) is not. Data condition — clean, structured records versus PDFs that look different every time. Judgment fraction — how many cases follow the rule and how many need a person; the more exceptions, the more review design. Consequence of a wrong answer — an internal correction, an apology to a customer, or a regulator; anything with money or compliance attached gets a human sign-off step, which costs build time. Volume — the same workflow at 20 a week and 400 a week is priced the same to build but pays back very differently. Ask any consultant to explain their price in these terms. If they can’t, the number is arbitrary.
An audit should end with something you can hand to another vendor: a workflow map with volumes and time-per-instance, a scored list of automation candidates (we usually find six to ten), and a build/buy/wait recommendation for each with hours saved, estimated cost and payback. A build should end with a system running inside the tools your team already uses, a human review step exactly where money or compliance is involved, handoff documentation, and — for us — a 30-day post-launch check-in. If a proposal doesn’t name these deliverables, you are buying activity. RAND’s 2024 study found that misunderstandings about a project’s intent and purpose cause more AI failures than any other factor, and were named by 84% of the practitioners interviewed. Written deliverables are the cheapest insurance against that.
Undisclosed pricing lets a consultancy quote what they think you will pay. Small businesses are especially exposed: the U.S. Chamber’s 2025 survey of 3,870 firms found 58% now use AI, up from 23% in 2023, and most are buying it for the first time with no reference point. We publish ranges for the opposite reason. If the numbers above are out of your budget, we would rather you know before the call. And if the honest answer is that a $12-a-month Make plan and a ChatGPT Business seat will do the job, we will say so — see Do I need an AI consultant, or can I just use ChatGPT and Zapier?
Take one workflow. Multiply the hours a week your team spends on it by a loaded hourly cost, then by 48 working weeks. A coordinator spending 12 hours a week re-keying orders at a loaded $35 an hour costs about $20,000 a year; a $15,000 build that removes 80% of that pays back in roughly eleven months, and everything after is margin. Gartner’s survey of 822 early adopters reported average cost savings of 15.2% and productivity gains of 22.6% — useful ceilings, but your own hours are the number that matters. If the arithmetic doesn’t clear in under 18 months, don’t build it yet. That is exactly the calculation the operations audit runs for every candidate we find.
ChatGPT Business is $20 per user per month billed annually (minimum two seats) and OpenAI states it does not train on your workspace data by default. Zapier starts at $19.99 a month for 750 tasks, with a free tier of 100 tasks; Make’s Core plan is $12 a month for 10,000 operations. For a generic task inside one system, that stack is often the right answer and no consultant should tell you otherwise. It stops being the right answer when the workflow crosses systems, when the exceptions are the job, or when nobody in the building owns the automation once the person who built it moves on. That is where a fixed-price build earns its keep.
The operations audit is $2,500–$5,000 and stands alone. Most first builds land between $8,000 and $35,000; we will tell you before the audit if your situation is unlikely to justify that.
No. Every phase is fixed price and scoped in writing before work starts. If a build takes us longer than quoted, the price does not change.
Published ranges: roughly $30–$100+ per hour for marketplace engineers, $25–$49 for most agencies on Clutch, $150–$450 for independents and boutiques, and $350–$500+ for large firms. Ask what the hours will produce before comparing rates.
Plan on $10,000–$25,000 all-in for an audit plus one focused build, and expect payback inside 6–18 months if the workflow you pick eats at least 8–10 hours a week today.
These take a while to write. One email when the next goes up, nothing else.
Book a free 30-minute operations call. No pitch, just an honest conversation about where AI can help.