How Much Does AI Consulting Cost? Real Numbers, Engagement Models and What Drives the Price

People ask this question because every firm answers it with "it depends", and it does depend, but not on anything mysterious. AI consulting is priced the way any professional work is priced: by how much thinking you need, how much building you need, and who ends up owning the result. Here are the numbers we see in the market, the models firms use to charge, and the few things that actually move the figure.

The short version

• Advice alone, a strategy or readiness assessment: from a few thousand dollars for a focused two-week review to the low six figures for a large-enterprise program from a big-name firm.
• One working system, a single process automated or an internal tool built: typically $15,000 to $80,000 with a conventional agency over two to four months. We do a scoped first release in five days at a fixed price, which sits at the bottom of that range.
• A connected platform, several processes on shared data: $100,000 to $400,000 over six to twelve months at agency rates, less when it is built one process at a time and each one proves itself before the next.
• A multi-quarter transformation program with a large consultancy: seven figures, and most of it is not software.
• Ongoing support after a build: from a few hundred to a few thousand dollars a month. Ours starts at $999 a month, month to month.

Those ranges are wide because the market is wide. A twenty-person consultancy in Chicago and a global systems integrator are not selling the same thing, even when the slide deck says they are. What follows is how to tell which one you are buying.

What you are actually paying for

Strip the language away and an AI engagement is some mix of four things. Thinking: understanding your process, your data and where intelligence would change a decision. Building: writing and shipping the software. Integrating: connecting it to the systems you already run, which is usually where a project lives or dies. Owning: who keeps the code and the data when the invoice is paid. The price is the sum of how much of each you need, times the rate of whoever does it.

The single biggest driver is the ratio of thinking to building. A firm that only advises is selling hours of senior time, and senior time is expensive whether or not anything ships. A firm that builds is selling working software, which costs more per week but ends. The trap is paying advisory rates for a roadmap and then paying again for someone to build it.

What moves the number

• Scope. One process is a week. A department is a quarter. "The business" is a program with a steering committee. Every honest quote starts by asking which of these you mean.
• Data readiness. If the records the system needs live in four places that disagree, someone has to unify them before AI can do anything useful. That work is real, unglamorous and often half the budget on a first project.
• Integration surface. A standalone tool is cheap. A tool that must read from your ERP, write to your accounting system and notify your CRM costs more, and is worth more, because it removes the re-keying that was the problem in the first place.
• Who does the work. Offshore body shops, domestic agencies, boutique firms and big-brand consultancies charge in different orders of magnitude for the same task. The output is not the same either; you are paying for judgment as much as for typing.
• How it is built. Agentic AI tools have changed the cost of writing software. A team that uses them well, under senior engineers who own the architecture and review everything, delivers a working first release in days where it used to take months. Firms that have not made that shift are still quoting the old calendar.
• Ownership. Renting a black box is cheaper this year and more expensive every year after. Owning the code costs the same up front and nothing later.

The three ways firms charge

Most credible firms offer more than one commercial model, and the model tells you what the firm is optimising for.

• Time and materials or retainer. You pay for hours or for a monthly block of them. Flexible, and the right choice when you are extending your own team with people who know the tools. The risk is that nothing forces a finish line.
• Fixed price for a fixed scope. You pay an agreed figure for an agreed deliverable. The right choice for a first project, because it forces the firm to scope honestly and puts the delivery risk on them. This is how our five-day builds work: one page of scope, one price, nothing starts until you have read it.
• Partnership or revenue share. A milestone-billed build followed by a share of what the product earns. Little upfront, aligned incentives, and only sensible when the software is the business rather than a tool inside it.
• Managed service. A fixed monthly fee to run and improve the system after it is live. Predictable, and what most operators want once the first build has proven itself.

A worked example

A distributor with forty staff assembles quotes by hand from a price list, last quarter's quotes and a phone call to the warehouse. It takes a sales person about forty minutes a quote, they send twenty a day between them, and about a third go out with a mistake that costs margin or a callback. That is roughly thirteen hours a day of skilled time on a process that follows the same steps every time.

The big-consultancy answer is a CPQ platform: a six-figure licence, a six-month implementation, and a sales team that has to be retrained. The agency answer is a custom quoting tool at $40,000 to $70,000 over three months. Our answer is a working quoting engine in five days at a fixed price: it reads the real price list, checks stock, produces the quote in minutes, and follows it up until it is answered. If it saves even half of those thirteen hours a day, it pays for itself inside the first quarter, and the distributor owns it.

How to spend a first budget without wasting it

• Diagnose before you build. A free assessment or a short paid one tells you which process is worth the money. Skipping this is the single most expensive mistake, because you will build the loudest idea instead of the highest-return one.
• Insist on production, not a prototype. A pilot that never touches real data and real users teaches you nothing and costs almost as much as the real thing.
• Start with one process and a fixed price. It caps your downside at a week and gives you a before-and-after you can measure.
• Tie every dollar to a number. Hours saved, errors removed, revenue recovered. If the firm cannot tell you which number their work will move, keep looking.
• Keep the code. Whatever the model, the source, the specs and the data should be yours from day one.

What we charge

Since this is a post about prices, here is ours. Scoping is free: a conversation about which process hurts, and an honest answer about whether it is worth building. The first release is a fixed price agreed before we start, for a working version of one process, live for one team, in about five days. You own the code. After that, support and enhancement start at $999 a month, month to month, or your own team can take it forward, or we can train them to build the same way. Larger programs are phased so every phase is separately priced and separately deliverable. The details are on the five-day build page.

The point of pricing it this way is not that it is cheap. It is that you find out whether custom software pays off in your business for the cost of one week rather than one quarter, and the answer is yours either way.