On this page
AI consulting pricing depends on how many processes the work covers, how complex your current systems are, whether custom development is needed and whether you want support after the build. That’s why quoting the same figure to two businesses usually means misleading one of them.
The question I hear most often in discovery calls is “roughly how much?”. It’s a perfectly reasonable question, but the honest answer isn’t a number; it’s a list. In this article I open up that list: which factors make up the figure in a proposal, which ones cause surprises later, and what you can do on your side to keep the cost down.
Why a fixed price list is misleading
You’ll see AI consulting packages laid out like website packages: “starter, standard, enterprise”. They’re convenient for selling, but they create two problems.
First, the package scope comes from the seller’s standard offer, not your need. A business struggling with reporting can end up being sold “three training sessions and a chatbot”. Second, the real cost items sit outside the package: integration with your existing software, tidying up the data, or monthly tool subscriptions.
That’s why I don’t publish a fixed price list. I look at how the work runs first, then send the scope and price in writing. It may look like a way of delaying the number, but in practice it prevents the wrong scope being sold.
Six factors that set the price
1. How many processes are in scope?
This is the biggest factor. Automating a single reporting flow is very different from tackling five separate processes from sales to accounting at once. Starting with a single process as a pilot keeps the cost down and gives you real data for the next steps.
2. How easily can your systems be connected?
If your software allows outside connections, the build moves quickly. If there’s old software, a closed system or a dashboard you can only export from by hand, building the bridge can become the largest part of the work. I check this during discovery, and if something can’t be connected, I say so at the start.
3. The state of your data
If the data is in one place and organised, AI or automation can latch onto it easily. If it’s spread across several people’s laptops, in different formats and duplicated, most of the time goes into tidying it rather than building. This item is invisible in most proposals but is the most common reason projects overrun.
4. Off-the-shelf tools or custom development?
Much of what businesses need can be solved by configuring existing tools properly, and that’s much cheaper. Where off-the-shelf tools fall short, custom AI solutions come in; then scoping, prototyping, testing and handover stages are added and the cost rises noticeably.
5. Training and handover
For your team to use the system and take it over when needed, training and documentation are required. It’s important to see this as a separate line in the proposal. A setup without handover looks cheap at first, but it leaves you dependent on whoever built it.
6. Ongoing support and tool costs
Maintenance after the build is optional; if you want it we set up a monthly arrangement, and if not, the system stays with you. On top of that, the software used may have its own subscription or usage fees. These aren’t part of the consulting fee; they’re billed directly to your account and should be listed separately in the proposal.
The cost factors in one table
| Factor | What pushes the cost up | How you can bring it down |
|---|---|---|
| Scope | Several processes at once | Start with a single pilot process |
| Integration | Closed or legacy systems | Share the list of tools you use at the start |
| Data | Scattered, duplicated records in different formats | Gather the data in one place beforehand |
| Development | A specific need off-the-shelf tools can’t meet | Be open to trying existing options first |
| Training and handover | A team that has never used the tools | Name one person to own the process |
| Support and tools | Constantly changing requirements | Measure usage and need together |
The last column is there on purpose. Part of the cost depends on your preparation, and most of that preparation needs no budget at all. To assess how ready you are, see six signs you’re ready to invest in AI.
Common pricing models
Three models are common in consulting work, and each suits a different situation.
- Project-based, paid in stages. For work with a clear scope. It’s split into stages such as discovery, build and handover, with a concrete output at the end of each. This is the model I use most often.
- Time-based. For discovery-heavy work where the scope can’t be fixed at the start. It’s transparent, but the budget can be harder to predict, so setting a cap is good practice.
- Monthly support. For maintenance, improvements and small new flows after the build. It shouldn’t be mandatory; the system should keep running without it.
Whatever the model, the terms should be written into the proposal. Work that proceeds on a verbal agreement ends in a dispute about scope.
Questions to ask when comparing two proposals
If you have two different proposals in hand, look at these questions before the numbers:
- Which process does the proposal cover, and with what measurable output?
- Will the current situation be measured before the build? If not, how will the gain be shown?
- In whose name will accounts and access keys be created?
- Are documentation and handover training included?
- Are monthly tool and usage fees listed separately?
- If you don’t buy support after the build, does the system keep working?
If the answer to any of these is unclear, the proposal that looks cheaper may end up costing more. The third question matters most: a system built on accounts you don’t own becomes a problem the day you stop working with that person.
Thinking it through with an example
As an example, imagine a ten-person team that compiles a weekly sales report by hand from three different dashboards. The sensible first step for this team is a single pilot: collecting the data automatically and having the report ready on the agreed day.
The questions that set the cost of this pilot are the same as the list above. Do the three dashboards allow outside connections? Is the report format the same for everyone, or does each manager want something different? Who checks the report after the build? If these answers are clear, the scope stays narrow and the cost predictable. If they’re unclear, clarifying the process first is the cheaper starting point.
Frequently asked questions
Can you give a price range for AI consulting?
Not before the discovery call, because a range given without knowing the scope usually creates the wrong expectations. After the call I send the scope, timeline and price in writing. The 20-minute discovery call is free and comes with no commitment.
Are software fees included in the price?
No. Subscription and usage fees for automation platforms, language models or other software are billed directly to your account. I list these separately in the proposal and we choose the lowest-cost option that does the job.
Is it possible to start on a small budget?
Yes. A pilot that starts with a single process is possible on a small budget in most cases. The real risk with a limited budget isn’t that it’s too small but spreading it across several areas. Getting a proper result in one area is both cheaper and more instructive.
Do you guarantee results?
No, and I’d suggest being wary of anyone who does. What I do is measure the current situation before the build and repeat the same measurement afterwards. If there’s a gain, the numbers show it; if not, they show that too.
Let’s clarify the scope
Let’s clarify the scope before talking numbers
The AI consulting process starts with a discovery call. At the end, either I send you the scope and price in writing, or I tell you plainly that you don’t need me for this.
Updated:


