The honest answer is: it depends on what you are actually asking for. But that is not very useful when you are trying to work out whether to budget a few hundred pounds or a few thousand. So here is a more practical version: the range is wide, the drivers are predictable, and if you understand what pushes a project up or down the scale, you can get a reasonable sense of where your situation sits before you speak to anyone.
What I can give you here are realistic ballparks based on what these engagements actually involve. Not vendor list prices. Not marketing copy. Just what the work costs to do properly.
What drives the cost of an AI consulting engagement
AI consulting cost breaks down into three components: scoping time, build complexity, and whether the engagement is advisory-only or advisory plus delivery.
Scoping time is how long it takes to understand your business well enough to give useful advice. If your processes are well-documented and your systems are straightforward, scoping is fast. If nothing is written down, every team member does things differently, and your data sits across five disconnected tools, scoping takes longer. That time costs money regardless of what you build afterwards.
Build complexity is the main variable in any implementation project. A document processing tool that reads a single structured document type and pushes data to one system is a fundamentally different piece of work from a multi-system workflow that handles exceptions, learns from corrections, and integrates with a live ERP. Both are "AI." The costs are not remotely comparable.
Advisory-only versus advisory plus build matters because advice and delivery require different time commitments. An advisory engagement produces a roadmap, recommendations, and a clear picture of what is worth building and what is not. An advisory-plus-build engagement goes further: the recommended work actually gets built, tested, and handed over. One costs consultancy day rates. The other costs consultancy day rates plus development time plus integration effort.
Typical cost ranges by engagement type
These are realistic UK market figures for SME-scale engagements. They are not guarantees. Every situation is different. But they give you a working frame.
These figures reflect what the work actually involves, not vendor marketing rates.
| Engagement type | Typical cost range |
|---|---|
| AI readiness assessment / process audit | A few hundred to around £1,500 |
| Full AI strategy and implementation roadmap | £1,500 to £4,000 |
| Document processing tool (build) | £3,000 to £8,000 |
| Custom automation workflow (build) | £5,000 to £15,000+ |
| ERP add-on or integration (build) | £8,000 to £20,000+ |
| Fractional advisory retainer (monthly) | £500 to £1,500 per month |
The plus signs at the upper end of some ranges are not a hedge. They reflect genuine variability: a workflow that touches three systems and handles several exception types costs materially more than a single-system, single-document-type version of the same concept.
What makes it more expensive
These factors reliably push a project toward the top of its range or beyond it.
- Undocumented or inconsistent processes. If the work is done differently by different people, the first job is to agree on the correct process before any automation begins. That is scoping time that would not exist if the process were already defined and followed.
- Multiple system integrations. Each integration point adds development time, testing, and ongoing maintenance surface. A tool that needs to read from one system and write to one other is simpler than one that coordinates across four.
- Real-time processing requirements. Batch processing is generally cheaper to build than real-time or near-real-time pipelines. If the use case requires immediate output, the architecture is more involved.
- Poor data quality. If the data the AI will work with is inconsistently formatted, incomplete, or spread across sources that have never been reconciled, data preparation work comes before any AI build. That can be a significant line item in its own right.
- Compliance-sensitive environments. Healthcare, legal, financial services, and regulated manufacturing add documentation, audit, and testing requirements that straightforward commercial projects do not have.
What keeps the cost down
The inverse of the above, largely. But there are a few things worth being deliberate about.
- Start with one well-defined use case. Scope creep is the main cause of AI projects running over budget. Pick one process, define its boundaries clearly, and build that. Expand later once you have a working foundation.
- Use off-the-shelf tools where they genuinely fit. Not every problem needs custom code. If an existing tool handles 80% of the requirement acceptably, building custom for the remaining 20% often costs more than it is worth. Part of a good advisory engagement is working out which is which.
- Document the process first. Before any technical work starts, write down exactly how the process currently runs, including exceptions. This saves scoping time and produces a cleaner specification for the build.
- Consolidate your data before you build. If the data is clean and accessible, build time drops. If it is not, clean it first. Trying to handle messy data inside an AI tool adds complexity to every layer of the project.
The cheapest AI project is not the one with the lowest quote. It is the one scoped tightly enough that it actually gets finished, works as intended, and does not need rebuilding six months later.
How to think about return, not just cost
The right question is not "how much does this cost?" It is "how long until this pays for itself, and what does it return after that?"
Here is a worked example from my own business. At Vanda Coatings, we processed a high volume of supplier and job-related documents manually. Each document took time to open, check, cross-reference, and enter. At 40 documents per day, 12 minutes each, with a fully loaded staff cost of £20 per hour, the annual cost of that process was around £26,000. That is not an estimate. That is a time study.
An automated document processing solution, built to extract the relevant fields and push them for human approval rather than manual entry, cost in the range of £3,000 to £5,000 to build. Annual maintenance and API costs run to approximately £500. The payback period from go-live was six to eight weeks.
After that payback point, the system produces net savings of around £20,000 per year, year after year, for a one-time build investment. The question was never whether £4,000 was a lot of money. The question was whether £4,000 against £20,000 per year made sense. Put that way, it is not a difficult decision.
Apply the same logic to your situation. Take one process you are considering. Work out what it currently costs in actual staff time at fully loaded rates. Compare that to the build range for that type of automation. If the payback is under 12 months and the benefit repeats annually, the case is strong. If the payback stretches beyond 18 to 24 months, the project needs sharper scoping or a different approach.
Why a quote without a scoping call is meaningless
Any consultant who gives you a fixed price before understanding your processes in detail is guessing. They may be guessing in good faith, but they are guessing. The number they give you reflects their assumptions about what you need, not the actual shape of your situation.
The problem with a guessed quote is not that it will necessarily be wrong. The problem is that you have no way to know whether it is right. And if it is wrong in the wrong direction, you either overpay for something undersized or underpay for a proposal that will run over budget and under-deliver.
A proper scoping conversation takes 30 to 60 minutes. It covers: what the process currently is, what systems are involved, what the data looks like, what the expected output is, and what success looks like. From that conversation, a consultant can give you a range with enough confidence to be useful for budgeting. Without it, any number is illustrative at best.
The cost ranges in this article are starting points for thinking. They are not quotes. Before you decide whether AI consulting is worth pursuing for your business, have a proper scoping conversation and get a range that reflects your actual situation rather than a generic template.
For a breakdown of what's included at each stage, see how the consulting works. If you want to talk through your specific situation before committing to anything, book a free call.
Sources
- British Chambers of Commerce, "The Turning Point for SMEs", September 2025
- Department for Science, Innovation and Technology (DSIT), AI investment data, 2024
- OECD, "Generative AI and the SME Workforce", 2025