AI consultant in Burton upon Trent
I’m Martin Holecek, an independent AI consultant and full-stack developer based in Burton upon Trent. I help small businesses across Staffordshire, Derby and the Midlands work out where AI genuinely saves time, prove it with a small measured pilot, and implement it properly.
Everyone keeps talking about AI. Where would it actually help?
An AI workflow review answers this with numbers: what each repetitive task costs you now, and where AI (or something simpler) would make a measurable difference.
Practical AI services, from first review to working solution
Everything starts small and measurable. You always know what a stage costs, what it delivers, and what the numbers say before committing to the next one.
AI workflow review Stage 1 · Diagnose
We go through the repetitive admin your team actually did last week: emails, quotes, documents, reports, data entry. Remotely that is a video call with examples you send over in advance; on site, around Burton and the local area, I watch how the work really happens. Either way, I map each task, estimate what it costs you per year in staff time, and give you a short written report on where AI would genuinely help, where simpler automation would do a better job, and what to try first. The fee is credited against a pilot or implementation if you go ahead.
AI workflow pilot Stage 2 · Prove it
A small fixed-price first project: one clearly defined process, improved with AI and measured properly. We agree the process, measure how long it takes today, build a small working solution, and measure again. For example: an enquiry email arrives, AI drafts the reply and the job record, your staff check and send. You get real before and after numbers for your own business, not a vendor’s case study, before committing to anything larger.
AI implementation & integration Stage 3 · Build
When a pilot proves its value, I build it into your real systems: connected to your database, your admin system and your existing workflow, with proper access control. This is where being a full-stack developer matters. I can build the software around the AI, not just the prompt. Connecting an assistant to your own live data is covered further down the page.
AI tools setup & staff training Standalone
For teams already using ChatGPT, Claude, Microsoft Copilot or similar in an unstructured way. I set up business accounts properly: usage guidelines, what is safe to share, reusable prompts for your common tasks, role-specific workflows, and hands-on training for your staff. How the training sessions work is covered further down the page.
Do you need the review before a pilot? No. If you already know which process you want to improve, we can start straight at the pilot; it always begins by measuring the task as it works today. The review is for when you are not sure where AI would help, and its fee is credited against a pilot or implementation if you carry on. Whatever a pilot produces is yours to keep: the working setup, the prompts and the before and after numbers.
Ask your business questions in plain English, and get answers from your own data
Most AI advice stops at drafting text, which is the small win. The bigger one is giving an assistant controlled access to your own live business data, so it answers from your real records instead of guessing. That works for you asking how the business is doing, and for your team asking what to do next.
What this looks like in practice
For the person running the business, it replaces waiting for a report. I have built this for a manufacturing business. The owner can sit down on a Friday and ask how this month compares with the last one, where capacity is tight next week, which customers have quietly gone quiet, or which jobs made money and which lost it. The answer comes back in a minute or two, from the live system, without anyone being asked to pull the numbers together first.
For the team, it removes the digging. The same connection answers the day to day questions: what is on the machines now, which jobs are held up waiting on materials, what this customer ordered last time and what it cost. Real orders and real stock as of this morning, not a general model’s best guess.
It works in the other direction too. Once an assistant can read the data, it can be given a small number of carefully chosen actions as well, such as drafting a quote based on what that customer ordered last time, ready for a person to check and send.
Underneath it is a Model Context Protocol (MCP) server, an open standard for giving AI assistants controlled access to a system. Because it is a standard rather than one vendor’s feature, the same connection works with Claude, ChatGPT and anything else that supports it. You are not tied to whichever assistant happens to be ahead this year.
Access is granted one capability at a time. Reading stays read-only, anything that changes your data is added deliberately and separately, and people only reach what their role should reach.
The kind of thing people ask
What the owner asks
- How does this month compare with the last one?
- Which customers are quietly ordering less than they used to?
- Where is capacity tight next week?
- Which jobs made money, and which ones lost it?
What the team asks
- What is running on the machines right now?
- Show me every job still waiting on materials.
- Summarise this customer’s history before I ring them.
- Draft a quote for this customer based on their last one.
Every answer comes from the business’s own live records, not from the model’s general knowledge.
This needs somewhere for the data to already live. If you run an admin system, a database, or anything with an API, this is a connector built on top of what you have. If the business still runs on spreadsheets and shared folders, that system has to come first, which is the custom business systems side of what I do. The AI layer goes on afterwards, once there is something solid worth asking questions of.
Typical tasks worth reviewing
These are the kinds of everyday tasks where AI assistance tends to pay for itself. Your review looks at your actual workload, not a generic list.
What does a repetitive task actually cost you?
Before building anything, put a number on the problem. This is the same calculation an AI workflow review starts with.
Try it with one of your tasks
- Staff time on this task
- 5 hours/week
- Cost per year
- £5,200
- Cost after AI assistance
- £2,080
- Estimated saving per year
- £3,120
A worked example: quotation preparation
- 12 quotes a week, at 25 minutes each. That is 5 hours of staff time every week.
- At £20 an hour, the business spends roughly £5,200 a year preparing quotes.
- An AI-assisted drafting step cuts each quote to around 8 minutes. The same work now costs about £1,700 a year.
- That is an estimated saving of around £3,500 a year, from one task.
The point is not the exact figure. The point is that once the cost is named, “should we try AI for this?” becomes a business decision instead of a hype decision.
Illustrative figures only. An AI workflow review measures your real numbers, and a pilot verifies the saving on your own work before any larger commitment.
Training where your team does the work, not watches slides
Most AI training is a demonstration. Someone shows a clever tool for an hour, everyone nods, and on Monday the work happens exactly the way it always did. These sessions are built the other way round.
Built around your kind of work
Sessions run on a fictional company with a realistic workload drawn from your industry rather than a generic tutorial. The material looks like the job your team actually does, so what they work out in the room still applies when they get back to their desks.
Your team does it themselves
Everyone works through the material hands-on and makes the calls for themselves: what to hand to AI, what to keep well away from it, and what to check before anything reaches a customer. People remember the decisions they made, not the ones they watched somebody else make.
They leave with something usable
Each session ends with guidelines and prompts written for the roles that were in the room, plus a clear sense of where AI earns its place and where it quietly gets in the way. Practical habits, not a slide deck nobody opens again.
Run on site around Burton upon Trent and the Midlands, or remotely for teams further afield. Sessions are shaped around your team’s actual roles and quoted after a short call, so you are not paying for material that does not apply to you.
Small steps, each one earning the next
Free initial call
A short 20 to 30 minute conversation about your business and the repetitive work your team deals with. We decide together whether a review is worth doing. No obligation.
AI workflow review · from £300
We map your repetitive tasks and put a yearly cost on each one. Remote by video call from £300, or on site around Burton from £500. You get a short written report: where AI would measurably help, where simpler automation wins, and what to try first.
AI workflow pilot · from £900
One process, improved and measured. Time per task before, time per task after, in your business with your real work. The result is evidence, not a promise.
Implementation
What the pilot proves gets built properly: integrated with your systems and data, with sensible access control. Quoted individually before any work starts.
Ongoing care
If MHDev hosts what we build, a monthly care plan from £79/month keeps it hosted, backed up, updated and supported.
If the review shows AI is not the right answer, I will say so. You keep the report and the numbers either way. Sometimes the best recommendation is simple automation or a proper database rather than AI.
AI is not always the answer, and I am fine with that
Many AI agencies sell exactly one thing: AI. My main business is building custom systems for companies that have outgrown spreadsheets, so I have no reason to force AI into a workflow where a simple database, an integration or a well-designed admin screen would work better.
When AI is the right tool, the same skills mean it gets implemented properly: connected to your systems and your data, with sensible controls, rather than a demo that falls over after the first week.
Every pilot has a before and after number. You see the saving in your own business before spending more.
Claude, ChatGPT, Microsoft Copilot, or no AI at all. The recommendation follows the problem, not a vendor deal.
Full-stack development background: databases, APIs, admin systems and integrations, not just prompts.
Clear fixed starting prices, small stages, and a genuine willingness to say “you do not need AI for this”.
An AI consultant based in Burton, working across the Midlands
I live and work in Burton upon Trent (often written Burton on Trent). For local businesses the first conversation can happen in person over a coffee. I also work with companies in Derby, Swadlincote, Lichfield, Tamworth and Birmingham, and remotely across the UK.
You talk directly to the person who reviews, builds and supports the work. No sales team, no outsourcing, and no pressure to buy tools you do not need.
Areas I cover
Tell me about the tasks that eat your week
The best first step is a short, free call. Tell me about the repetitive admin in your business, and we will work out together whether an AI workflow review is worth it.
Prefer email? Write to hello@mhdev.co.uk.