Most managing directors I speak to have been told AI will change everything. What they want to know is simpler. Where would it make us money, what would it cost, and how quickly would we see it? This is how we answer that for a client, using work we have done, in the order we do it.
Where we look first
We start with the money, and we look for it in four places.
What you pay for today
Software licences, outside suppliers, the hours your team spend retyping and chasing. If you rent a system per user, we work out what owning your own would cost against that bill.
Your own data
We look at the real orders, stock and invoices, with your permission, and count where they disagree. It tells us more than any workshop.
Any deadline a customer has set
If a big customer wants digital orders by a date, that date sets the plan. The part that has to work goes first.
The people doing the job
We follow an order through the business with whoever handles it. The person who follows it through is the person who builds the fix.
At Miranda, a glazing supplies merchant in Kent, we mapped how the trade counter works and automated the budget checks, order matching and invoice preparation. Invoices are now prepared every day from the orders, and the operations team approves each one. No order goes more than 24 hours without an invoice, where the gap used to be days or weeks.
Where the money usually turns up
For the firms we work with, it tends to be the same few jobs: admin that gets retyped, orders and invoices, replies to customers, checking figures, and stock that the shop and the till disagree on. Here is what changed when we fixed them.
| The job | What changed | Proof |
|---|---|---|
| Admin: invoices raised by hand, days after the order | Invoices prepared every day from the orders. A person approves each one before it goes. | Miranda |
| Orders and invoices by email, spreadsheet and phone | Orders arrive digitally and are checked. Invoices go back in the customer’s format. | Everything Boxed |
| Customer enquiries waiting for someone to reply | A researched, personal reply drafted within three minutes, checked by a person before it is sent. | Revitt (our own) |
| Reporting and checking: weeks of spreadsheet comparison | AI sorts each difference into explainable or needs investigating. Every decision is recorded. | Global HRIS |
| Shops and till quietly disagreeing | Faults traced and fixed: missing order emails, a sales report showing zero revenue, a security hole in the till link. | Mitchell’s Vintners |
Source: Revitt client work and our own systems, 2025 to 2026. Each name links to the full story.
The mess is the work
You don’t need to tidy your data or agree a new process before you talk to us. Sorting that out is a big part of what we get paid for.
Global HRIS, a payroll consultancy, proved two payrolls matched by comparing spreadsheets by hand, over weeks. The files came from SAP, ADP, Workday and others, in CSV or Excel, all laid out differently. We built software that reads the lot, lines them up and uses AI to sort each difference. Employees’ personal details never reach the model.
Everything Boxed ran its orders on email, spreadsheets, phone calls and PDFs. When its biggest customer later couldn’t find 18 invoices, we traced every one and reconciled them to the penny. The reason they couldn’t be searched was a quirk in how the reference was built, so we quoted a small fixed price to fix it for good.
At Mitchell’s Vintners, two online shops are fed from the till. One fault we found made a sales report show zero revenue for products that had sold. Finding and fixing that sort of thing is the job, and it comes before any forecasting.
How fast it goes live
We ship the part that has to work first and add the rest in phases. Everything Boxed had a fixed deadline from its biggest customer. We started on day one, met the date and kept the account. Halo Traffic got a first release scoped at four weeks, and its fleet control system went live under five months after our first proposal. A recent phase of fixes for Halo was quoted at 10 working days.
We can go at that pace because we build with our own AI tooling. One of our experienced people plans each job and checks everything the AI does before it reaches you. We are much faster and cheaper than we were 18 months ago, and the quality is higher. It also means a small job is now worth automating. Making those invoice references searchable was a small job.
Fixed prices, and when we do discovery first
The first step is a free call. We go through the jobs that eat your team’s week and tell you which are worth automating and what each would save.
After that, the build is a fixed price against a written definition of what done looks like. Small pieces, like the job above, get a fixed quote straight away. Larger builds start with a discovery phase, priced for the job, so the build itself can be fixed-price.
If a phase delivers less than we agreed, we don’t bill for the missing part. On one Halo phase a feature was waiting on a message format from the client, so we swapped it for extra work we had already done and charged the original price.
Some of the return is not hours saved. Miranda and Everything Boxed each kept a customer that required digital ordering and invoicing. The review puts the revenue at risk in the business case, next to the hours.
Keeping your data yours
Anything that sends, changes or deletes waits for a person to approve it, and every step is recorded. Personal details are kept out of the model where they are not needed.
If the data can’t leave the building, the AI doesn’t have to either. Open models such as Llama, Qwen, Mistral and Gemma can run on a machine you own, with no per-use fees, and we can set that up for you. We run them on our own kit for our own work, alongside hosted models from OpenAI, Anthropic, Google and Microsoft. We pick whichever does your job best for the money.
What we run ourselves
We use the same methods ourselves. Every enquiry to our website gets a researched reply drafted in three minutes, day or night, and a person reads it before it goes. Our agents and private models run on our own machines, with rented GPUs for the heavy work.
Who looks after it
Once it is live, we look after it on System Care, a monthly fee for updates, monitoring and fixes. Improvements are priced separately, as a written change request, so you always know what you are paying for. At The Embroidery Shed no release goes live unless a full restore from backup has been tested recently.
Some clients take it in-house instead. We helped Oldham Engineering hire its own specialist, handed the system over and ended our retainer.
Keeping the build to what you need
A hotel group wanted rugged Zebra scanners for a stock-take app. We said first that phones would do and the scanners added cost. When they chose the scanners anyway, we specified the Wi-Fi model rather than the 5G one, because the sites run on Wi-Fi. Halo had camera control in scope, and after discovery with the camera provider we left the feeds display-only. For Global HRIS’s move to AWS we recommended one simple application measured in production over a more elaborate set-up.
We don’t replace things that work. The Embroidery Shed kept WooCommerce behind a new shop front, and Oldham kept its own software and had it modernised in two-week blocks.
Want a rough idea first? The free AI readiness check asks ten questions and tells you where we would start.