Most people start by looking at tools. They read about something impressive, watch a demo, or hear what a competitor is supposedly using, and then try to find a use for it in their business. That is the wrong order, and it is why a lot of early AI projects produce no lasting value.
The right order is: understand what your team does, identify where pattern-based work is consuming time, then find the technology that addresses that specific thing. The technology is the last decision, not the first.
Start with what your team does, not what AI can do
Spend an hour listing every repetitive task your team handles in a typical week. Not the interesting work, not the decisions that require judgment, just the things that repeat: filling in the same fields, sending variations of the same email, transferring information from one place to another, checking the same kind of document for the same kind of information.
Write each task down in one sentence. "We manually copy delivery addresses from emailed purchase orders into the job management system." "We draft individual follow-up emails after each site survey." Do not worry about what is solvable yet. Just get it on paper.
This list is your starting point. Everything that follows depends on it. If you want a more structured way to map your business before you start, our business performance framework covers the six areas where most SMEs lose time and money.
Rank by time, not by frustration
Once you have your list, add rough time estimates. Not how annoying the task is, how long it actually takes. Twenty minutes per occurrence, ten occurrences a week, is 3.5 hours. That is worth pursuing. Two hours once a month is two hours a month. That is a different conversation.
Prioritise the tasks that happen most frequently and take the most cumulative time. The emotional weight of a task is not a reliable guide to its actual cost. Some of the most automatable time sinks in a business are tasks nobody complains about because they have become invisible.
Some of the most automatable time sinks in a business are tasks nobody complains about because they have become invisible.
Check whether something already exists
Before building anything custom, find out whether an off-the-shelf tool already does what you need. Off-the-shelf is almost always faster to implement and cheaper to maintain than a custom build. The exception is when your process is genuinely specific to your business in a way that no generic tool can accommodate, or when data privacy requirements make a hosted SaaS product unsuitable.
When you search for tools, search for the specific task, not for "AI tools for business." You want to find tools built to solve the exact problem you have identified, not tools that market themselves broadly. Specificity in your search will save you hours of evaluation time.
Test before you commit
Every tool worth considering has a trial period. Use it with real tasks from your business, not the demo tasks the vendor provides. The vendor's demo is designed to show the best case. Your actual data and your actual processes will surface the real limitations.
Involve the people who will use the tool in the test. Their objections are information, not obstacles. If your team finds the tool confusing, or the output is not good enough to use without heavy editing, that matters. A tool that your team does not trust or does not use produces no value regardless of how impressive the demo was.
The common mistake
The most common mistake I see is businesses picking a tool because it looks impressive and then trying to find a use for it. This produces activity rather than results. A tool that does something genuinely useful for your specific situation is worth paying for. A tool that you are using because you feel you should be using AI is a cost with no return.
The question to ask before any purchase is: which specific task on my list does this replace or speed up, and by how much? If you cannot answer that question, do not buy it yet.
What the process actually looked like at Vanda
At Vanda Coatings, we had the same problem most businesses in the construction industry have: admin was increasing faster than output. Every job generated paperwork. Every stage of a project required someone to move information from one place to another, and that information often ended up in multiple places in slightly different forms. When I sat down and measured the duplication properly, I found around 30,000 points of duplicated data across our processes. That number made the cost of doing nothing concrete rather than abstract.
I looked at off-the-shelf ERP systems. None of them were built around how a commercial recoating business actually works. Forcing our workflows into their structure would have meant changing how we operated to suit the software, rather than the other way round. I decided a bespoke build was the right call, and went back to university to do an MSc in Computing specifically because I wanted the skills to build it properly rather than depend on someone else to build something I could not maintain or improve myself.
The build versus buy question got answered by the specifics of our situation, not by any preference for one approach over the other. If your processes are standard and well-documented, off-the-shelf will almost always be faster and cheaper. Ours were not standard enough, and the cost of bending them to fit a generic product would have been higher than the cost of building something that fitted us.
The decision that made the most difference to the outcome was involving employees from the start. The people doing the work every day knew exactly where the friction was. They knew which parts of a process were pointless duplication, which bits always went wrong, and what information they were missing when they needed it. Most of the features that have made a genuine difference came from suggestions by the people using the system, not from me sitting at a desk deciding what they should want. Features only get built if they have a genuine usefulness: improving efficiency, reducing cost, or saving time. Not because they look good in a demo.
The result has been at least five hours a week saved across the business on administration, with staff now spending that time on things like marketing and client work. The system has had AI features added progressively to reduce admin burden further, and it is still being improved. The improvement cycle has not stopped because the people using it keep finding things that could work better.
If you have done this analysis and want a second opinion on whether your shortlisted tools are the right choice, or whether a custom build makes more sense for your situation, that is a short conversation. See how the consulting works, or book a free call and bring your list.
Sources
- British Chambers of Commerce, "The Turning Point for SMEs", September 2025
- DocuSign Digital Maturity Report, 2024
- ONS, "Management practices and AI in UK firms", March 2025