Where to actually start with AI (it isn't the impressive bit)
The right first AI project is not the most valuable one, it is the most valuable one you can finish. What makes a good candidate, why sector matters, and a free tool that gives you a ranked shortlist for your business.
"We should be doing something with AI" is the single most common sentence we hear at the moment, usually followed by a pause, because nobody is quite sure what the something is.
That pause is reasonable. There are genuinely hundreds of things AI could do in an average business. The problem is not a shortage of options, it is that the options arrive undifferentiated, and the impressive-sounding ones are usually the worst place to start.
Why businesses start in the wrong place
Left to browse, most people gravitate towards whatever sounds most transformative. Predictive analytics. A custom chatbot trained on the whole business. Something with a dashboard.
These are the hardest, slowest and least certain things on the list. They need clean data, proper design and a real budget, and they take long enough that enthusiasm runs out before anything ships. We have seen businesses spend a year on one and end up with a demo.
Meanwhile the boring items sit there untouched. Meeting notes written automatically. Invoices read straight into the accounts system. A drafted reply waiting for every routine enquiry. None of it is exciting. All of it works in a fortnight, on tooling you may already own.
The right first project is not the most valuable one. It is the most valuable one you can finish.
What makes a good candidate
A task is worth automating when it is repetitive, rule-based, high volume, low judgement, and recoverable when it goes wrong.
Pulling figures off a supplier invoice hits all five. So does triaging an inbox, summarising a long document, or drafting a routine letter from a template. The volume justifies the effort, the rules are clear enough for a machine to follow, and a mistake is caught and corrected rather than being expensive and invisible.
A task is a poor candidate when it needs judgement, when the volume is too low to justify building anything, or when an error would be costly and hard to spot. Plenty of things are technically automatable and still not worth automating. Anyone who will not tell you which is which is not being straight with you.
Sector matters more than people assume
Generic advice is why so much of this feels irrelevant. "Use AI to personalise your marketing" means very little to a construction firm halfway through a tender.
The genuinely useful use cases tend to be specific to how a sector works.
A recruitment agency gets most of its value from CV screening against fixed criteria, matching open roles to an existing candidate database on meaning rather than keywords, interview notes written up against a competency framework, and job adverts drafted in a consistent format. Four things, all achievable, all attacking the same bottleneck.
A construction firm gets it from tender documents read and broken into a response matrix, site notes dictated on a phone and turned into consistent daily reports, and method statements drafted for a competent person to review and approve.
A care provider gets it from rota checking, evidence mapped against inspection questions, incident themes surfaced across a quarter rather than each incident being closed in isolation, and policies rewritten so staff can actually follow them.
An accountancy practice gets it from invoice data extraction, management accounts commentary, and job costing analysis that shows which clients are genuinely profitable.
Same underlying technology, completely different starting points.
Get a shortlist for your business
Rather than another list of everything, we built something that narrows it down.
Three questions: what sort of business you are, how many people work there, and which parts of the business you care about. It returns a ranked shortlist of twenty automations, drawn from a bank of seventy-four, every one of which an SME could realistically get running inside a quarter.
The ranking is deliberately biased. Impact counts most, then a bonus if a use case is specific to your sector, then low effort to break ties. Anything that only makes sense at larger headcounts gets pushed down rather than hidden, because knowing what is coming later is useful.
Each item shows an effort rating and what it typically runs on, named honestly. A lot of them say "already in your licence", because for most businesses on Microsoft 365 Business Premium a good share of this needs configuration rather than new subscriptions. It takes about a minute and nothing is sent to us.
Twenty ideas is still not a plan
The list is a starting point, not a programme. What to do with it is straightforward and slightly unglamorous.
Pick one low-effort item. Time the process as it stands today, honestly, because you cannot prove a saving without a before. Apply the tool. Run it alongside the manual process for a fortnight so you are comparing rather than hoping. Measure again. If the numbers are not there, stop and pick another one.
Then do the next. The businesses that get real value do this four or five times in a year and end up with genuine change. The ones that try to do it all at once have a working group and no working automations.
If you want to know whether your business is in a fit state to start at all, the AI Readiness Assessment covers the data, governance and process groundwork that decides whether any of this sticks. If you want to know what it might be worth, the AI ROI Calculator will tell you. And if you would like someone to work through the shortlist with you, get in touch.
Written by CT1 Technologies
