There is no shortage of things AI can do. The useful question is narrower: which repetitive, expensive task in your business happens often enough that automating it changes a number you care about?
The projects that succeed share a shape — high volume, clear rules, tolerable error cost, and a human in the loop for the edge cases. The ones that fail usually automate something rare, judgement-heavy or unmeasured.
First response, day and night
Answering enquiries within minutes is the highest-value automation for most sales-driven businesses, because response speed is directly tied to conversion. An assistant on WhatsApp or your website can answer the first questions, qualify the enquiry and book a call at 2am on a Sunday.
The rule that makes it work: the assistant handles the known questions and hands over to a person the moment the conversation moves beyond them. Customers accept an automated first reply; they do not accept being trapped in one.
Turning documents into data
Supplier invoices, delivery notes, ID documents, medical reports, insurance forms — anything that arrives as a PDF or photo and gets re-typed into a system is a strong candidate. Extraction is now reliable enough for production when a human confirms anything below a confidence threshold.
Payback here is easy to calculate: hours per week multiplied by the loaded hourly cost, plus the errors that no longer reach your accounting or your patients.
Lead scoring and routing
If your team treats every enquiry identically, the good ones wait behind the noise. Scoring on real signals — source, budget indication, urgency, language, past behaviour — and routing accordingly means your best salesperson sees the best lead first.
This is unglamorous and effective. It rarely needs a large model, it needs your data organised and a rule set that reflects how you actually qualify.
Voice, where volume justifies it
A voicebot answering routine calls — opening hours, order status, appointment changes, basic triage — earns its keep when call volume is high and the questions are repetitive. Below a certain volume, a good FAQ page and a shared inbox do the same job for far less.
Be honest about your call mix before building one. If most calls are unique or emotionally sensitive, a voicebot will damage the experience rather than improve it.
Where it usually wastes money
Content generation with no editorial standard, chatbots trained on a website that was already unclear, dashboards nobody opens, and 'AI strategy' work that produces slides rather than a working process. All expensive, all common.
Also avoid automating a broken process. If your quote takes four days because three people must approve it, an AI writing the quote faster changes nothing. Fix the process first, then automate it.
How to scope a first project
Pick one task, measure it for two weeks before you build anything — volume, time per item, error rate — then automate it with a human review step and measure again. If the numbers do not move, you have lost weeks rather than a year.
Keep the data yours, keep a fallback path when the automation is unavailable, and write down what the system must never do on its own. Those three rules prevent nearly every AI project disaster we have been called in to fix.
Key takeaways
- Automate what is frequent and expensive, not what is interesting.
- First-response automation usually delivers the fastest measurable return.
- Always keep a human in the loop for low-confidence and sensitive cases.
- Never automate a broken process — fix the process first.
- Measure before and after, on a task you have baseline numbers for.
Read the full solution:AI & automation →
Services:Custom Software →
