Automation2 Jun 2026 · 9 min read

Business process automation: where to start

Not "AI on everything." Which processes to automate first, what to leave alone, and how to check if it pays off.

Business process automation rarely fails because of technology. It fails because the company starts with the wrong process — usually the most visible one, not the most profitable. This piece shows how to pick the first process, what not to automate, and how to estimate ROI before you sign anything.

info
According to McKinsey Global Institute (2017), about half of all paid work activities are technically automatable with existing technology, yet fewer than 5% of jobs can be fully automated. That is why picking specific processes matters more than "automating the company".

What makes a process worth automating

A good candidate has four traits at once. Miss one and the payoff shrinks while the project drags.

  • Repeatability — run dozens or hundreds of times a week, not once a quarter.
  • Volume — the more operations, the faster the build cost is recovered.
  • Clear rules — you can state "if X then Y"; where human judgment is needed, automation assists rather than replaces.
  • Cost of error — a process where mistakes cost money (penalties, lost leads, refunds) gains twice from consistency.
warning
Don't automate chaos. If a process isn't documented, automation just speeds up the mess. Tidy it up first, then automate.

Six processes to automate first

These recover cost fastest in our experience, because they combine high volume with clear rules.

ProcessExampleTypical gain
Lead routing & qualificationForm → scoring → assign to repFaster response, zero lost leads
First-line supportVoice/chat agent for FAQ and statusesLess team load, 24/7 coverage
Reminders & confirmationsSMS/WhatsApp before an appointmentFewer no-shows, fewer manual calls
Document generationQuotes, invoices, contracts from templatesHours saved, consistent data
System syncCRM ↔ ERP ↔ accountingNo more double data entry
ReportingAuto-reports from Analytics / CRMDecisions on data, not gut feeling

What not to automate (yet)

  • Processes that change every week — stabilize them first.
  • Decisions needing real judgment (negotiations, disputes, regulatory exceptions).
  • Low-volume processes — the build cost will not pay back sensibly.
  • Everything at once — the first project is one process, not the whole company.

How to check if it pays off

A simple model is enough for a yes/no. Count saved hours per year, multiply by the hourly cost, subtract build and run cost.

ts
// Annual return from automating a process ROI = (saved_hours_per_year × hourly_cost) − (build + 12 × monthly_cost)
tip
If payback is under 6–8 months, it is a good first candidate. Over 18 months — shelve it and start with something higher-volume.

Where to start in practice

Start with discovery: map processes, data and integrations, and pick one process with high volume and clear rules. Build it to production in 2–4 weeks, measure the result on real data, then expand. Lowest risk, fastest return.

info
At CyberNinja we start with a free scoping call: we show what already works in your industry and point to the process that pays back fastest. Book a call: https://cyberninja.digital/book

Newsletter

Like this kind of breakdown?

Once a week — a short letter with numbers from our deployments.