Insights · AI & Business Automation

What to Automate First: A Practical Way to Find Automation Wins

The best first automation is rarely the most visible one. A simple method for mapping, measuring and scoring processes points you to the work worth automating.

16 October 2026 4 min read AAVORAA Team

Most businesses know they should be automating more, but few know where to begin. The usual mistake is starting with the most visible idea – a chatbot, or a tool someone saw in a demo – rather than the work that quietly consumes the most hours. A short, structured method finds better starting points.

Map what actually happens

Before choosing any tool, write down how work really moves through the business. Pick one area – sales, operations, finance or customer service – and trace a single piece of work from start to finish: an enquiry, an order, an invoice, a delivery.

Note every step, who does it, which system it touches and where information is copied from one place to another. The written procedure and the actual practice are often different. Spreadsheets on someone’s laptop, WhatsApp groups used as task lists, data re-typed from email into accounting software – this is where automation opportunities hide.

  • What triggers the process, and who starts it
  • Every system the work passes through
  • Every point where someone copies, re-types or reformats data
  • Every approval or judgement a person makes
  • Where work waits, and for how long

Measure hours and the cost of errors

For each process, estimate two things: how many hours it takes per week across the team, and what it costs when it goes wrong. Rough figures are fine. Ask the people who do the work; they usually know exactly which tasks eat their afternoons.

Error cost is often the more revealing number. A mistyped GST number, a missed follow-up or a wrong stock count takes seconds to cause and days to correct. A task that takes modest time but produces expensive mistakes can be a better candidate than one that takes longer but rarely goes wrong.

Score each candidate

Once you have a list, score each process on four factors, using a one-to-five scale for each, and multiply the scores. Multiplying rather than adding pushes the strongest candidates clearly to the top and sinks anything that is weak on any one factor.

The highest scorers are usually frequent, time-consuming, predictable tasks where an error is easy to catch. Those are your first projects. Processes that score low on rules or high on risk aren’t ruled out; they simply come later, once the team has experience with and trust in the systems.

  • Frequency: how often it happens – hourly, daily, weekly, monthly
  • Time: how long each instance takes
  • Rules-based: whether the steps can be written as clear rules or rely on judgement
  • Risk: how contained the damage is if the automation gets it wrong (lower risk scores higher)

Begin at the handoffs between systems

In our experience the quickest wins sit at the joins between tools rather than inside any one of them. A website form that emails a person, who types the details into a CRM, who then has to remember to follow up – that chain has three handoffs, and each is a place where time is lost and leads slip.

Common starting points are form-to-CRM routing with automatic follow-up, scheduled reporting that pulls from several systems, and document intake, where invoices, delivery notes or purchase orders are read and entered automatically. For V3 Data Solutions, a B2B data firm, the work combined workflow automation, CRM integration and automated reporting. The result was 60% less manual workload — 40 hours saved every week.

For ESSA Garments and Taqua Silks, the handoff problem lived in the warehouse. We built a warehouse and POS platform with document intake, GRN, inventory with piece QRs, 33 reports and a floor app, replacing manual GRN and Tally entry. The gain came from removing re-entry between the floor, the stockroom and the accounts.

Where AI fits, and where plain rules are better

AI is useful where inputs are messy or varied: reading supplier invoices in different layouts, classifying incoming emails or tickets, extracting fields from scanned documents, and drafting first versions of replies, summaries or reports for a person to review.

Where the logic is clear – route orders above a threshold to a manager, assign leads from a region to a named person – plain rules are cheaper, faster and easier to audit. Using AI for a decision a simple condition could make adds cost and unpredictability without adding value. The strongest systems usually combine both: AI to read and sort, rules to route and act.

Keep people in the approvals

Automation should remove repetitive effort, not accountability. Payments, credit decisions, customer commitments, price changes and anything with legal or regulatory weight should keep a named person in the loop, at least until the system has a long record of getting it right.

A good pattern is to let the system prepare everything – collect the data, check it, draft the output – and leave a person to approve it in one step. You keep the time saving and preserve judgement where it matters.

Measure before and after

Record the baseline before you change anything. Without it, you will have opinions about whether the project worked rather than evidence. After launch, measure the same things for several weeks and talk to the people affected; their feedback usually points to the next process worth tackling.

  • Hours spent per week on the process
  • Time from trigger to completion
  • Errors or corrections per month
  • Volume handled without adding staff

Key Takeaways

  • Map the real process, including spreadsheets and WhatsApp workarounds
  • Score candidates on frequency, time, rules and risk, and multiply
  • The quickest wins usually sit at the handoffs between systems
  • Use AI for messy inputs and drafting; use rules for clear decisions
  • Keep people on approvals and measure a baseline before you start

Let’s Grow Together

Turn Ideas Into a Plan for Your Business.