Where to look for automation opportunities
The best candidates are usually ordinary tasks that happen frequently, follow understandable rules and use data your business already holds. Ask the people doing the work where they copy information, wait for approval, rebuild the same report or chase the same update.
Document the current process before discussing software. Include the trigger, each decision, the systems involved, the person responsible and what happens when something goes wrong. This often exposes a simpler process improvement before any automation is needed.
The practical checklist
Repeated data entry
The same customer, order or project details are typed into more than one system.
Routine reports
Someone exports data, cleans a spreadsheet and rebuilds the same report on a daily, weekly or monthly cycle.
Templated follow-ups
Standard messages depend on a known status, date or missing item but are still sent manually.
Approval chasing
Work pauses while someone asks for, records and follows up an approval through email or chat.
Document generation
Quotes, agreements, certificates or packs are assembled from information that already exists elsewhere.
Status updates
Teams or customers repeatedly ask for progress that could be derived from the system managing the work.
Scheduling
Appointments or assignments are coordinated manually even though availability and rules are known.
Reconciliation
People compare records between platforms to find missing, duplicated or inconsistent entries.
Alerts and exceptions
A person checks a dashboard or inbox for conditions that could instead trigger a targeted alert.
Team handoffs
The next person waits for a manual message even though a clear event means the work is ready.
Score each opportunity before you build
Give each candidate a score from one to five for frequency, time spent, error or delay risk, transaction volume and rule stability. A frequent, time-consuming process with stable rules is usually a stronger starting point than an occasional task that depends on judgement.
Then record delivery effort, integration constraints, data quality and the consequence of a failure. This prevents an attractive time-saving estimate from hiding a fragile dependency or a risky automated decision. Prioritisation should compare likely value with delivery and operational risk.
What should not be automated
Keep people responsible for work that depends on empathy, negotiation, safeguarding, ambiguous evidence or a decision with serious consequences. Automation can assemble information and prompt the next step, but that does not mean it should make the final decision.
Do not automate a process simply because it is frustrating. If the rules change every week, the data is unreliable or no one owns the outcome, first simplify the process and establish responsibility. Otherwise the automation will make the confusion happen faster.
Run a small, observable pilot
Choose one bounded workflow with an owner, a known baseline and a manual fallback. Log what the automation does, make failures visible and review exceptions with the team using it. Measure time, rework and delay before and after rather than relying on a broad claim about productivity.
Once the workflow is reliable, use what you learned to assess the next opportunity. A sequence of well-owned improvements is usually safer and more valuable than attempting to automate an entire operation in one project.
Make the thinking useful.
We can help you review the current situation, decide what needs to change and put the next improvement into practice.
Business automationFirst Steps automation case study