Someone in your business copies customer details between systems every day. Someone else dreads preparing the quarterly report. Meanwhile, several new projects are waiting for clients to send the information needed to get started.
All three sound like candidates for automation. Each has an enthusiastic supporter. With time and budget for one improvement, which deserves attention first?
The answer depends on what the work costs your business, what it prevents from happening and how reliably a better process could run.
The strongest automation opportunities combine a meaningful business benefit with a process you can improve reliably at a sensible total cost. Frequency matters, but so do delays, mistakes, customer experience and the judgement involved.
This guide offers a practical way to compare opportunities and choose a first project you can justify.
Start with the result you want
“We need to automate our admin” is too broad to guide a useful investment.
A clearer objective describes something that should change:
- Customer details reach the right systems without repeated typing.
- New projects are ready to start by their agreed dates.
- Managers receive accurate figures before they make a decision.
- Staff can see which enquiries still need a response.
These outcomes give you something to assess. They also help you recognise when automation is only part of the answer.
Before building anything, ask whether the task still needs to exist. A spreadsheet might duplicate information already available in your customer system. A report might contain pages that nobody uses. A second approval might survive from an old policy that no longer applies.
Remove unnecessary work, simplify what remains and check whether your existing software already offers the function you need. If clients repeatedly omit the same information, a clearer request or a better form may be the first improvement.
Automation becomes more valuable when it supports a process that already makes sense.
Five questions that reveal the best opportunities
1. How much work does this create over a normal month?
A two-minute task can be expensive when hundreds of records pass through it. A frustrating afternoon once a year might be a lower priority.
Count how often the task happens, how many people touch it and the time they actually spend doing it. Include checking, correcting mistakes and moving information onwards. Avoid counting the same work twice when several people describe their involvement.
A useful starting calculation is:
Number of occurrences × average hands-on time = current workload.
Use a representative sample. If work varies, record straightforward cases and exceptions separately. An average based only on the easiest cases will understate the problem; an estimate based on the worst week may exaggerate it.
Then identify which parts could realistically disappear. Reviewing an unusual request may remain necessary even when collecting and organising its details becomes automatic.
2. What is waiting for this task to happen?
Two different clocks run in a business process: the time someone spends working and the time that passes before the next step can begin.
Sending a reminder may take three minutes. Waiting for the information it requests may hold up a project for four days.
Look at what depends on the task. Does a delay prevent someone from preparing a quote, scheduling delivery, starting paid work or issuing an invoice?
Also establish what causes the wait. Automation could help if nobody notices missing information until Friday. It may have less effect if the client needs a week to decide.
Faster administration will only improve the overall process if it addresses a real constraint. Getting every document ready sooner achieves little if the work must still wait a month for an unavailable specialist.
3. What happens when someone gets it wrong?
An incorrect internal label and an incorrect delivery address have different consequences.
Examine recent mistakes. How often do they happen? Who corrects them? Do they create rework, delay a customer or lead to an avoidable expense?
Automation may reduce copying mistakes and missed steps, but its own errors can repeat across many records. The assessment should cover both the manual problem and the checks a replacement process would need.
For example, copying an approved customer record could be straightforward. Automatically changing payment details deserves much stronger controls.
A task with serious consequences may be worth improving even when it happens infrequently. In that case, the automation might collect evidence, flag inconsistencies, and request approval while leaving the final decision to a person.
4. How much judgement does the work require?
Separate routine preparation from the decision itself.
Gathering a customer’s details, finding the relevant documents and assigning a review task may follow clear rules. Deciding whether to offer unusual commercial terms may depend on experience, negotiation and the wider relationship.
Automating preparation can still add substantial value. Consider a person who spends twenty minutes finding and organising information before making a five-minute decision. The decision can stay with them while AI removes much of the preparation.
AI can help when information arrives in emails, documents, or notes that need interpretation. Allow for checking and correction when estimating the benefit. A draft that needs extensive rewriting may save less time than expected.
Aim for a clear division of responsibility: what the system prepares or does, and what a person must decide.
5. Is the process ready to support automation?
A valuable problem is not always ready to automate immediately.
Check whether the necessary information is accessible, the systems can exchange it, and the team agrees on what should happen. Someone also needs responsibility for resolving exceptions and approving changes.
For customer records, that might mean agreeing which system holds the correct address and how an existing customer is recognised. For onboarding, it could mean defining the information required for each project type.
Missing details are often fixable. The distinction is whether preparation is manageable or changes every time someone uses it.
If your team cannot yet agree on when a project is ready to start, a small process improvement should come before automating that decision.
Three everyday tasks, compared properly
The following examples are illustrative. The figures are assumptions to show how the assessment works, rather than results from client projects or estimates of what your business will save.
Example one: copying customer details every day
Suppose a team copies 20 customer records or updates between applications each working day. Each takes two minutes. Across a month with 20 working days, that is 13 hours and 20 minutes of manual work.
The volume is consistent, the information is already digital, and the destination fields are known. This is a promising candidate.
A proposed automation could transfer the required details, check for an existing record and flag incomplete information for review. Before building it, the team would agree which records should move and which system takes priority when details differ.
Assume the new process still requires an average of 30 seconds of staff attention per record, plus an hour of general monitoring each month. That leaves four hours and twenty minutes of work and releases nine hours a month.
Those are estimates to test. If exceptions take longer, the benefit falls. If a connection requires an expensive software upgrade, the costs rise.
Assessment: A strong first pilot if the connections are practical and the released time can be put to useful work. The scope is clear, and the result is measurable.
Example two: preparing an occasional report
Now consider a quarterly report that takes four hours to assemble. Everyone dislikes it, but it consumes 16 hours a year.
Suppose automation could reduce preparation and checking to one hour each quarter. That would free up 12 hours annually before considering any maintenance time.
A substantial custom build could be difficult to justify on that time saving alone. A saved report, reusable template or existing export function might provide most of the benefit at a lower cost.
The answer changes if late or incorrect figures have serious consequences. A report used for an important decision or a required deadline may deserve attention because of its reliability and timeliness.
It also matters whether the report changes each quarter. Automating a stable set of figures is a different proposition from automating analysis that asks new questions every time.
Assessment: Improve the preparation first. Consider a larger automation only when the decision it supports, the consequences of mistakes or the low cost of implementation make the case worthwhile.
Example three: chasing information that holds up new projects
Imagine a business starts 12 projects a month. Staff spend an average of fifteen minutes per project checking what is missing and following up: three hours a month.
That is less manual effort than the daily data copying. However, several projects routinely lose days because nobody realises a required document is missing until the work is due to begin.
A better process could request the correct information when the customer agrees to proceed, maintain a visible checklist, remind them about outstanding items and notify the team when the requirements have been met.
It should distinguish a file being uploaded from a document being checked and accepted. Reminders should also pause when the customer has raised a question, or someone is resolving a problem.
The potential value extends beyond the chasing time. It may help protect booked start dates and make scheduling more dependable.
But that benefit needs testing. Earlier reminders cannot guarantee an earlier response. Starting work sooner does not automatically create additional revenue, particularly if the team is already fully booked.
Assessment: Potentially the most valuable opportunity if missing information is a genuine cause of disrupted work. Measure readiness and delays as well as minutes spent sending emails.
| Task | Main opportunity | What would make it a priority? |
|---|---|---|
| Daily customer data copying | Release recurring staff time and reduce rekeying | Consistent volume, reliable connections and manageable exceptions |
| Quarterly report preparation | Reduce preparation and improve consistency. | A simple implementation or important consequences for late or incorrect reporting |
| Chasing project information | Prevent avoidable delays to starting work. | Evidence that missing information is disrupting schedules and can be collected more effectively |
On these assumptions, data copying offers the clearest first pilot. Document collection could take priority if it substantially affects project delivery. The quarterly report deserves a simpler improvement before a major investment.
The comparison changes when the facts change. That is why a decision based only on how annoying a task feels is unreliable.
Count the full cost and be honest about the benefit
An automation has costs beyond the initial build. Depending on the solution, allow time to understand the process, prepare data, test, train, and handle software subscriptions, usage charges, support, and changes as the business evolves.
Also account for the staff work that remains. Reviewing exceptions, approving decisions and correcting problems belong in the new workload estimate.
Return to the customer-record example. It released nine hours a month after allowing for review and monitoring. At an illustrative total employment cost of £25 an hour, that represents £225 of staff capacity each month.
For comparison, suppose implementation costs £900, and ongoing charges are £75 a month. These are invented figures for the calculation, not a quotation or market benchmark.
Over the first year:
- Released capacity would be valued at £2,700: nine hours × £25 × twelve months.
- Implementation and ongoing charges would total £1,800: £900 + twelve payments of £75.
- The difference would be £900 of estimated net value, assuming the released time can be used productively, and the estimates hold.
That does not mean £900 of additional cash will appear in the bank account. If salaries stay the same, payroll spending has not fallen. The case depends on what happens with the capacity: reducing overtime, avoiding external support, handling more work or giving customers better attention.
Scattered minutes are not always as usable as a clear block of time. Ask the team what would actually change in their day.
Use a conservative scenario too. If the net time released were only five hours a month, its annual value at £25 an hour would be £1,500. The same £1,800 first-year cost would exceed that time value. Other benefits might still justify the project, but the time-saving case alone would be weaker.
Keep different benefits separate. Earlier invoicing may improve the timing of cash receipts without increasing profit. If saved hours enable extra work, avoid counting their full labour value and any resulting profit as separate benefits. Explain how the improvement would produce a return.
Choose a manageable first improvement.
After assessing the candidates, give each one a practical next step:
- Pilot now: The benefit is meaningful, the process is clear enough, and the cost appears proportionate.
- Prepare first: The opportunity is worthwhile, but information, responsibilities or rules need attention.
- Simplify or leave manual: A smaller change meets the need, or the likely benefit does not justify the current cost.
Keep the first pilot focused on an outcome you can observe. “Transfer approved customer records accurately” is easier to evaluate than “automate the entire customer journey”.
Specify what starts the process, what successful completion looks like and which cases need a person. Decide who will notice a failure and how the team will complete the work if the automation is unavailable.
Some temporary failures can be retried; other problems need investigation or corrected information. Recovery should also avoid sending the same message or creating the same record twice.
These arrangements are part of delivering the expected benefit. They help prevent an apparently useful automation from creating a new checking job for the team.
Measure whether everyday work actually improves.
Record a baseline before changing the process. Use a normal period of work and include difficult cases, rather than relying on a single smooth demonstration.
Choose measures that match the reason for the investment:
- Customer records: Staff time per record, missing updates and duplicate or incorrect entries.
- Reporting: Preparation time, corrections and whether the report is ready when needed.
- Document collection: Follow-up effort, time to obtain accepted information and the proportion of projects ready by their planned start date.
Check the overall outcome. A system can report that every step ran successfully while still putting the wrong address in the wrong record. It can send every reminder on schedule while projects remain blocked.
Ask the people doing the work whether the improvement removed effort or moved it elsewhere. Include the customer’s experience where the process affects them.
If the pilot works, you have evidence for expanding it. If it needs adjustment, a limited scope makes that easier. If the benefit fails to materialise, you can stop before committing to a larger build.
Choose the task that changes something important.
The best first automation should clearly explain why it matters: hours released for useful work, fewer costly corrections, more dependable follow-up, or fewer projects waiting unnecessarily.
Before commissioning anything, aim to complete this sentence:
“We want to automate this part of the process because it currently causes this measurable problem. We expect this improvement, we know what people must still do, and we will judge the result using these measures.”
If those details are clear, you have the beginnings of a business case. If they are vague, a closer look at the process will usually be more useful than adding more technology.
For some businesses, the right first step will be removing daily rekeying. For others, it will be getting missing information in before a project is due to start. Both can be worthwhile when the improvement addresses the work that actually holds the business back.
Have a few processes competing for attention? Contact Company Compass to discuss the work, the delays and the systems involved, and explore where automation could make a worthwhile difference.