How to Calculate Whether an Automation Project Is Worth It

Insights · 10 min read

An automation idea can sound obviously worthwhile: save ten minutes here, remove a spreadsheet there, send reminders automatically. The difficulty begins when someone asks, “What will this actually return?”

Return on investment is often presented as one confident percentage. Real operations are less tidy. Time saved does not always become cash saved. Employees need training. Exceptions still require attention. Licences renew. A vendor may quote only the initial build while leaving out internal preparation and future maintenance.

A useful automation business case should therefore do more than justify a purchase. It should help the organisation decide what to automate, how much to spend, what to measure and when to stop.

IMDA reported that Singapore SMEs adopting AI-enabled solutions under the Productivity Solutions Grant achieved average cost savings of 52% in 2024. That is encouraging evidence of potential value across supported adopters, but it is not a guaranteed return for a particular business. Your result will depend on workflow volume, implementation quality, adoption, risk and what happens to the capacity released.

This guide provides a practical calculation method that an SME can use before committing to a project.

Start with the decision, not the percentage

Before opening a spreadsheet, define what decision the analysis must support. Common questions include:

• Should we proceed with this automation at all?
• Should we build a custom workflow or use an existing product?
• Which of three candidate processes should go first?
• Is a larger scope justified after a pilot?
• Did the project deliver the result originally approved?

The required precision depends on the decision. A two-week pilot may need a simple range. A multi-year contract affecting sensitive operations needs a more detailed model, risk assessment and scenario analysis.

Avoid deciding the answer first and building a spreadsheet to support it. Record assumptions and use ranges where evidence is weak.

The core automation ROI formulas

Use four basic calculations.

1. Total annual benefit

Annual benefit = usable labour capacity + avoided error cost + avoided delay cost + other measurable benefit

“Usable” is important. If a workflow saves 300 hours, the business only receives value when those hours reduce overtime, avoid hiring, increase billable capacity, improve customer response or allow important work to be completed.

2. Total cost

Total cost = implementation + licences + internal project time + training + maintenance + expected risk cost

Calculate first-year and recurring costs separately. Implementation may be one-off; licences and support continue.

3. Return on investment

ROI = (total benefit − total cost) ÷ total cost × 100%

An ROI of 25% means the net benefit equals one quarter of the cost over the period measured. Always state the period: first-year ROI and three-year ROI can tell very different stories.

4. Payback period

Payback period = initial investment ÷ monthly net benefit after launch

This simplified formula works when monthly benefit is reasonably stable. For seasonal businesses or phased deployments, calculate cumulative cost and benefit month by month.

Step 1: Establish the current-state baseline

Do not estimate benefits from memory alone. Observe the current process for a representative period and capture:

• Number of cases.
• Active handling time per case.
• Waiting time between steps.
• Number of people involved.
• Error and rework frequency.
• Escalations and missed deadlines.
• Overtime, temporary support or outsourced cost.

Separate active handling time from waiting time. If an invoice waits two days for approval, automation may improve cycle time without saving 16 hours of labour. Both benefits matter, but they are valued differently.

Use a sample large enough to include exceptions. Measuring only the cleanest week will overstate performance. Note seasonal peaks, month-end routines and periods when a key employee is absent.

Step 2: Calculate labour capacity released

The basic formula is:

Annual hours released = cases per year × minutes saved per case ÷ 60

Then multiply by a realistic loaded hourly cost. Loaded cost may include salary plus employer contributions and other employment costs relevant to the decision.

Next, apply a realisation factor. This is the percentage of released time that the business expects to convert into something valuable.

Suppose four employees each spend 25 minutes a day on repetitive data entry over 220 working days:

• 4 employees × 25 minutes × 220 days ÷ 60 = approximately 367 hours.
• At S$35 per loaded hour, the theoretical value is about S$12,845.
• If the organisation expects to use 60% of the time for customer work and higher-value administration, usable capacity is about S$7,707.

The 60% factor prevents a common mistake: treating every saved minute as money in the bank. Time arrives in small pieces. Five minutes saved between calls may be less usable than half a day released from a monthly reporting routine.

Describe how capacity will be used. “Staff can focus on higher-value work” is not enough. Name that work: respond to enquiries within one day, process 15% more cases without hiring or remove six hours of monthly overtime.

Step 3: Value errors and rework

Automation may create more value through consistency than speed. Estimate:

Avoided error benefit = current errors × average correction cost × expected reduction

Correction cost can include employee time, credits or refunds, additional delivery cost, management review and customer recovery. Use direct evidence when available.

For example, if a process produces 24 preventable errors a year and each costs an average of S$150 to investigate and correct, the current annual cost is S$3,600. If a controlled workflow is expected to reduce those errors by two-thirds, the model includes S$2,400 in annual benefit.

Be careful with severe but rare events. Do not assign a dramatic number without a defensible probability. Keep high-impact risks visible in a separate risk assessment even when the expected monetary value is uncertain.

Step 4: Value speed and visibility

Faster cycle time can improve cash flow, customer experience or operational planning. However, not every day saved has a direct dollar value.

Possible measurable benefits include:

• Invoices issued earlier and paid sooner.
• More enquiries contacted before they go cold.
• Fewer expedited shipping or late-processing charges.
• Lower backlog during peak periods.
• Reduced time spent asking for status.

Choose the measure closest to the business outcome. If an approval workflow reduces turnaround from four days to one, track whether this changes order completion, revenue timing or customer complaints. Do not invent a financial value merely to make the business case larger.

Some benefits should remain non-financial but still influence the decision: a better audit trail, reduced dependency on one employee or improved service continuity. List them separately and explain their importance.

Step 5: Count the complete cost

An incomplete denominator makes every ROI look attractive. Include the following categories.

Implementation cost

This includes discovery, process mapping, configuration, development, testing, documentation, deployment and project management. If a vendor offers a fixed price, clarify what counts as a change request.

Licence and infrastructure cost

Include platform subscriptions, per-user licences, usage charges, storage, connectors, AI consumption and any required higher-tier plan. Model likely growth in volume.

Internal project time

Employees will attend workshops, prepare data, test cases, review outputs and answer questions. Multiply estimated hours by loaded cost. Internal time is not free merely because no invoice is issued.

Training and adoption

Include user training, manager communication, revised procedures and the temporary productivity dip while people learn the new flow. If the automation changes responsibilities, allow time to redesign work—not only demonstrate buttons.

Maintenance and change

Business rules, templates, staff and connected systems change. Budget for monitoring, incident response, small improvements, licence administration and periodic review. A workflow that nobody can maintain becomes a future replacement project.

Risk and fallback

Consider expected downtime, incorrect processing, privacy incidents, vendor dependency and manual fallback effort. For high-consequence workflows, security and governance are implementation requirements, not optional overhead.

A worked first-year example

Consider the data-entry workflow above.

Estimated annual benefits

• Usable labour capacity: S$7,707.
• Avoided error cost: S$2,400.
• Avoided late-processing and follow-up cost: S$1,200.
• Total annual benefit: S$11,307.

Estimated first-year costs

• Process mapping and implementation: S$4,500.
• Internal testing and project time: S$1,000.
• Training and documentation: S$700.
• Annual licences: S$1,800.
• Support and maintenance allowance: S$1,000.
• Total first-year cost: S$9,000.

The first-year net benefit is S$2,307. The first-year ROI is:

(S$11,307 − S$9,000) ÷ S$9,000 × 100% = approximately 25.6%.

The recurring picture may be stronger because the implementation cost does not repeat in full. If benefits remain stable and recurring annual cost is S$2,800, the second-year net benefit would be about S$8,507 before any major change cost.

This example is illustrative. Its most important feature is not the percentage; it is the separation of assumptions. A reviewer can challenge the realisation factor, error reduction or maintenance allowance without rebuilding the entire model.

Use three scenarios, not one forecast

Create conservative, expected and optimistic cases.

The conservative scenario might assume lower adoption, less time saved, more maintenance and a slower launch. The optimistic scenario may assume higher volume and faster adoption, but should remain plausible.

If the project only looks worthwhile in the optimistic case, run a smaller pilot or reconsider the scope. If it remains worthwhile in the conservative case, the decision is more resilient.

Test the variables most likely to change:

• Volume.
• Minutes saved.
• Realisation factor.
• Error reduction.
• Licence growth.
• Implementation delay.
• Ongoing support cost.

This sensitivity analysis reveals what must be measured during the pilot.

Do not confuse capacity, cost savings and revenue

These terms are often mixed together.

Capacity released means employees have time available for other work. It is valuable but does not automatically reduce expenditure.

Cost savings means the organisation spends less—for example, lower overtime or an avoided outsourced fee.

Cost avoidance means future spending, such as an additional hire, may no longer be necessary.

Revenue contribution means the released capacity enables more sales or billable delivery. Attribute only the portion reasonably connected to the automation and avoid counting the same benefit twice.

A credible business case labels each category honestly.

Define the measurement plan before approval

The project should specify how benefits will be verified after launch. Record the data source, owner, baseline period, review dates and target for each measure.

A practical scorecard may include:

• Median handling time.
• Cases completed per employee.
• Error or rework percentage.
• Average cycle time.
• Number of exception cases.
• Hours spent maintaining the workflow.
• User adoption rate.
• Customer or staff feedback.

Review at 30, 60 and 90 days. Early results may be distorted by training and cleanup, so avoid declaring success after one good week. Equally, do not let an underperforming workflow continue indefinitely because the original business case looked strong.

When an automation can be worth doing without a high ROI

Financial return is not the only valid reason. A project may be justified because it improves safety, data protection, auditability, service continuity or regulatory compliance. It may remove a single point of failure or create the operational foundation for later growth.

In these cases, state the primary objective clearly and choose the least costly approach that meets it. Do not disguise a risk-reduction project as labour savings.

Make the business case testable

The best ROI model is not the one with the largest number. It is the one that helps the business learn.

Use real baseline data. Discount theoretical time savings. Count internal effort and maintenance. Separate financial and non-financial value. Show conservative and optimistic scenarios. Most importantly, connect each assumption to a measure that can be checked after launch.

An automation project is worth it when the expected operational improvement exceeds the complete cost and risk—and when the organisation has a credible plan to realise that improvement.

That conclusion is more useful than “automation saves time.” It tells the team what must be true, what to monitor and when to change course.

Sources and further reading

• IMDA: Singapore Digital Economy Report 2025 highlights
• Enterprise Singapore: Productivity and digitalisation