How To Write An RPA Business Case: ROI And Payback For An Automation Pilot

Quick answer: An RPA business case is a decision document that asks executives to fund a software-robot pilot. It proves three things: the current process has costly, measurable pain points; the bot fixes them with people still in control; and the numbers work. Show volumes, a TCO table, a benefit table, then ROI and payback with formulas and stated assumptions.

That is the short version. The longer version is where most students get stuck, because the brief usually hands you an interview transcript rather than a spreadsheet. You have to turn “about ten minutes” and “maybe twice a month” into a cost model a finance director would trust.

This guide uses the SMU ACCT433 business case document brief as the running example. The scenario is an accounts receivable team that builds past-due customer lists by hand from Xero and Excel, while Account Managers type each reminder email.

Short on time? Send us your brief and our accounting and information-systems experts will walk you through your own assumptions and calculations.

What an RPA business case must contain

Most RPA business case briefs follow the same six-part logic, whether they come from a university module or a corporate PMO. Each section feeds the next, so a weak current-state section quietly breaks your ROI three sections later.

Notice what is missing: bot architecture, selectors, UiPath or Power Automate screenshots. The ACCT433 brief says explicitly that the focus is why to invest. Markers reward restraint here.

Step 1: Turn the interview into a current-state cost model

The current-state section of an RPA business case is where you earn or lose the whole paper. Start by listing every number the interview gives you, then fill gaps with assumptions you label clearly.

Build a driver table first

In the ACCT433 scenario the interview gives you roughly 200 Account Managers, at least two requests each a month, about 10 minutes of AR time per request, around 50 customers per AM, and 5–8 minutes per reminder email. It does not tell you how many customers are overdue. That gap is yours to fill – say 20% of 50, so 10 customers per request – and you must flag it as illustrative.

1. Requests a year: 200 × 2 × 12 = 4,800.
2. AR time: 4,800 × 10 min = 800 hours.
3. AM time: 4,800 × 10 emails × 6.5 min = 5,200 hours.
4. Rework: an assumed error rate × emails × minutes to fix.
5. Multiply hours by a loaded hourly rate for each role.

The surprise in this case is that Account Managers carry about 80% of the cost. A student who only automates the AR list misses most of the value. That kind of insight is what separates a “Good” from an “Excellent” current-state analysis.

Mark the pain points on the As-Is map

The rubric wants the pain points on the map itself, not in a separate paragraph. Number them (PP1, PP2…) on the boxes where they happen: the manual master list, matching by customer name with no ID, missed rows during filtering, informal exception handling, and template emails sent with the wrong name or address. Swimlanes for AM, AR Clerk and the systems (Xero, Excel, Gmail) make the hand-offs visible.

Step 2: Design the bot and the human-in-the-loop split

RPA suits structured, rule-based, repetitive work that runs across existing applications – software robots operating the same screens a person would (van der Aalst et al., 2018). It does not suit judgement. Your solution section should show you know the difference.

Keeping the send decision with the AM is not a weakness. The transcript says AMs sometimes hold back a reminder to protect a relationship. Automating that away would make the solution look naive, and the review literature repeatedly flags data quality and governance as the places RPA projects stumble (Syed et al., 2020).

Step 3: Cost the pilot properly with a TCO table

In an RPA business case, total cost of ownership is every cost of running the bot over your chosen horizon, not just the licence. Use the six categories most briefs list: software licensing, infrastructure or cloud, implementation and development, testing and deployment, training and change management, and ongoing maintenance and support.

Split one-off from recurring costs. Show the basis for each line (days × day rate, users × hours × rate). Never present an invented figure as a vendor quote; call it an illustrative estimate and, if you can, point to a published price list as the source of your licensing model. In the worked sample, one-off costs come to S$55,000 and recurring costs to S$38,600 a year, giving a three-year TCO of S$170,800 – all illustrative.

Step 4: Value the benefits without overclaiming

Here is my strong view: the single most common error in a student RPA business case is treating every saved hour as cash. Freed staff time is capacity. Unless someone leaves or a hire is avoided, payroll does not fall.

Research on RPA in accounting firms found that automation mostly changed what staff did rather than removing jobs (Cooper et al., 2019). So, in your RPA business case, apply a benefit realisation factor – 50% is a defensible base case – and say why. Then run a sensitivity line. In the sample, gross labour value released is S$279,840 a year; at 50% realisation the business case counts S$139,920.

Also report the operational benefits the brief asks for: hours saved (4,864), FTE capacity released (about 2.7 at 1,800 hours per FTE), processing time per request (75 minutes to 19), and rework reduction. Add non-financial benefits – audit trail, consistency, staff and customer experience – briefly.

Step 5: Calculate ROI and payback period in your RPA business case

RPA ROI is the net benefit over a stated period divided by total cost over the same period. Payback period is the time until cumulative net benefits recover the up-front investment. Both must use figures that already appear in your cost-benefit tables, with the horizon and assumptions written beside the formula.

ROI worked example

ROI = (Total benefits − Total costs) ÷ Total costs × 100. Over three years, with benefits starting in month 4: (384,780 − 170,800) ÷ 170,800 = 125.3%.

Payback worked example

Investment at go-live = S$55,000 one-off + three months of running costs = S$64,650. Monthly net benefit after go-live = (139,920 − 38,600) ÷ 12 = S$8,443. Payback = 64,650 ÷ 8,443 ≈ 7.7 months after go-live, or about 10.7 months from project start.

A sensitivity line strengthens the RPA business case recommendation: in this model the three-year ROI only reaches zero if less than about 22% of the released value is realised. You can see the full layout in the ACCT433 RPA business case sample (for reference only). All figures in this guide are illustrative assumptions, not vendor quotes or real company data.

Stuck on the cost model or the map? Submit your RPA business case brief for guidance.

Where students lose marks on an RPA business case

One pattern I see every term: students who work full time draft the summary first, then bend the numbers to fit. Do it the other way round. Write the executive summary last, once the payback figure actually exists.

If you want a second pair of eyes on your figures, our accounting assignment help team reviews calculation trails, and our business process management assignment help experts can check your As-Is and To-Be maps.

Final checklist before you submit

• Executive summary states the process, the cost today, the pilot, ROI, payback and a vision statement
• As-Is map with numbered pain points on the steps
• Driver table with sources: interview vs illustrative assumption
• Bot vs human table, every pain point mapped to a fix
• TCO across the relevant cost categories, one-off and recurring split
• Benefits in hours, FTE and dollars, with a realisation factor
• ROI and payback formulas, values, time period and assumptions
• A clear proceed / do not proceed recommendation with conditions
• File named as the brief requires, AI-use declaration completed

Frequently asked questions

How do I calculate ROI for an RPA business case?

Take total benefits over a stated period, subtract total costs over the same period, divide by total costs and multiply by 100. Use benefits that are already in your cost-benefit table, state the horizon (often three years), and explain any realisation factor you applied to labour savings.

What is a good payback period for an RPA pilot?

There is no fixed target in most university briefs. What matters is that the payback is calculated correctly and traced to your tables. Many organisations look for payback within the first year or two, so show the figure from both go-live and project start, and test it with a pessimistic case.

Can I make up numbers if the interview doesn’t give them?

You can make assumptions, but you must label them clearly as illustrative and explain the reasoning, for example “20% of 50 customers are past due”. Never present an assumption as a fact or a vendor quote. Markers reward transparent assumptions and penalise figures that appear from nowhere.

Should the bot send reminder emails automatically?

Usually not in a first pilot. Where staff use judgement – such as an Account Manager holding back a reminder for a valued customer – keep a human approval step. Let the bot draft the emails accurately and let the person who owns the relationship decide and send.

Do I need to explain how the bot is built?

No. A business case explains why to invest, not technical design. Describe which steps are automated and which stay with people, and how that fixes each pain point. Technical detail belongs in a later process design document, not in this paper.

Can I use ChatGPT for my ACCT433 business case?

Only within your course rules. The ACCT433 brief requires an AI-use declaration and a usage record of prompts and outputs, and it allows only fictional or supplied data in AI tools. You remain accountable for every word and figure, so check and declare any AI help honestly.

Get the numbers right before the deadline

The fastest way to lose marks on an RPA business case is a payback figure nobody can recalculate. Build the driver table tonight, before you write a word of the summary. Our experts provide model answers and structured guidance – you stay responsible for your own submission, and your files and identity are kept confidential. Submit your RPA business case brief, explore SMU assignment support or read student reviews.

About the expert

William Tan holds a Master’s degree in Accountancy from the National University of Singapore and, according to his profile, works full-time as an accountant with one of the biggest accounting firms in Singapore. His profile lists financial accounting, cost accounting, auditing and taxation among the topics he has written on, and 8 years of academic writing.

References

Cooper, L. A., Holderness, D. K., Sorensen, T. L., & Wood, D. A. (2019). Robotic process automation in public accounting. Accounting Horizons, 33(4), 15–35. https://doi.org/10.2308/acch-52466

Syed, R., Suriadi, S., Adams, M., Bandara, W., Leemans, S. J. J., Ouyang, C., ter Hofstede, A. H. M., van de Weerd, I., Wynn, M. T., & Reijers, H. A. (2020). Robotic process automation: Contemporary themes and challenges. Computers in Industry, 115, Article 103162. https://doi.org/10.1016/j.compind.2019.103162

van der Aalst, W. M. P., Bichler, M., & Heinzl, A. (2018). Robotic process automation. Business & Information Systems Engineering, 60(4), 269–272. https://doi.org/10.1007/s12599-018-0542-4