AI ROI Calculator: Building the Business Case

Who this is for

Business leaders who need to justify AI investment to stakeholders, boards, or themselves, with real numbers rather than vendor promises.

Question this answers

How do I calculate whether an AI project is worth the investment, and what payback period should I expect?

What you'll leave with

A step-by-step framework for calculating AI project ROI
How to quantify the cost of your current process
Realistic impact estimates for common AI use cases
Example calculations for three common scenarios

Why calculate AI ROI?

Every AI vendor will tell you their solution delivers "massive efficiency gains" and "transformative outcomes." But when it comes time to sign the proposal, you need actual numbers, not enthusiasm.

An ROI calculation does three things:

Justifies the spend. To your board, your CFO, or yourself. A $60K AI project is easy to approve when the payback is 8 months.
Right-sizes the scope. If the ROI doesn't work for a $100K project, maybe a $30K version targeting the highest-value workflow does.
Sets realistic expectations. Saving 200 hours per year is worth celebrating. Promising 2,000 hours and delivering 200 is a failure of expectation management, not technology.

The ROI framework

The framework has four steps. Each one takes 10–15 minutes with the right people in the room.

Step 1: Cost of the current process

Before you can calculate savings, you need to know what the current process costs. This is almost always measured in staff hours.

Input	How to measure
Hours per week on the task	Ask the team. Track for one week if unsure.
Number of staff involved	Count everyone who touches the process.
Fully loaded hourly cost	Salary + super + leave + overheads ÷ working hours. Typically $50–$80/hr for admin, $80–$150/hr for professionals.
Error/rework cost	How many errors per month? What does each one cost to fix? Include downstream impacts.

Formula: Annual process cost = (hours/week × staff × hourly cost × 48 weeks) + (errors/month × error cost × 12)

Step 2: AI impact estimate

How much of the current cost will AI eliminate? This is the hardest number to estimate, so use conservative benchmarks:

AI use case	Conservative estimate	Optimistic estimate
Document processing (invoices, forms)	50% time reduction	75%
Email triage and routing	40% time reduction	65%
Knowledge search (RAG)	30% time reduction per lookup	60%
Approval workflow automation	35% cycle time reduction	55%
Report generation	60% time reduction	80%
Data entry automation	55% time reduction	75%

Use the conservative estimate for your business case. If the ROI works with conservative numbers, it'll definitely work in practice. If you need optimistic numbers to justify the project, the project is marginal.

Formula: Annual savings = Annual process cost × impact percentage

Step 3: Total investment

Include everything, not just the build cost:

Cost item	Typical range
Discovery and scoping	$3K–$8K
Development and deployment	$15K–$70K (depending on complexity)
Annual hosting and infrastructure	$3K–$10K/year
Annual maintenance and updates	$5K–$12K/year
Internal staff time during project	40–80 hours (valued at their hourly cost)
Training and change management	$2K–$5K

Formula: Year 1 total cost = Build cost + hosting + maintenance + internal time + training

Step 4: Calculate ROI and payback

ROI: (Annual savings − Annual ongoing costs) ÷ Year 1 total cost × 100

Payback period: Year 1 total cost ÷ (Annual savings − Annual ongoing costs) × 12 months

A healthy AI project typically shows:

ROI of 100–300% over the first year
Payback period of 4–12 months
Year 2+ ROI improves significantly (no build cost, only ongoing costs)

Example calculations

Example 1: Invoice processing automation

A mid-size business processes 200 supplier invoices per month. One AP clerk spends 25 hours per week on data entry and matching.

Current annual cost: 25 hrs × $55/hr × 48 weeks = $66,000
Conservative AI impact: 55% time reduction = $36,300 savings/year
Year 1 investment: $35K build + $6K hosting/maintenance + $4K internal time = $45,000
Payback: 45,000 ÷ (36,300 − 6,000) × 12 = 17.8 months

Marginal in year 1, but Year 2 ROI is strong with only $6K ongoing costs against $36K savings. Also factor in error reduction and faster payment cycle benefits.

Example 2: Email triage for professional services

A firm receives 150+ emails per day across shared inboxes. Two admin staff spend a combined 4 hours per day sorting, forwarding, and responding.

Current annual cost: 20 hrs/week × $50/hr × 48 weeks = $48,000
Conservative AI impact: 45% time reduction = $21,600 savings/year
Year 1 investment: $22K build + $4K hosting/maintenance + $3K internal time = $29,000
Payback: 29,000 ÷ (21,600 − 4,000) × 12 = 19.8 months

Also marginal purely on time savings, but add the value of faster response times (won leads, happier clients) and it becomes compelling.

Example 3: Internal knowledge search (RAG)

An organisation with 50 staff who each spend 30 minutes per day searching for internal procedures and policies.

Current annual cost: 50 staff × 2.5 hrs/week × $70/hr × 48 weeks = $420,000
Conservative AI impact: 35% reduction in search time = $147,000 savings/year
Year 1 investment: $55K build + $8K hosting/maintenance + $6K internal time = $69,000
Payback: 69,000 ÷ (147,000 − 8,000) × 12 = 6 months

Strong ROI. This is why RAG projects are popular. The time savings multiply across every person who uses the system.

Common mistakes

Using vendor numbers. Vendors cite best-case scenarios from their best clients. Use your own data and conservative estimates.
Ignoring ongoing costs. AI systems need hosting, monitoring, and maintenance. Budget 15–25% of the build cost annually.
Counting hours that won't be recovered. If AI saves someone 30 minutes per day but they just fill the time with other tasks, the saving is real only if you redeploy that capacity deliberately.
Forgetting internal time. Your team will spend time on requirements, testing, feedback, and training. Include it.
Only counting time savings. Quality improvements (fewer errors, better compliance, faster customer response) have real value. Include them even if they're harder to quantify.

Next steps

Pick your most promising AI use case and run through this framework. If the payback is under 12 months on conservative numbers, you have a strong business case. If it's over 18 months, consider whether a smaller scope or different use case would work better.

Need help with the numbers? Book a free consultation and we'll work through the ROI calculation with you honestly, including the scenarios where the answer is "not yet."

Key takeaways

AI ROI is calculated like any other business investment: current cost vs projected savings vs project cost.
The biggest mistake is overestimating AI impact. Use 40–60% improvement, not the 90% vendors promise
Include all costs: build, deploy, maintain, train, and the staff time during implementation
Most well-scoped AI projects pay back in 6–12 months. If the payback is over 18 months, the scope might be wrong
Quality improvements and risk reduction have value too. Don't only count time savings.

AI ROIBusiness CaseAI StrategyCost Analysis