Why should you start with the cost of doing nothing?
The strongest business cases for automation start not with what the AI will do, but with what the current process actually costs. This requires specificity: not "we spend a lot of time on this" but "we spend 14 hours per week on this, distributed across 3 people at an average fully-loaded cost of $65/hour, which equals $47,000 per year in salary cost alone."
Add the error cost: how often does the current process produce an error, what does correcting that error cost in time and any downstream consequences, and what's the annual total. Add the opportunity cost: what would those 14 hours per week produce if they were redirected to higher-value work — additional client capacity, faster response times, more strategic work that currently doesn't get done.
The baseline cost is almost always larger than leadership expects, because nobody has ever added it up before. Adding it up is the first job of the business case.
How do you build the AI automation ROI model conservatively?
AI business cases get killed by optimistic projections. If you claim 90% time savings and deliver 65%, you've failed the expectation even though 65% is a strong result. Build your projections on the conservative end and you'll have room to outperform.
A reliable framework: estimate 50–70% reduction in manual time for the targeted process (this is a typical range for well-scoped automations), apply it to the baseline cost you calculated, subtract the build cost and any ongoing infrastructure costs, and calculate the payback period. For most automations of meaningful scale, the payback period is under 6 months. For high-volume processes, it's often under 90 days.
What non-obvious benefits should you quantify in the business case?
Beyond direct time savings, automation produces benefits that are real but harder to quantify — and including them strengthens the case. Error rate reduction: if the current process has a 10% error rate and errors cost an average of $200 each to resolve, at 500 processes per month that's $10,000/month in error resolution. A well-built automation with a 1% error rate eliminates $9,000 of that monthly.
Scalability: the automation handles 2x the volume without additional headcount cost. If your business is growing, the cost of not automating compounds — each new hire you need because the manual process doesn't scale is a direct cost that automation eliminates. Cycle time reduction: if your lead intake process takes 4 hours instead of 24, that's a faster response that directly affects close rates. Even a 5% improvement in close rate on the affected leads has a revenue value that belongs in the business case.
How do you address the risk objections finance will raise?
Finance and operations will have three objections: what happens if it breaks, what's the implementation risk, and what's the exit if it doesn't work. Answer all three proactively.
For break risk: every well-built automation includes error handling, alerting, and a human fallback path. If the automation fails, the process reverts to human handling and the failure is immediately visible. You don't discover it a month later. For implementation risk: scope the first build as a focused, time-bound engagement with defined milestones and a clear deliverable. You're not buying an enterprise software suite; you're commissioning a specific build. For exit risk: the system runs in your environment on your accounts. If you shut it down, the process goes back to manual. Zero lock-in.
How do you pick the right first AI project for a business case?
Choose a process where the ROI is large and the build risk is low. High frequency, well-defined steps, clear inputs and outputs, a metric you can measure before and after. This isn't necessarily your most important automation — it's the one that will produce the most convincing proof of concept for the next one. The goal of the first build is to win the argument for the second and third builds as much as it is to deliver value directly.