How to Build an Honest ROI Model for AI Automation

Before you invest in any AI system you should be able to model the financial impact yourself. This walks through the inputs, the assumptions, and the formulas, and it is deliberate that every assumption in it is one you set rather than one we hand you.

Start with your current baseline. You need four numbers: monthly lead volume, contact rate (the share of leads you actually reach), appointment rate (the share of contacted leads that book), and close rate (the share of appointments that become deals). If you do not know them, pull ninety days out of your CRM before you go any further. A model built on guessed baselines is worse than no model, because it feels like evidence.

Next, establish your current cost structure. Add up everything you spend on lead follow-up: wages, your own management time valued at a rate you would genuinely accept, software subscriptions, phone costs, and anything else attached to the function. That total is what any alternative has to be compared against.

Now the honest part. To model the improved scenario you have to pick an assumption about what changes, and you should pick it yourself. Industry-wide multipliers are handed around freely and almost none of them come with a sample, a date range, or a definition, which makes them unusable for a decision you are betting money on. Set a figure you would be willing to defend to a partner, then set a second one at half that, and look at both.

The formulas are simple. Deals a month today equals lead volume multiplied by contact rate multiplied by appointment rate multiplied by close rate. The modelled version applies your assumption to whichever variable you believe changes, and holds everything else exactly where it is. Be explicit about which variable you moved, because moving contact rate and moving close rate are very different claims about what a system does. Automation changes what happens before the conversation. It does not change what happens inside it.

Multiply the difference in deals by your average assignment fee to get the modelled revenue difference. Then subtract the system cost (our published plans are $497, $697, or $997, plus metered usage at our published rates) and add back whatever cost you actually remove on the human side. That is your model.

Run it twice: once at your assumption and once at half of it. If the decision still makes sense at half, it is a decision. If it only works at the optimistic figure, you have not built a model, you have built a hope. The calculator on this site is built the same way, with every field empty until you fill it, for exactly this reason.