TL;DR
- A drive-thru AI ROI case that holds up is built from four lines, all per store: labor hours saved, average ticket lift, turnover and cost.
- For each line, use your own inputs, compare them with Hi Auto’s reported range (3-8 labor hours saved per store per day, average ticket +~1.5% with a 1-2% range, and turnover reductions of 17 to 30%, measured before and after deployment against a control group), and answer the CFO’s likely question.
- Run the case at the low end of every range, and set pilot criteria in advance: completion and accuracy together, measured as an average across live stores.
- A good case is one your CFO can rebuild from their own numbers.
The vendor deck ends. The CFO looks up and asks one question: what are your assumptions?
The conversation stalls. The technology may work well. The trouble is a case built on someone else’s averages, with inputs nobody in the room can check and a cost side that is vague. A CFO cannot rebuild that, so a CFO cannot approve it.
This guide gives you a method you can defend line by line. It works for Hi Auto’s AI order taker and for any drive-thru AI you are weighing. Every line is per store, and every line uses your inputs. You will not find a worked example with invented numbers here, or a single blended ROI figure. Those are what a CFO distrusts first.
The four lines
Think of a small sheet with four rows. Each row has the same three parts:
- Your inputs. The numbers only you know, such as your hourly cost, your average ticket and your annual hires.
- Hi Auto’s reference range. A reported figure, always shown as a range and never as a single blended number.
- The CFO question. The pushback that line will face, so you can answer it before the meeting.
Nothing in this guide adds the four lines into one total. That step belongs to you, with your own numbers.
Line 1: Labor hours, priced at your own cost
This is the line a CFO will look at hardest, and the one you can defend most precisely.
Your inputs. Start with two figures:
- The labor hours your drive-thru order taking uses in a day, in one store.
- Your fully loaded hourly cost, meaning base wage plus payroll taxes, benefits and any other cost you carry per employee.
For a reference point on wages, the U.S. Bureau of Labor Statistics (BLS) reports average hourly earnings for food services and drinking places of $21.89 for all employees and $20.20 for production and nonsupervisory employees, as of July 2026, preliminary. Payroll taxes and benefits sit on top of a base wage, which is why Hi Auto’s own labor example is a fully loaded $25/hr. Use your own number in place of either.
Hi Auto’s reference figures. The reported figure, and one illustration:
- 3-8 labor hours saved per store per day.
- For illustration, 4 hours a day at a fully loaded $25/hr (wages, taxes and benefits) is about $3,000 a month per store. Actual savings vary by market and labor cost.
The dollar figure is an example only. Your own hourly cost and your own hours decide the real number.
To build your own line, price the hours at your own fully loaded cost, starting at the low end of the range, and use the result as your conservative case. The illustration above shows the shape of the calculation. Our sales per labor hour post on labor scheduling and profitability covers how labor hours relate to sales.
The CFO question: “How do the hours turn into dollars?” Answer with your own cost per hour, and show your inputs on the page.
Ryan Weaver, CEO of Lee’s Famous Recipe Chicken, described how he looks at it:
“The technology pays for itself with the labor hours we are able to take out of our budget every week, and then the sales lifts, which we think is about 1-2% is just kind of gravy on top of that.”
Ryan Weaver, CEO, Lee’s Famous Recipe Chicken. Nift “Marketing Bites: Restaurant Growth Unwrapped” podcast, Episode 32, 1 April 2025. Watch the clip
Run this line with your own inputs.
Line 2: Average ticket, per store
Your inputs. Your own average ticket size in one store, and your daily orders in that store. Your check average is the same idea by another name, and the glossary covers the difference.
Hi Auto’s reference. Average ticket +~1.5%, within a 1-2% range, per store.
Run your model twice: once at the low end of the range and once at the high end. Keep the two results separate. A blended midpoint looks tidy, but it is harder to defend and harder to stress-test.
Why does the lift happen at all? Lee’s Famous Recipe Chicken reports upsell on 100% of orders with Hi Auto, compared with roughly 20% with human order takers. A person on a headset at a busy window is also handling food, cash and the window, so an offer on every single order is a hard ask of the job. The business case only needs the outcome, and you do not need to model how the offer is chosen.
The CFO question: “Is that range an average across stores, or a best case?” Ask the vendor for the basis: which stores, which period, and how price changes and new menu items were handled. Until you hear back, use the low end.
Run this line with your own inputs.
Line 3: Turnover, priced at your own replacement cost
Turnover savings are real, but they arrive slowly, which makes them easy to leave out of a case.
Your inputs. Two numbers, both yours:
- Hires per store per year for the drive-thru positions. Use the count of hires, not a rate.
- What one replacement costs you, fully loaded: recruiting, onboarding, training time and the dip in output while a new person learns the job.
We do not offer a replacement cost here, because it depends on your market, your training model and how you count manager time. Use yours. Our glossary entry on employee turnover in QSR defines the term.
Hi Auto’s reference. Hi Auto reports turnover reductions of 17 to 30%, measured before and after deployment against a control group. Use the range, and never treat it as typical.
The CFO question: “What is that reduction measured against?” Ask the vendor for the measurement method and the baseline period before any figure goes into the model, and apply the low end of the range to your own baseline.
Plan for a lag. Turnover savings show up in recruiting spend and training time over quarters, not in the first weeks. Build that into the timeline so the CFO is not surprised.
Line 4: The cost side
A CFO will not approve a case with a blank cost line.
Ask any vendor for a written quote that covers every cost, before you build the model.
Then add the internal costs that often get missed:
- A POS system. Confirm that your POS works with the AI order taker before the pilot starts. Our product page describes how the system connects.
- A drive-thru audio system. The equipment at the order point has to work with the setup.
- Reliable internet. Each store needs a steady connection.
- Team time. Onboarding, training and the pilot all take your people’s time, even when no invoice arrives.
A CFO thinks in total cost, not only the vendor price. List every cost before you call the model complete.
How to stress-test the case
Two habits separate a defensible case from a hopeful one.
Run the model at the low end of every range. Rebuild it with 3 labor hours a day, the low end of the ticket lift and the low end of the turnover reduction. If the case still works, it is defensible. If it only works at the high end, your CFO will find that before you do. This discipline earns trust with a finance team.
Set pilot success criteria before the pilot starts. Write down a threshold for completion and one for accuracy. A system with high completion and low accuracy produces wrong orders at scale. A system with high accuracy and low completion needs staff to step in more often. Neither delivers the labor savings you are building the case around. Reliability at scale means both numbers hold across all live stores.
Across 100M+ orders processed per year, Hi Auto reports 93%+ completion and 96% accuracy. When you evaluate any vendor, ask for their average across all live stores, not their best locations, and ask how they measure it. Hi Auto’s Buyer’s Guide advises asking every vendor for the measurement methodology, because the method matters as much as the number.
Measure your pilot as an average across all pilot stores, and agree the decision in advance. If the pilot meets both criteria, the scale decision is ready to go. If it does not, you learn that before a chain-wide commitment.
Timeline and what the CFO signs off on
A Hi Auto pilot goes live in about 8 weeks, and getting from pilot to scale takes 90 to 120 days. Treat these as planning inputs, since every POS setup, audio setup and approval process is different. Put a decision date in the case, so the CFO can see when the pilot data arrives and what happens next.
The business case can fit on one page. Here is a layout, with no numbers filled in:
| Row | What goes in it |
|---|---|
| Labor | Your inputs, Hi Auto’s reference range, your estimate |
| Ticket | Your inputs, Hi Auto’s reference range, your estimate |
| Turnover | Your inputs, Hi Auto’s reference range, your estimate |
| Cost | Vendor cost from the written quote, plus internal costs |
| Pilot criteria | Completion and accuracy thresholds, measured as an average across live pilot stores |
| Timeline | Pilot live date, scale decision date |
A page built this way invites scrutiny, which is exactly what makes it credible.
The shift in thinking: a case your CFO can rebuild
The strongest business case has the most checkable inputs, whatever the size of the number. When your CFO can rebuild each line from your store data and the reported ranges, the meeting becomes a review of assumptions the CFO already understands.
Call to action
Build the four lines with your own numbers, then talk to the Hi Auto team about pricing in writing and the reported ranges.