73% of restaurant brands are now investing in AI. Only 9% report meaningful impact. Those are Qu's own numbers from the 2026 State of Digital, and the gap is real.
That gap isn't a technology problem. It's a budget problem. The brands not seeing results aren't using worse AI; they're planning for it the wrong way.
We asked Joel Abdinoor, Qu's CFO, what he'd tell a restaurant finance leader sitting down to budget for AI right now.
What Leaders Need to Know About AI Costs & Budget Forecasting for 2027
Q: What's actually going wrong with AI budgets?
Joel: Finance teams are applying the wrong playbook. We're used to budgeting for restaurant tech a specific way: negotiate a per-location price, sign a contract, put a stable number in the P&L. That's worked for POS, loyalty, and kitchen displays for decades.
AI doesn't behave that way. The more it works, the more it costs. Your best stores, the ones doing the most orders, the most upsell, the most guest engagement, generate the most AI activity. When you apply a capital-budget mindset to what is really an operating expense, the invoice at the end of the quarter is going to surprise you.
Q: How bad does it get?
Joel: About four out of five enterprise finance teams overran their AI budget last year (Cloud Intelligence). And across industries, total AI spend more than doubled year over year. The pricing didn't change much. Usage did. That's the piece most budgets weren't built for.
What happened is that volume grew faster than anyone planned. A brand rolls out AI at the drive-thru; it works, so six months later they add kiosks and online too. Now the same technology is handling three times as many interactions, and the invoice reflects it. Nobody had modeled that compounding, so by Q3 someone opens a spreadsheet and asks what happened.
Q: What should a restaurant operator do differently?
Joel: Two things.
Stop treating it like a capital decision. There's no set-it-and-forget-it AI budget. Review it quarterly, the same way you review labor or food cost, because the economics keep changing. The cost to run a ChatGPT-quality model dropped from $20 per million tokens in late 2022 to $0.07 by late 2024 (Stanford University). Usage grew even faster, so total spend still went up. If you set your AI budget in January and never revisit it, you're guessing by October.
And don't build your budget against today's footprint. Build it for where you'll be in 18 months with twice the locations, twice the channels, peak volume rather than average. That's the number you'll actually be defending.
How to Structure the AI Line in Your Budget
For brands actively deploying AI, four rules that hold up at scale:
- Set a range, not a fixed number. Budget your AI spend based on the difference between what it costs today and what it would cost at peak volume. Review that range quarterly against actual usage, not just the contract amount. The brands that get blindsided are the ones treating AI like a utility bill that never changes.
- Connect it to a revenue number. If your AI spend can't be tied to real business outcomes like check average, order volume, or labor hours saved within 90 days, it's still a pilot. Set that threshold before you scale, not after.
- Know what's included vs. what's extra. Break your AI budget into two lines: what you pay for the platform, and what you pay per use. Those have very different growth curves. When your order volume doubles, your per-use costs can double too. Treating them as one number is how budgets get surprised at quarter-end.
- Budget for upgrades, not just access. AI platforms improve over time. Some vendors include those upgrades in your contract. Others bill for them separately. Know which one you have and put a line in the budget for it every year regardless. The brands that don't are the ones hit with a migration fee they never saw coming.
Three questions to ask before you sign any AI contract
The operators who got this right asked different questions before they committed. Here are the three that matter most:
1. Is AI bundled into the platform fee, or does it generate its own invoice?
Usage-based pricing means every order, recommendation, and interaction costs something extra. Bundled pricing means AI is included in what you already pay. Bundled is predictable. Usage-based is where most budget surprises come from. Ask explicitly before you sign.
2. What does this cost at twice my current scale?
Get this number from your vendor before you commit. Double the locations, double the channels, peak volume. Vendors who've priced this fairly give you the answer quickly. The ones who haven't will need time to "check with the team." That delay tells you something.
3. What does it cost to leave, and what happens to my data?
Exit costs don't appear on any invoice until you need them. Know upfront: what does it cost to end the contract early, can you take your data with you, and who owns the AI trained on your operational history?
The bottom line
The brands seeing real AI results, that 9%, share one thing in common. They evaluated platforms on total cost, not monthly fee. They modeled at scale before signing. And they chose platforms where AI was included in the product, not billed separately on top of it.
Buy right or buy twice.
