Dine Brands Brings AI to 3,500 Restaurants — Quietly

I was on the phone with a franchisee operator in central Ohio when the WSJ piece dropped. He runs three Applebee’s. Mid-sentence — we were talking about labor scheduling — he said, “hold on, I’m getting a Slack from corporate, they’re announcing the AI thing.” Then a beat. “Oh. It’s a chatbot for the printers.”

That is, almost verbatim, the Dine Brands AI strategy. And it’s the right one.

CIO Justin Skelton walked the Journal through a deployment that will eventually reach 3,500+ restaurants across Applebee’s, IHOP, and Fuzzy’s Taco Shop — and roughly 300 franchisees who collectively own most of the system. There is no drive-thru voice agent. No robot expediter. No “AI-powered menu engineering” press release with a Bain logo on it. Instead: a tech-support chatbot for franchisees, a personalization engine for offers, and a manager-assist tool that handles routine back-office tasks. The PYMNTS write-up hit the same morning; Slashdot picked it up by lunch; Restaurant Technology News had the operator-side framing by afternoon.

The contrarian thesis, before we get into the math: the most boring AI announcement in casual dining this year is also the most defensible one. Skelton is not trying to win an analyst-day slide. He is trying to make 300 franchisees less furious on a Tuesday night when the kitchen printer eats a ticket. Everything else flows from that.

Let me show my work.

The printer is doing more work than you think

The anecdote everyone is going to quote — and you’ll see it in three more recaps by Monday — is Skelton’s “broken printer” example. A franchisee walks back, kitchen printer is dead, dinner rush is starting, and instead of calling a help desk and waiting on hold while a tech reads from a script, they ask a chatbot. The bot knows the model, knows the firmware, knows the integration to the POS, knows which of three power cables is the one that wiggles loose on the GP-80 thermal units IHOP rolled out in 2023. It walks the manager through the fix in ninety seconds.

That is unglamorous. It is also where the money is.

Here is the back-of-the-envelope. Dine Brands runs roughly 3,500 restaurants. Assume — and this is conservative — that each location generates two help-desk-worthy IT or equipment incidents per week. That is 7,000 tickets a week, or about 365,000 a year. Industry benchmarks for fully-loaded enterprise help-desk cost per ticket land somewhere between $15 and $22 depending on tier and geography. Call it $18 average.

365,000 × $18 = $6.57M per year in help-desk spend across the system.

A chatbot does not eliminate that line item. Nobody who has actually deployed one believes it will. But the realistic deflection rate for L1/L2 issues in a well-instrumented restaurant IT environment — printers, KDS, network, POS reboots, payment terminal handshakes — is 40-55%. Take the midpoint: 47%.

365,000 × 0.47 × $18 = $3.09M saved per year, gross.

Net out the platform cost (call it $400K-$800K depending on whether they built or bought, and Skelton hinted strongly at “bought, then fine-tuned”), and you’re at $2.3M-$2.7M of annual operating savings. On a system where corporate G&A runs in the $90M zone, that is real but not transformational.

So why is this the right first move? Because the savings is the least important part.

The franchisee-relations dividend

If you have ever sat in a franchisee advisory council meeting — and I have, more than I’d like — you know that the single most corrosive line item in a franchisor-franchisee relationship is unhelpful corporate technology. Loyalty programs that don’t reconcile. POS upgrades that brick the bar terminals on a Friday. A “new ordering platform” that requires three extra clicks to comp a meal.

Dine Brands has 300 franchisees. They are, collectively, the customer Skelton actually has to please — far more than any guest. And franchisees do not measure technology in NPS. They measure it in how angry am I at 9pm on a Tuesday.

A chatbot that reliably solves printer problems at 9pm on a Tuesday is, in franchisee currency, worth more than a personalization engine that lifts marketing yield 4%. It is the difference between “corporate is finally listening” and the next franchisee class-action.

I cannot overstate how often this gets missed by analysts. The casual-dining AI conversation in 2025 has been dominated by guest-facing applications — voice ordering, dynamic pricing, drive-thru. Those are the splashy ones. They also have the highest failure rate, the longest payback, and the most regulatory risk. A forthcoming May piece on the case against the AI premium walks through why public-market investors have been mispricing the customer-facing AI bets in QSR; Dine Brands’ announcement is the inverse trade. Internal tools, boring use case, defensible savings, franchisee goodwill thrown in for free.

The personalization engine is the actual moat

OK, the chatbot is the headline. The personalization engine is the asset.

Both Applebee’s and IHOP run mature loyalty programs — Club Applebee’s relaunched in 2023, IHOP’s MyHOP has been in market longer. Between them, Dine Brands sits on a guest database north of 25 million enrolled members and probably 60 million addressable identifiers when you count anonymized POS data and third-party delivery joins. That is a non-trivial corpus.

What Skelton described to the WSJ is a personalization layer that decides which offer goes to which guest at which time through which channel. The TMC Net writeup has the cleanest version of the framing: the AI is doing offer-targeting, not menu redesign.

The math here is more interesting than the help-desk math, but harder to underwrite from outside. Applebee’s posts roughly $4.3B in system sales in a normal year; IHOP runs around $3.4B; Fuzzy’s is a rounding error. Combined: $7.7B-ish.

Casual-dining loyalty programs typically drive 18-25% of transactions among enrolled guests, and personalization improvements at scale — when they work — move per-member frequency by 3-7%. Let’s split the middle at 5%.

If 22% of $7.7B touches loyalty, that’s $1.69B of loyalty-attributable sales. A 5% lift on loyalty frequency, applied to that base, is $85M of incremental system sales annually. Royalty rate at Dine Brands is around 4% of franchisee sales. Corporate take on that lift: roughly $3.4M of incremental royalty revenue per year.

Add it to the help-desk savings and you’re looking at $5.7M-$6.1M of annual EBITDA contribution from the announced AI stack. On a company that generated $200M-ish in adjusted EBITDA in 2024, that’s a 3% earnings lever. Not heroic. But it’s real, it’s defensible, and — critically — it does not require any guest to talk to a robot.

What’s missing — and that’s the point

Here is what Dine Brands explicitly did not announce:

• No drive-thru voice agent. (Applebee’s barely has drive-thrus. IHOP has takeout. Doesn’t matter.)
• No AI-powered dynamic pricing. (After the Wendy’s “surge pricing” PR disaster of 2024, nobody is touching this for at least another fiscal year.)
• No generative menu redesign. (Olive Garden tried it. The breadsticks stayed.)
• No “AI host” or table-management bot. (Smart. The economics don’t work below $200 average ticket.)
• No labor scheduling AI as a headline item. (Skelton mentioned manager-assist for “routine tasks,” which probably includes scheduling — but he buried it.)

The omissions are more strategically interesting than the inclusions. Skelton is signaling: we are not going to chase the announcements that competitors are chasing. He is implicitly comparing Dine Brands to the more aggressive QSR plays — and an upcoming May piece on Chipotle’s AI stack lays out the opposite philosophy in detail, where guest-facing automation has been the explicit headline. Dine Brands is choosing the back-of-house lane. That is a defensible strategic position for a casual-dining franchisor whose customers come for two-for-$25 and pancakes, not for novelty.

The franchisor math problem nobody talks about

There is a structural reason Dine Brands had to make this move boring, and it’s worth saying out loud.

Dine Brands is ~98% franchised. Corporate does not own restaurants. Which means: corporate cannot unilaterally deploy guest-facing AI at the store level. Every change that touches a guest interaction requires franchisee buy-in, a vendor approval process, and usually a brand-standards committee. That is why you have not seen Applebee’s roll out an AI drive-thru: they couldn’t, even if they wanted to. The governance overhead would eat eighteen months before the first restaurant went live.

What corporate can deploy unilaterally — or with light franchisee approval — is back-office and franchisee-facing tooling. A chatbot for franchisees does not require a vote. A personalization engine that runs on corporate-owned guest data and produces offers that franchisees opt into does not require a vote.

Skelton is playing the only game the franchisor governance structure actually allows him to play. The fact that it happens to also be the highest-ROI, lowest-risk AI use case is a happy coincidence. Or, more likely, the reason he got the CIO job.

Compare this to a fully company-operated chain like Chipotle, where the corporate parent can push a guest-facing change to every register on a Wednesday. Different structural reality, different AI strategy. Both can be right.

The Fuzzy’s question

One last thing, because it matters for the read-through.

Fuzzy’s Taco Shop is the small one. Dine Brands picked it up in late 2022 for around $80M. It’s a fast-casual, ~135 units, mostly franchised, mostly in the Sun Belt. It does not have the scale to justify a bespoke AI stack. Which means: Skelton is using Fuzzy’s as the integration test bed for whatever he deploys at Applebee’s and IHOP.

That is a tell. It says the platform is being built once and ported. It says Dine Brands’ AI roadmap is now a platform play, not a per-brand play. And it says any future tuck-in acquisition — and Dine Brands has been clear they’re shopping — comes with a built-in technology integration story that didn’t exist 18 months ago.

For an M&A-minded reader: this is the move that quietly makes Dine Brands a more credible roll-up vehicle. Not the chatbot. The fact that the chatbot was built to be portable.

Where this lands

So: 3,500 restaurants. 300 franchisees. A chatbot, a personalization engine, and a manager-assist tool. No drive-thru moonshot.

If you graded this announcement on press-release dazzle, it’s a C. If you graded it on franchisee-relations math, it’s an A-. If you graded it on EBITDA contribution per dollar of capex, it’s probably the best AI announcement in casual dining this year — because it’s one of the very few where the dollars actually pencil and the risk is contained.

The market is going to spend the rest of the week comparing this to Chipotle, McDonald’s, and whoever else is in the news cycle. That comparison is going to be unfair to Dine Brands in the short run and very fair to them in the long run.

Casual dining is not QSR. Franchisors are not operators. The right AI strategy for a 98%-franchised system with mature loyalty programs and a 300-franchisee advisory council is, almost by definition, the boring one.

Skelton picked the boring one. That’s why it’s worth writing about.

The Ohio operator I started this piece with? After he read the announcement, he sent me one more text: “If the printer thing works I’ll forgive them for the last loyalty migration.”

That is the entire business case, in one sentence, from the only customer who actually matters.

— Marcus edits The Bottom Line for TableTransfers. Tips: ma@tabletransfers.com.