How Presto runs proactive drive-thru voice AI with Bluejay

About Presto A leading voice AI company for drive-thru automation in quick-service restaurants. Presto takes orders for brands including Carl’s Jr., Hardee’s, Taco John’s, Wienerschnitzel, and Dairy Queen, live in hundreds of locations across the US and Canada and expanding fast. presto.com

Presto puts voice AI in the hardest place to run it: the drive-thru. Its agents take live orders at hundreds of quick-service restaurant locations across the US and Canada, for brands spanning burgers, hot dogs, tacos, seafood, and ice cream, across every kind of accent, integration, and lunch-rush chaos the physical world can produce.

The drive-thru is also the definition of last-mile AI. Roughly 70% of the quick-service restaurant industry runs through it, it operates day in and day out, and when something drifts, the impact rolls across stores by the hour. There is no batch job to re-run. There is only the next car.

“Bluejay is a foundational part of our observability layer. We are very focused on delivering AI to the last mile in the restaurant world, and with such a physical environment, it’s critical for us to be proactive versus reactive.”

Today, Bluejay analyzes millions of conversation minutes from Presto’s drive-thrus every single month, turning one of the most analog channels in retail into something Presto, and its restaurant brands, can actually measure.

The problem: the drive-thru is analog

Before AI, drive-thru observability meant standing in the parking lot. “Prior to any kind of AI going into the drive-through, you literally have to physically go to the drive-through and observe and sample,” Gupta said. “That’s the best kind of observability you can have.”

Voice AI changes that, but only if you can see inside the conversations. So as Presto entered 2026, Gupta made observability a core commitment: as CEO of a business that touches the physical world continuously, he wanted to see problems ahead of time, not after the impact had rolled across the system.

“Some point in the very near future, this will be a digital channel that can be run in the way that e-commerce can be run,” he said.

Why Bluejay: audio plus transcripts, and a point of view

Presto evaluated the market of observability players before committing. “We found Bluejay’s thinking and tooling to be the most sophisticated and advanced,” Gupta said. Two capabilities stood out.

The first is that Bluejay couples audio and transcript analytics together, “in a way that I think very few, if anybody else, does,” Gupta said. “A lot of other analytics or observability tools lean mostly on transcript. Layering in the audio for me is very interesting and important.” In a drive-thru, what a customer heard, and how, matters as much as what was said.

The second is a perspective on what actually matters. A phone call booking a hospital appointment, an order at a drive-thru, and a call to an HVAC contractor are three different kinds of conversation, and Gupta has little patience for tools that treat them the same.

“The wrong kind of observability company says: here’s the set of tools you can use to observe this thing, a one-size-fits-all approach. The right kind of observability company says: we understand there are nuances in each of these conversations, and we have to develop different metrics and really be thoughtful.”

The partnership sealed it. Bluejay has been forward-deployed with Presto from day one: spending a week at a time in Presto’s San Mateo offices, meeting Presto’s restaurant customers on the front lines, and co-developing the metrics alongside Presto’s own industry knowledge. “I fly out the minute one of our customers asks us to come,” Gupta said. “The fact that Bluejay embraces that same ethos has been a great testament to the kind of DNA they have as a company.”

Custom evals, from the CEO’s dashboard down

Working with Bluejay, Presto develops custom evals tailored to the drive-thru: metrics that let the team observe and measure parts of the business that were never observable or measurable before. Those metrics now sit inside Presto’s core internal dashboard, the set of numbers Gupta reads to run the company every day.

“There’s the old adage: you can’t manage what you can’t measure,” Gupta said. “That applies very well to a world where you’re managing a bunch of AI agents. You can’t really manage them if you can’t measure them well.”

The usage runs the whole org chart: Gupta digs into the analytics personally, and the commercial, customer success, and product teams all spend serious time in the Bluejay dashboard.

“Evals are an important part of how we function at Presto. We leverage them to find the needles in the haystack that simple deterministic systems just cannot discover at scale.”

Proving fixes work: interruptions, pre and post

Proactive only counts if you can close the loop, and interruptions are the clearest example of how Presto does it.

Presto works across many brands, integrations, and end customers, and how fluently those customers speak to an AI differs from place to place. In a handful of locations, interruptions, the agent and the customer talking over each other, had become a real issue. The team had ideas for fixing them. The question was proving the fixes worked.

“For us to be able to launch those tools and have Bluejay observe the impact pre and post was very valuable,” Gupta said. “We had a handful of interruptions; we implemented this, and interruptions have dropped dramatically. We basically don’t have interruptions.”

Just as important is who gets to see the proof. “We can see that in our product, but we can also show it to our customers, so they can say: okay, wow, you did this, and we can see the impact in real time.”

Wienerschnitzel: an X-ray, store by store

One place the loop runs today is Wienerschnitzel, the iconic American hot dog brand, where Presto is deployed across many stores. Working with Bluejay, Presto has built a handful of custom evals now running in the field, telling the team what’s happening across those stores and what would make the experience even better.

In any restaurant brand there’s a spectrum: most stores do really well, and a handful are the problem children. The hard part has always been telling why.

“Without the tooling Bluejay provides and the evals we’ve created, it’s much harder for me to analyze what’s going on. Now I can truly do an X-ray of what’s going on, and use that to benefit the whole product overall. The baseline just increases.”

Store isolation is what makes the X-ray possible. “Anyone who’s ever run a restaurant understands that treating stores the same is a big problem,” Gupta said. Store A and Store B perform differently even with the same AI layered on, and at Wienerschnitzel, where many franchisees own just a couple of stores each, being able to compare one franchisee to another matters. Bluejay gives Presto a new set of metrics that see through the business layer by layer, in a way the old ones never could.

From analog to digital: intelligence for operators

The metrics don’t stop at Presto’s door. Because the drive-thru has always been analog, the restaurants running it have had thin tools for understanding it. Voice AI changes what an operator can know about their own business, and Presto is revamping the toolkit it hands to store GMs, district managers, and franchise owners, additive, in part, because of Bluejay.

“One of the biggest game changers has been the ability to understand and track QSR shoppers in a way we simply haven’t been able to before. We can see patterns, identify opportunities, and dig into what’s actually driving performance across locations with an entirely new level of visibility.”

That visibility reaches all the way to the order window. Instead of hearing only from the guests who choose to fill out a survey, Presto’s teams see what’s actually happening in the conversation itself.

“Bluejay has given us a much fuller view of what the guest experience actually looks like in the drive-thru. Their evals surface what customers are reacting to at the order level, rather than only from the guests who choose to submit feedback. They’ve brought us data we can turn into actionable insight for both our internal teams and our operators, and that goes a long way toward building trust.”

“The beauty of this AI wave is that it is allowing us to quite fundamentally digitalize businesses that previously were analog,” Gupta said. Showing restaurant brands what observability means for their own operations is part of why Bluejay meets Presto’s customers in person: once operators see the metrics, they start to see how differently their business could run.

Build vs. buy: partnering with the best in class

Observability is core to Presto’s vision, and many CEOs in Gupta’s position would have reflexively built it in-house. He weighed it, and decided against it, for two reasons.

First, focus: the metrics Presto needs demand people who think about observability all day, from how eval prompts get written to how A/B tests are structured to how results are displayed so a restaurant business can act on them. Second, velocity: the space is evolving fast, and the learnings will compound with whoever is at the frontier.

The day-to-day working relationship backs the decision up. When Presto hit a problem with product telemetry, Bluejay’s team was accessible enough to help solve it overnight, “I highly recommend their product and their team,” Abbasi said. Thompson’s team says the dashboards surface insights that would previously have taken hours of manual analysis to uncover. And Gupta’s verdict on the rollout: “It’s delightful, just like seeing a blue jay in the wild.”

What’s next: more brands, and simulations

Presto is expanding with its existing brands nationwide and adding new ones, across seafood, hot dogs, burgers, ice cream, and every geography in between. “I’m a data guy at heart,” Gupta said. “I want to deliver the same magical experience to a Wienerschnitzel franchisee selling chili dogs in California as to a burger chain in the South.” The more dispersed the customer base, the more the observability layer matters.

The next step is turning that layer into a launchpad. Today Presto runs simulations on its own; the plan is to make the observability metrics the underpinning of a simulations module, so every change gets tested against the metrics that matter before it goes live. “That should be part of our standard deployment procedure,” Gupta said. It’s the same philosophy carried one step earlier in the pipeline: proactive, not reactive, before the product ever reaches a real car.