Chipotle's Automation Bet: Inside the Augmented Makeline and Autocado Roadmap

The lunch rush at the Chipotle on a corner of suburban Chicago last Thursday — a strip-mall anchor between a nail salon and a Verizon store, the kind of unit that makes up the long middle of the chain’s 3,500-plus restaurant footprint — was, to a first approximation, the same lunch rush you would have seen a year ago, or two. A line of fourteen people at 12:17 p.m. Three crew on the front line. A fourth in the back, prepping avocado, by hand, with the wide-bladed knives and the green plastic bowls that have been Chipotle’s signal of “we make it here” since the company was a single Denver storefront in 1993. The digital make-line overhead clicked through tickets. The cashier called names. The line moved.

And yet what I was looking at — what I had driven forty minutes through January slush to look at — was a restaurant that has spent the last two years quietly becoming the most automation-credible chain in American fast-casual. Not because anything in front of me had visibly changed. But because of two pieces of equipment that, in some other Chipotle in some other zip code, are already changing the shape of the work: a robot called Autocado, which prepares avocados for guacamole, and a piece of conveyor-and-vision hardware called the Augmented Makeline, which assembles digital orders — the bowls and salads that come in through the app, not the in-store line I was standing in.

I want to make a contrarian case in this column, and I want to be honest about its limits up front. The case is this: of the major U.S. restaurant chains that have made automation announcements over the last three years — and there have been many, most of them either marketing theater or pilots that quietly disappeared — Chipotle’s program is the one operators should be taking most seriously. Not because we have public proof that it works at scale. We don’t, yet. But because the structural conditions at Chipotle — the menu, the unit economics, the digital mix, the leadership transition — line up in a way that makes the bet legible. The limit on the case is that the company has not yet given the market a full accounting of what these pilots are doing to throughput, labor hours, or margin. The next chance to do that is the Q4 2024 earnings print in February, and what gets disclosed there will shape how the rest of the industry reads the playbook.

So this column is a case study in two parts. What we can say from what is already public, as of the last week of January 2025. And what we will be listening for when the print drops. Forward-context interpretation is flagged explicitly throughout — I am writing on a Tuesday in late January about a company whose next material disclosure has not happened yet, and operators reading this should treat the forward portions as anticipated, not as known.

Niccol’s legacy, decoded

Brian Niccol left Chipotle for Starbucks in August 2024. By the time he walked out the door — which I want to be careful to describe accurately, because the narrative around his departure has hardened in ways that are not entirely fair to either company — Chipotle had become the operating template that other fast-casual chains were quietly, then loudly, copying. Digital made up roughly a third of sales. The Chipotlane drive-thru-pickup format had reshaped new-unit economics. Throughput on the front line was the highest in the segment. And the company had, in late 2022 and through 2023, started making the kinds of automation announcements that, from any other chain, I would have read as headline-bait.

Two reasons I did not read them that way at Chipotle. The first is that the announcements were architecturally specific. Autocado was not “we are exploring robotics in the kitchen.” It was a piece of equipment that does one task — halve, pit, peel, and scoop avocados — chosen because that single task consumes a meaningful number of labor minutes per restaurant per day, and because the failure mode of getting it wrong is bounded. A bad avocado is a bad avocado. It does not, the way a chicken-cooking robot might, create a food-safety incident. The second reason is that the Augmented Makeline was, on closer inspection, a digital-order-specific solution. It does not replace the front line, where the customer interacts with crew. It sits in the back, where the bowls and salads ordered through the app are assembled and where, in a busy unit, half or more of the production volume now lives — invisibly to the in-store guest.

Niccol’s successor is Scott Boatwright, who was promoted from chief operating officer. The market’s reaction to the succession, in the weeks after the Niccol announcement, was a mix of share-price wobble and an expectation — reasonable, if you had been reading the company’s communications carefully — that the operating program would continue. Boatwright has been inside the throughput-and-format work since well before Niccol’s departure. He has spoken publicly, in earnings calls before this one, about the labor-deployment thinking behind the digital makeline. He is the right person, on paper, to continue the program. The question, and it is a real question, is what cadence of disclosure he and the new CFO posture will set on the automation pilots — because the prior cadence, while specific in architecture, has not been specific in operating metrics.

What Autocado actually does and what it doesn’t

I want to spend a paragraph being precise about Autocado, because the piece of equipment has been described in the trade press in ways that conflate it with much more ambitious robotics programs at other chains, and the conflation does the operator-reader a disservice.

Autocado is a single-task piece of automation. An employee loads whole avocados into a hopper. The machine sorts them, halves them, removes the pit, and scoops the flesh into a bowl. A human still mashes the flesh into guacamole, still adds the cilantro, lime, onion, and salt by hand, and still tastes and adjusts the final mix. The machine, in other words, does the prep work — the time-consuming and repetitive halve-pit-scoop part — and leaves the actual recipe to the crew. This matters for two reasons. First, the failure mode is bounded; if the machine produces a marginal scoop, the crew member who finishes the guac catches it. Second, and this is the operator point that I think is underappreciated: Chipotle goes through an enormous volume of avocados — the company has historically disclosed figures in the hundreds of millions of avocados per year across the system — and the per-restaurant labor minute reduction from doing the prep mechanically is, in principle, a directly measurable line.

What we do not know publicly, and what I am hoping the Q4 print clarifies, is the actual measured reduction in prep labor in the test units versus control units; the unit-cost of the machine and the depreciation schedule the company is using to evaluate the return; and the rate at which the equipment is being deployed system-wide, both in terms of restaurants per quarter and in terms of which restaurants are getting prioritized. The company has spoken in past calls about an intention to broaden the pilot. The mechanics of that broadening — capex, cadence, hurdle rates — are the operator detail that has not yet been disclosed.

I will go further and say something I do not see being said in the trade press, which is that Autocado is interesting less as a piece of robotics and more as a piece of organizational evidence. A chain that builds a custom robot for a single ingredient is a chain that has done the labor-time math at the SKU level. Most chains have not. Most chains are still operating on the level of “we need to take hours out of the kitchen” without a clear view of which ingredient, which task, which station to attack first. Chipotle’s choice to start with avocado was not arbitrary, and the choice itself tells you the company has a methodology. That methodology is, in many ways, more transferable to the rest of the industry than the specific machine is.

The Augmented Makeline pilot, contextualized

The Augmented Makeline is a separate program and, I would argue, the more strategically important one. Where Autocado attacks a single prep task, the Augmented Makeline attacks the digital-order assembly line — the back-of-house production flow that produces every bowl, salad, and burrito ordered through the app or through third-party delivery aggregators. It is, in concept, a conveyor with vision and dispensing. Bowls travel along the conveyor and the system places ingredients — the digital order is decoded into a sequence of dispenses — and a crew member at the end of the line finishes the assembly, adds the items the system does not yet handle well (the cheese-and-lettuce-and-salsa stage, in particular), and lids and bags the order for handoff.

To understand why this is strategically significant, you have to understand the structural problem the digital channel created at Chipotle, which is the same structural problem it has created at every digital-heavy fast-casual operator. In the pre-digital model, a Chipotle line was a single line: the customer arrived, walked the line, customized at each station, and paid. Throughput was bound by the speed of customer decisions and the cadence of crew motion. The introduction of the digital channel — initially a side benefit, then a third of sales, then in some urban units half or more — created a second line of production that the front line had to accommodate. Operators dealt with this through a second physical makeline in the back, staffed separately. That second makeline did not reduce front-of-house throughput, but it doubled the production complexity of the unit. The crew now had to manage two production flows, two ticket queues, two sets of timing constraints, against a single ingredient prep.

The Augmented Makeline is, in operating-architecture terms, the answer to that doubling. It is automation that fits the production flow that the digital channel created — which is to say, it is automation aimed at the back, not the front. The front line stays human. The customer experience does not change. The cost structure of the second production line, which had been steadily eroding the unit-level margin advantage of high digital mix, becomes addressable. This is, in my view, the underappreciated piece of Chipotle’s program. The chain is not automating to remove the crew that customers see. It is automating to eliminate the cost of the production line that customers do not see — the line that the company’s own digital strategy created.

What we do not yet have public visibility on, and again, the Q4 print is the next opportunity, is the throughput of the Augmented Makeline in the pilot units relative to the manual digital line; the labor reallocation that results — whether the crew member that the Augmented Makeline displaces is moved to the front line, used to extend hours, or removed from the labor model; and the capital cost per unit, which determines whether the rollout cadence makes sense at the corporate-cost-of-capital level or only at a more aggressive hurdle rate.

The company has previously framed Augmented Makeline expansion in terms of restaurant counts for the year ahead, but those counts — the public ones I have seen referenced — are forward targets, not realized rollouts, and I want to be careful not to cite them as accomplished facts. As our subsequent operator case study on Sweetgreen automation argues in a later piece, the gap between announced pilot expansion and actual restaurant-deployed equipment is, across the segment, much larger than the press would suggest. That is not a critique specific to Chipotle. It is a feature of how automation rollouts work in physical restaurants, where construction sequencing, equipment lead times, and crew retraining all gate the cadence below what the announced numbers might imply.

What the Q4 2024 earnings print needs to clarify

The Q4 2024 earnings report, scheduled for mid-February, is the next event that will materially change what operators can say about Chipotle’s automation program. I want to enumerate, as specifically as I can without pretending to know what the company will choose to disclose, what I will be listening for. I am writing this as a watch-list for operator-readers, not as a forecast.

The first thing to listen for is whether the company offers a system-wide count of Autocado-deployed restaurants as of year-end 2024, and a forward target for year-end 2025. The previous calls have offered directional language. A specific count, with a specific target, would shift the disclosure from program-existence to program-cadence.

The second is the same disclosure for the Augmented Makeline, with the additional question of how the company is choosing which restaurants get the equipment. Is it being deployed to high-digital-mix urban restaurants where the production-doubling problem is most acute? Is it being deployed to new builds as a standard? Is it being retrofit into existing units, and if so, on what construction schedule? Each of these choices has different operating-cost implications, and the choice the company has made is, in itself, a piece of information about how confident management is in the unit-level return.

The third is the labor-line discussion. Chipotle has historically discussed labor in terms of percent of sales and in terms of crew-hours-per-restaurant. The cleaner disclosure, which I do not necessarily expect but would welcome, would be a like-for-like comparison of crew-hours in automation-enabled units versus control units, normalized for sales mix. Anything in this direction would let analysts and operators model the program. Without it, the rollout is opaque at the financial level even if the architecture is clear.

The fourth is the capex disclosure. Chipotle’s annual capex is typically discussed in terms of new units, remodels, and digital infrastructure. A separate line for automation capex — or at least a sized callout within the digital infrastructure line — would let the market see the corporate-level commitment in numbers. The absence of such a callout, conversely, will be readable.

The fifth, and this is the one I am genuinely uncertain about, is how Scott Boatwright will frame the automation program in his first full-year-earnings-cycle as CEO. Niccol had a public-facing persona around the program. Boatwright is, by reputation, the operator. The framing he chooses — whether the program is presented as a strategic capability or as a tactical labor-management lever — will signal how the new leadership wants the market to value the work. I will be watching the framing as carefully as I watch the numbers.

I want to flag explicitly: everything in this section is forward. I am describing what I will be listening for, not what has been said. Operator-readers should not infer that any of the disclosures I am listing for are actually expected. They are the disclosures that, in my view, the program needs.

What we should not say yet

This is the section where I do the work I most want the trade press to do more of, which is the work of marking the boundary between what is knowable and what is not. From the Tuesday-late-January-2025 vantage of this column, the following are not yet established facts, regardless of how confidently you may have seen them stated elsewhere:

We do not yet know what Chipotle’s same-store sales did in Q4 2024. We will know in February. Anything written in the present tense about Q4 same-store sales is forward.

We do not yet know what the comparable-traffic trend has been in the first weeks of Q1 2025. The next disclosure of comparable traffic will be in the Q1 earnings cycle, in late April or early May. I have seen later-dated trade press make claims about Q1 2025 traffic dynamics; from my Tuesday-late-January vantage, that reporting is forward and should be treated as such.

We do not yet have publicly disclosed figures for the size of the Augmented Makeline deployed footprint at year-end 2025. The figures that have circulated in trade press conversations are forward targets from previously-disclosed Chipotle commentary, not accomplished rollouts. The distinction matters because operators reading those numbers as completed milestones will draw inferences about deployment velocity that the underlying disclosure does not support.

We do not yet have CEO-level commentary from Scott Boatwright across a full Q4 cycle. His framing of the automation program, the leadership signal he wants to send, and the operating priorities he chooses to emphasize are all forward.

I am stating this list bluntly because the structural problem in trade-press automation coverage is the collapse of forward, current, and retrospective into a single present-tense narrative. The story I have been telling in this column has been deliberately staged to keep those tenses separate. The Q3 2024 disclosures are retrospective. The Autocado and Augmented Makeline architectures are current. The Q4 print, the Boatwright framing, and the year-end-2025 rollout figures are forward. Operator-readers who want to model this program for their own purposes — and I get the emails, you are out there — need to keep the tenses separate too.

For the company’s own current disclosures, the canonical reference is the investor-relations page at ir.chipotle.com, which is where I would direct operator-readers who want to verify the pre-Tuesday-January-28-2025 record for themselves.

Operator takeaways for fast-casual CFOs

I want to close with the part of this column that the operator-readers actually came for, which is what to do with the case study if you are running a fast-casual chain or a multi-unit operator and trying to think about your own automation roadmap. I am going to be opinionated. I am going to be wrong in places. Push back; that is what tips@tabletransfers.com is for.

Take the methodology, not the equipment. Autocado is not a thing you buy. It is custom-built for Chipotle’s avocado volume. The transferable lesson is the methodology: pick a single ingredient, measure the labor minutes it consumes per restaurant per day, model the failure modes, and start there. Most chains do not have an ingredient-level labor-minute model. Building one is, in my view, the precondition for any serious automation program. Chipotle’s program is impressive less because of what the robots do and more because of the apparent rigor of the analytical work that selected the robots.

Automate the back, not the front. The Augmented Makeline story is, at its core, a thesis about where automation pays off in a customer-facing restaurant. It pays off in the production flow that the customer does not see, which in digital-heavy chains is now half or more of the volume. The instinct to put the robot at the counter, where the press release writes itself, is the wrong instinct. The robot belongs in the back, on the digital line, where the customer experience does not have to absorb its failure modes and where the cost being addressed is the cost the digital channel created.

Match the rollout cadence to the actual constraint. Across the segment, announced automation rollouts move at roughly twice the speed of actually-deployed automation rollouts. The constraints are construction sequencing, equipment lead times, crew retraining, and — the one most operators do not flag — back-of-house power and refrigeration upgrades that the new equipment often requires. A CFO sizing a multi-year automation program should size the construction calendar before sizing the equipment order.

Watch the leadership-framing signal. When a new CEO inherits an automation program, the framing they choose in their first full-cycle earnings discussion is informative. Strategic-capability framing implies a long-cycle commitment to the program as a competitive advantage. Tactical-labor-lever framing implies the program is being valued at its near-term unit-level return. Both can be the right framing depending on the chain, but they imply different capex cadences and different organizational investments. For Chipotle specifically, the Boatwright framing in the upcoming print is the signal to listen for. In a later piece we publish on Chipotle’s evolving roadmap, I will return to whatever the framing turns out to be and to what it means for the rest of the segment.

Do not overweight the avocado. The press around Chipotle automation has been disproportionately about Autocado, because it is photogenic and because guacamole is a brand-defining ingredient. The Augmented Makeline is the operationally more important program, by a wide margin. If you are reading the chain’s strategy through the prep-room robot rather than through the digital production line, you are reading the wrong story.

Hold the case study under low confidence until the Q4 print. I want to be explicit, as I close, that the case I have been making in this column is structurally credible but is operating under low-confidence conditions on the specifics. The architecture is public. The cadence is not. The rollout numbers I have seen in trade-press circulation are forward targets, not accomplished milestones. The Boatwright framing has not yet been delivered across a full earnings cycle. None of this invalidates the case. It does mean that the case is not yet a verdict. The Q4 print, and the Q1 print after it, will move the case toward a verdict in one direction or another, and I will be writing through both prints in this column.

The lunch rush I was watching last Thursday ended at about 1:05 p.m. The line shortened. The crew turned to restocking. The digital tickets — the bowls and salads going out the back, into the hands of drivers I could see through the glass — kept coming, at a cadence that I want to point out is invisible from the front of house. That is the production line Chipotle is automating. That is the line the operator case study is really about. The avocado robot is the photograph. The makeline behind the wall is the program.

— Priya files The Operator. Tips: tips@tabletransfers.com.