Latch Names AI Beside a $10M-$12M Savings Target, With No Headcount

Latch reported on November 12 that its reduction in force, due to finish in Q4, should save about $10M-$12M a year, with Q3 revenue down 18.9% to $14.1M. The release gives no headcount and does not say AI caused the cut.

Latch Names AI Beside a $10M-$12M Savings Target, With No Headcount

On November 12, Latch, which trades as DOOR (LTCH), reported that the reduction in force it began in August will finish in the fourth quarter and is worth about $10M-$12M in annualized operating cost savings. The Q3 release puts revenue at $14.1M, down 18.9% from a year earlier.

It does not say how many people are leaving. It does say AI is embedded in how the company builds software, handles support and runs internal operations. That pairing, a cost cut with no headcount and an AI paragraph beside it, is the whole story, and the release never connects the two.

The quarter

Software revenue was $6.3M, up 18.1%. Hardware was $3.1M and professional services $4.7M. The three lines add to the $14.1M total, which means software is about 45% of the business and the other 55% is where the decline sits. Net loss was $20.5M and adjusted EBITDA loss was $3.8M. Cash stood at $17.6M on September 30.

Two charges sit in the quarter. A $13.6M non-cash goodwill impairment wrote off the entire remaining goodwill balance. Restructuring cost $1.2M, which is what Q3 has booked so far for a reduction that is not finished. Separately, the company completed the sale of certain DOOR Property Management assets on October 30, and the release discloses an SEC settlement with a $1.0M civil penalty under a cease-and-desist order dated October 1.

The savings math

These are our numbers, not the company’s. Q3 revenue annualizes to about $56M, so $10M-$12M of savings is roughly 18% to 21% of the current revenue run rate. The $3.8M quarterly adjusted EBITDA loss annualizes to about $15M, and the savings would cover somewhere between two-thirds and four-fifths of it.

That is a large share of the gap from one program. It also means the cut, taken alone, does not close the gap. Revenue fell 18.9% in the same quarter the savings were being set up.

What the release says about AI

Three statements, all in the company’s own words. AI remains embedded across software development, customer support and internal operations. AI-assisted development has nearly quadrupled development output. And the company says AI contributes to higher productivity and a leaner cost structure, with plans to extend that automation into property operations.

Two things are absent. The release gives no headcount for the reduction, and it never says the reduction happened because of AI. We treat this as AI named beside the cost line and not as the stated reason. That distinction matters. A reader who files this under AI replacing workers is reading more than the document supports, and a reader who files it under no AI angle is ignoring a paragraph the company chose to put there.

Where it sits in the cohort

Five9’s November 5 release added back $4.3M of reduction-in-force costs and also gave no headcount. SoundHound’s plan cut about 27% with AI unmentioned. Trade Desk booked its charge and kept AI out of the explanation.

Latch differs from all three in one way. In SoundHound and Trade Desk, AI went unmentioned in the explanation for the cut. In Five9, AI showed up as product revenue, not as a claim about cost. Latch is the one where the document that reports the cut also credits AI with higher productivity and a leaner cost structure, and it does so in a quarter when revenue fell 18.9%.

Why the missing headcount matters

A savings figure without a headcount is hard to audit. $10M-$12M a year could be a dozen senior salaries or a much larger group of junior roles, and the release gives no way to tell. It also means the AI claim cannot be tested against the cut. If the quadrupled development output is real, the sensible place to look for it is engineering headcount, and that number is not here.

The $1.2M restructuring charge is the only other handle. Against $10M-$12M of annual savings it looks small, but the reduction is not finished, and Q4 will carry the rest of the cost. The honest reading today is that the program is partly booked, fully priced on the savings side and silent on people.

What it means for careers

The release lists where AI is already used, which is software development, customer support and internal operations. It does not list where the cuts landed. Do not read the first list as the second. What it does give is a direction: the next extension is property operations, the part of the business that has just been partly sold.

For people in those three functions at small software and hardware companies, the exposure signal is the output claim. A development team that says it has nearly quadrupled output gives its board a reason to ask what the team needs to be. That is a question about future headcount, not a finding about past cuts.

The reduction is due to finish in Q4 2026. The 10-Q or the 10-K is where an employee count would normally show up, and it is the number that would turn this from a savings target into a measured workforce change. Until then it stays a target.

What to watch

Four things settle most of the open questions. First, the employee count in the next periodic filing, compared with whatever the company reported a year ago. Second, whether Q4 results show the restructuring charge rising beyond the $1.2M booked so far. Third, whether the $10M-$12M savings figure holds or gets restated once the reduction closes. Fourth, what the plan to extend automation into property operations turns out to mean in practice, given that part of that business was sold on October 30. If the company starts giving headcount alongside its AI productivity claims, this stops being a cost story with an AI paragraph and becomes a measured displacement case.

Keep reading