EliseAI Raises $350 Million to Push AI Deeper Into Apartment Operations

On September 29, 2026, EliseAI announced that it has raised $350 million at a $4 billion valuation. The company’s release says the financing was led by Andreessen Horowitz and Bessemer Venture Partners, with participation from Ontario Teachers’ Pension Plan, Sapphire Ventures and Navitas Capital, and Ontario Teachers’ published the same announcement on its own site. Sameer Dholakia, a partner at Bessemer, joined EliseAI’s board as part of the financing. EliseAI says it will use the capital to automate more of its customers’ operations and to grow its engineering, deployment and sales teams across its North American offices, with San Francisco becoming a second engineering hub alongside its New York headquarters. Neither release discloses the round’s structure, how much of it is new primary capital, or any customer count.

The company describes its housing business as AI for every stage of renting a home, automating leasing, resident services, maintenance and renewals. It says operators first adopted the platform for leasing and resident communications and have since asked it to take on a broader range of processes. It restates two figures from earlier announcements: that it passed $200 million in annual recurring revenue in June 2026 after doubling revenue for five consecutive years, and that its platform powers one in six U.S. apartments, with more than 30 million Americans having interacted with it. Those are company figures from a private company, not audited results and not a PSV finding, and the release does not define what “powers” means for a unit. EliseAI also runs a separate healthcare business for specialty physician groups, which this brief does not cover.

The funding follows a product change that matters more to an operator than the valuation. On September 3, EliseAI announced Apollo, which it calls its first agentic AI teammate: a single agent that it says can perform any task in the Elise platform. The launch release gives two examples, a leasing agent asking Apollo to reschedule every tour for a colleague out sick, and a maintenance technician asking which units have recurring issues. It says Apollo has access to the full settings surface of the platform, so it can take real action rather than only answer questions, and that it inherits each user’s existing permissions exactly. The company also says Apollo has run through thousands of tests of everyday scenarios and says so when it does not know something. Those are company claims about its own product; PSV has not tested Apollo.

Put together, the two releases describe a vendor moving from a chat layer in front of the leasing office to software that changes the property’s own settings, with $350 million behind taking on more of the operation. That shifts where the risk sits for an owner. When the agent inherits a person’s permissions, the permission map becomes the control: a regional manager’s access, handed to an agent that acts on plain English requests, can reach every property that manager covers. And a vendor that says it runs one in six U.S. apartments is a concentration for the whole sector, since a model change, an outage or a pricing change lands on many owners at once.

The artifact is an agent authority register for each property: what the AI vendor is contracted to do, what it is able to do, and who approved the difference. Inputs: the vendor order form and master agreement, the platform’s role and permission export for every user at the property and above it, the vendor’s published product documentation, the property’s leasing and fair housing policies, and a sample of the agent’s conversation and action logs for the last thirty days. Output: a table of every action the agent took or can take, grouped by leasing, tours, maintenance, renewals, billing and settings, with the permission it ran under, whether a person approved it first, and whether a log exists; plus the contract terms on data export, audit log retention, model change notice and termination. Every line cites the document or export it came from.

The reviewer is the regional property manager with the asset manager, and the approval gate is simple: no new category of agent action goes live at a property until that pair signs off on the permission it will run under and on the log that will show what it did. An assistant does the volume work well: reading conversation logs at scale, flagging actions taken under broader permissions than the task needed, comparing contract language to what the platform actually allows, and drafting the register. It should not decide which permissions a person or an agent should hold, and it should not grade whether a leasing conversation met fair housing rules. PSV promises no savings or outcome from any of this.