Filmore announced early access to a Copilot agent for Microsoft Teams on 14 September 2026. A sales representative at an equipment dealership sends a direct message asking about a contractor, a jobsite or a market, and the agent replies in the thread. The release puts it plainly: “The agent returns relevant demand information directly in the conversation, helping the representative decide which accounts and projects deserve attention.”
As product announcements go this is a small one. There is no model to evaluate, no benchmark, no named customer. What makes it worth a paragraph in a construction technology newsroom is who is shipping it, where they are shipping it, and to whom — all three of which are unusual.
Who
Filmore is based in New Orleans and was founded by Martin Roth and Josh Nguyen. Both came out of Levelset — which they worked at rather than founded; Levelset was founded by Scott Wolfe Jr as zlien — and Filmore’s about page is specific about their roles: Roth was “Former CRO of Levelset, where he built the sales and marketing organization from early stage through acquisition by Procore for $500M,” and Nguyen “Led AI Data Products at Procore after joining through the Levelset acquisition.”
The Levelset outcome is a matter of public record rather than founder folklore. Procore’s own announcement states that it acquired Levelset “for a purchase price of approximately $500 million, consisting of approximately $425 million in cash, subject to customary adjustments for working capital, transaction expenses, cash and indebtedness, and approximately $75 million in Procore common stock.” The deal completed in November 2021.
That matters here because Levelset was, underneath the lien-waiver software it sold, a data business. It knew who was on which job, who had been paid and who had not, because the paperwork it processed said so. Filmore is the same move in a different category: assemble a proprietary picture of who owns what equipment and who is about to need more, then sell access to it.
To Whom
The customer is an equipment dealer or rental company — specifically, per the release’s own description, “the intelligence layer for equipment dealers and rental companies.”
This is a genuinely under-covered part of construction technology. Much of what this newsroom writes about is sold to the contractor: the software that plans the work, reads the drawings, or watches the site. The dealer channel sits one step to the side of that, and it is large — the machines on every job in the archive were bought or rented from someone, and that someone runs a sales organisation with quotas, territories and a CRM that, as Filmore’s own about page argues, “showed what happened last quarter, not what was about to happen next week.”
The adjacent stories this newsroom has covered approach the same machines from the contractor’s end. Gritt raised $26 million to put autonomy on equipment contractors already own or rent. Xpanner raised $18 million selling automation as a subscription rather than as iron. Filmore is pointed at the other side of the same transaction — not making the machine smarter, but making the person selling it better informed.
Where
The agent lives in Microsoft Teams, and that is the actual product decision.
Roth’s quote in the release makes the reasoning explicit: “Equipment sales reps do their best work in the field, talking to customers. They can’t sell equipment when they are digging through spreadsheets or trying to learn new software,” and “By bringing Filmore into Microsoft Teams, we can meet reps where they already work. They can ask a direct question, get useful context, and spend more of their day building relationships with customers.”
What follows is analysis. The prevailing pattern in construction AI over the last two years has been to build the agent inside the vendor’s own application — Autodesk’s assistant inside Forma, Procore’s agents inside Procore, Trunk Tools’ Cortex inside its own product. That makes sense when the vendor already owns the daily surface. It makes much less sense for a company whose users spend their day in a truck and whose employer’s daily surface belongs to Microsoft. Shipping into Teams rather than asking a dealership to adopt another login is the cheaper path to the only thing that matters for a sales-intelligence product, which is whether anyone opens it.
The risk of that choice is the mirror image: a business built on someone else’s distribution surface is a business exposed to that surface’s terms. That is a well-understood trade and Filmore is presumably making it knowingly.
What the Release Does Not Say
There is no named customer or design partner in the announcement, no usage or accuracy figure, no pricing, and no general-availability date — early access is the stated status, with the release directing interested dealers to contact the company. Filmore has not publicly announced a funding round that this newsroom could find, and the release names no investors.
The data sources are described only in the general, and the release is careful about whose data is whose: it says “Equipment dealers have access to large amounts of useful data, but much of it is scattered across public records, project databases, dealer systems, service departments, telematics feeds, and individual representatives’ notes” — that is the dealer’s problem being described, not an inventory of Filmore’s own inputs. Of Filmore itself the release says only that it “connects equipment ownership, project activity, and dealer data into one picture.” Its site advertises coverage across all states with more than twenty data layers including permits, projects and UCC filings. Those are the company’s own characterisations of its coverage, and this newsroom has not independently tested any of them.
The checkable facts are the 14 September 2026 announcement date, the early-access status, the Teams surface, the founders’ identities and their Levelset roles, and the terms of Procore’s 2021 acquisition of Levelset as stated in Procore’s own filing.
