Claybird Sends AI Agents to Cold-Call Janitors and HVAC Techs So You Don't Have To

The Macro: Commercial Service Sales Is Still Painfully Manual

There is a massive gap between how B2B SaaS companies sell and how commercial service contractors sell. A SaaS company has Salesforce, Outreach, Apollo, ZoomInfo, and a dozen other tools in the stack. A commercial janitorial company has a guy named Dave who drives around business parks leaving flyers on doorsteps.

I am not exaggerating much. The commercial services industry is enormous. Janitorial services alone is a $90 billion market in the US. Add HVAC, pest control, landscaping, pressure washing, and the other trades, and you are talking about hundreds of billions in revenue generated by companies that mostly find new customers through word of mouth, door-to-door visits, and cold calling from a spreadsheet.

The outbound sales tools that exist were built for tech companies selling to other tech companies. Apollo and ZoomInfo have great data on software engineers and VPs of Marketing. They have terrible data on facility managers at strip malls and property management companies that oversee 200-unit apartment complexes. The ICP is different, the buying process is different, and the data sources are different.

Clay (the company, not Claybird) has done well building a flexible data enrichment and outbound platform, but it requires significant setup and expertise. Instantly and Smartlead handle email sending at scale but do not help you find targets. None of these tools understand the specific signals that matter for commercial service sales: new construction permits, business license filings, property management changes, or facility expansion announcements.

The Micro: Permit Data Meets AI Cold Outreach

Claybird was founded by Saad Jamal and Abdullah Nauman. Saad previously led ML infrastructure at Tesla Autopilot and trained video models with Ashish Vaswani (yes, the “Attention Is All You Need” Vaswani). Abdullah built ML systems at Google for Search Ads. They came through Y Combinator’s Fall 2025 batch with a four-person team in San Francisco.

The product has three components they call Pinpoint, Trigger, and Convert. Pinpoint scans service areas and identifies target businesses using permits, directories, and buying signals specific to commercial services. Trigger deploys AI agents that send personalized emails, make calls, and leave voicemails to decision-makers. Convert pre-qualifies leads and delivers them to a dashboard ready for quoting.

The numbers they are sharing publicly are aggressive: 4x more qualified commercial opportunities, 80+ hours of monthly prospecting eliminated, and $2M in inbound pipeline generated for clients. They serve janitorial, pest control, HVAC, pressure washing, landscaping, window cleaning, hood cleaning, and grease trap cleaning companies.

What makes this interesting is the data layer. Most outbound tools start with a generic business database. Claybird starts with permit filings and industry-specific directories. A new construction permit is one of the strongest buying signals in commercial services because someone will need to hire a cleaning crew, an HVAC maintenance company, and a pest control service before that building opens. If Claybird can identify those signals faster than competitors, the outreach arrives at exactly the right moment.

The Tesla and Google pedigree is unusual for a company selling to janitorial companies, but it actually makes sense. The problem is fundamentally about processing large amounts of unstructured data (permits, directories, business filings) and generating personalized outreach at scale. That is ML infrastructure work.

The Verdict

I like this one. The commercial services market is underserved by sales technology, the buying signals are distinct from what existing tools capture, and the founding team has serious technical depth. The question is not whether the problem is real. It is whether Claybird can build reliable data pipelines across dozens of municipalities that each publish permits in different formats.

Permit data is notoriously messy. Every county has its own filing system, its own data format, and its own timeline for making records available. Building a comprehensive, real-time view of commercial construction activity across the US is a genuine technical challenge. If Claybird solves it, they have a defensible moat. If they cannot scale it beyond a few metro areas, the product stays niche.

In thirty days, I want to see how many metro areas they cover and whether the permit data is genuinely real-time or weeks behind. In sixty days, the metric that matters is close rate. Are the AI-qualified leads actually converting to signed contracts? Commercial service sales have long cycles and relationship dynamics that AI outreach might struggle with. A facility manager who has used the same cleaning company for ten years is not going to switch because of a well-written cold email. In ninety days, I want to see retention. If contractors are staying on the platform and expanding their territories, the data layer is working. If churn is high, the leads are not qualified enough.

The commercial services industry is ripe for this kind of tool. I just hope Claybird is building the boring data infrastructure alongside the flashy AI agents, because the data is what will actually make this work.