Open any AI venture pitch deck right now and count how many slides reference the Fortune 500. The consensus is obvious: enterprise contracts are bigger, sales cycles are predictable enough, and the buyer has a budget line for AI experimentation. Every fund and every founder is fishing in the same pond.
We think this is a mistake. Not because enterprise is wrong — it's just crowded. The real white space is the 33 million small and medium businesses in the United States that have never had software built for them, can't afford to hire engineers, and have been stuck with off-the-shelf tools that don't fit their actual workflow.
What the SMB workflow actually looks like
Consider a regional HVAC company doing $8M in revenue. They have a phone that rings, a dispatcher who schedules jobs in a notebook, three technicians, and a billing person who manually invoices customers from a spreadsheet. Every category of software ever built has tried to sell them something — Salesforce, ServiceTitan, QuickBooks. Most of it sits unused because the workflows don't match how the business actually operates.
An agentic system can. It can answer the phone, parse the request, check technician availability, schedule the job, dispatch the right person, generate the invoice, follow up on payment, and reconcile the books. End to end. With one onboarding conversation and zero ongoing software training. That's the demo that sells itself, and that's the demo nobody is building for SMBs because the conventional wisdom says SMBs don't pay.
Why the consensus is wrong about SMB monetization
The conventional wisdom is wrong. SMBs absolutely pay — they just pay differently. They pay per outcome, per call answered, per invoice collected, per hour saved. The pricing model is closer to a contractor than a SaaS vendor, and that scares away every VC trained on enterprise ARR multiples.
The SMB unit economics agentic founders should anchor on
- •$200–$2,000 per month average contract value, paid monthly without procurement.
- •Outcome-priced add-ons (per booking, per recovered invoice) that often exceed the base subscription.
- •Sub-30-day sales cycles when the demo is live and the buyer is the owner.
- •Net revenue retention above 110% when the agent measurably grows the business.
- •CAC under $1,000 via channel partners, trade shows, and warm SMB networks.
The aggregate market is enormous
Here's what's underappreciated: the aggregate market is enormous. SMBs spend over $700B annually in the US alone on labor that agentic systems could automate — front desk staff, dispatchers, schedulers, billing clerks, intake coordinators, basic legal review, bookkeeping. Even capturing 1% of that is a $7B market that almost nobody is competing for.
And distribution is dramatically easier than enterprise — you can reach an SMB owner via Facebook, a trade association, a franchise newsletter, a single warm intro from a peer, or organic search for "answering service for HVAC." Compare that to a 14-month enterprise procurement cycle with three rounds of security review and a CIO who has not actually used the product.
The verticals we're watching
- •Home services — HVAC, plumbing, electrical, roofing, landscaping. Phone-driven, dispatch-heavy, billing-broken.
- •Health & wellness — med-spas, dental, physio, mental health practices. Intake, scheduling, insurance, follow-up.
- •Professional services — accounting firms, small law practices, insurance brokerages. Document-heavy and ripe for autonomous review.
- •Food & hospitality — independent restaurants, catering, event venues. Reservations, ordering, supplier coordination.
- •Logistics & trades — freight brokers, towing, equipment rental. Quote-to-cash workflows that today live in text messages.
Why now
Three things have crossed thresholds in the last twelve months that make SMB agentic AI suddenly viable. First, voice agents are now indistinguishable from a competent human receptionist on routine calls. Second, integration with the messy SMB tooling stack — QuickBooks, Calendly, Stripe, ServiceTitan, Square — has become a one-day project instead of a one-quarter project. Third, SMB owners themselves have used ChatGPT enough that the demo no longer requires education. They get it before you finish the sentence.
We back the founders who see this. Not because the consensus is wrong about enterprise being lucrative, but because the consensus is wrong about everyone else having to compete there. The biggest agentic AI businesses of the next decade will be built on top of SMBs that the rest of the venture market wrote off as ungovernable. We disagree, and we're putting capital behind it.
