A New Era for SMB Tech
Selling software in many SMB verticals used to be untenable. Low ACVs, high CAC. But AI tools can break through the obstacles that SaaS could not, ushering in a new wave of SMB tech.
As I’ve discussed previously, new business applications in the U.S. have exploded since the pandemic. They ran at roughly 470,000 per month through 2025 and into early 2026. That’s a pace of about 5.6 million applications a year – well above the pre-pandemic baseline of 3.5 million. The cumulative installed base of SMBs now sits at 36.2 million, accounting for 46% of private sector employment and roughly 44% of GDP.
That installed base has historically been the wrong end of the venture market to build for. SMBs were too fragmented to reach efficiently, too price-sensitive to monetize, and too small to support the seat-based ACVs that made enterprise SaaS work. AI changes each of those constraints. The structural reasons that made SMBs hard to build for are unraveling.
The Market Shift
SMB software TAM globally sits between $300B and $430B depending on how you size it, growing at a 7-8% CAGR – faster than horizontal enterprise software.
But software TAM is a data point more suited to a prior generation of tools. Total U.S. labor spend is roughly $10.5 trillion, against software spend of around $313 billion – a 33:1 ratio. SMBs feel that ratio more acutely than anyone, because they can’t absorb software costs the way an enterprise can, and finding workers is hard enough as it is. There are 8.5 million open jobs in the U.S. against 6.5 million unemployed workers, and 41% of SMBs report trouble filling vacancies. AI tools that sell labor get to price against the $10.5T pool, not the $313B one.
SMB has also remained the last truly fragmented buyer base in B2B. Enterprise has been consolidated, sliced, and resold repeatedly. Lab companies clamor for enterprise contracts because they’re massive – enterprise GTM is the most effective way to scale into a trillion-dollar valuation. Meanwhile, SMB stayed fragmented because the customer acquisition cost (CAC) math didn’t work – too expensive to acquire one at a time, too small to monetize once acquired. AI can lower CAC through programmatic distribution and lowering the cost-to-serve through agentic onboarding and support. But more importantly, AI is the market expander that raises contract values and makes LTV/CAC ratios more tenable for a wider swathe of SMB verticals.
And the playbook is established. Toast, Procore, and ServiceTitan (to name just a few) each built billion-dollar-plus outcomes by going narrow on one SMB vertical. Now, in the age of AI, there are more verticals that can support these types of exits.
AI’s True Impact
Let’s take a look at the qualitative dynamics that change the SMB math:
Pricing moves from seats to outcomes: Most U.S. small businesses have zero employees. Among those with employees, the median is still under ten. Per-seat pricing caps ACV at a few hundred dollars a year in this segment, which is why traditional SaaS economics were tricky. AI changes the pricing surface. For example, a five-person practice that would never pay $200/month per seat will pay $3,000/month for a system that replaces a part-time admin. Selling work, not software, fixes the ACV problem – which fixes the LTV/CAC problem that has historically made SMB un-venture-backable.
The race to build new systems of record is still early: I wrote in AI and the Next Generation of Systems of Record about why legacy SORs are vulnerable: AI-native systems accumulate data through workflow execution rather than data entry. In enterprise, that means displacing entrenched Salesforce and Epic deployments, which is a more difficult fight. In SMB, the incumbent is usually a spreadsheet, a notepad, or nothing at all. Switching costs from “nothing” are zero.
Integration processes are a lot smoother: SMBs run on a chaotic mix of tools, many of them lacking APIs. Historically, this killed any platform play: you couldn’t get clean data in or push automation out. Agentic browser automation, which I covered previously, removes that constraint. Agents navigate web interfaces with reasoning and flexibility, plugging into the SMB stack as it exists rather than requiring vendors to rebuild it.
The Math
One objection to building tools for SMBs is the churn dynamics. On the surface, churn numbers look grim: SMB-focused SaaS churns at 3-7% monthly, compared to 1-2% monthly for enterprise. Annualized, that’s 30%+ logo churn for a typical SMB SaaS company.
But that’s the average of a bimodal distribution. Horizontal point solutions sold to SMBs churn brutally. Embedded vertical platforms, less so. Best-in-class vertical SaaS players hit 96% gross retention because their platforms become the operational backbone of the business they serve. You can’t churn off the thing that runs your business.
Now we can see the full picture. In this new era, ACVs will be higher. Customer lifetimes will be longer. Loaded customer acquisition costs will be lower. The LTV/CAC dynamics are much more favorable. Using the previous examples, at $200/month per seat with 7% monthly churn, the math is hopeless. At $3,000/month for an agentic workflow with 96% gross retention, the same business supports venture-scale potential.
The Path Towards Defensibility
Structurally, SMBs share characteristics that are attractive for early-stage builders and investors. They’re resource constrained, which means in most cases they don’t have the engineering bandwidth internally to apply horizontal models to their specific vertical and workflows in a way that makes economic sense.
And counter-intuitively, the niche components of specific SMB verticals and workflows are features, not bugs, of what makes building for these SMBs venture-backable. The model providers are trying to grow into and exceed trillion-dollar valuations; building tools for billion-dollar markets doesn’t move the needle for them.
Eventually, these service providers will layer on additional defensibility in ways that are akin to the vertical SaaS giants who have come before them: financial infrastructure, a compounding data advantage, strategic partnerships, and the social proof of having delivered before for a specific group of customers.
The Takeaway
The niche verticals that were once written off as too small are now attractive places to build. Domain depth – and all that comes with it – is the catalyst towards defensibility.
On the other hand, the SaaS-pocalypse has negatively impacted every horizontal SaaS company, precisely because the model companies are coming after horizontal applications. The same way that Salesforce became a giant, but there was still room for ServiceTitan – in the AI era, there’s room for Anthropic and vertical AI platforms in SMB. And the pie is an order of magnitude larger – or more.
For the businesses themselves – the 36 million small companies that bolster our economy – capital is about to flow towards service providers that enable their growth.
New business applications are at all-time highs, and they’re not going anywhere. This piece was about why building tools for SMBs is a good idea today, but the case will only grow stronger in the coming years. AI will reduce headcount bloat at the enterprise level, and we’ll see more solopreneurs and small business owners as the workforce evolves.
This is a market I’ll be watching closely.