Long AI Is Long SaaS

// Anatomy of the Bear Case

The "AI kills SaaS" thesis has become a recurring theme in technology investment circles, and it is not entirely without merit. The bear case runs as follows: AI agents, powered by increasingly capable language models, will progressively replace the workflow-automation value proposition of enterprise SaaS. Why pay $50,000 per year for a point solution that automates one workflow when an AI agent can perform the same function dynamically, without a license? The argument draws specific force from the "unbundling" dynamic: many SaaS products are, at their core, structured interfaces sitting on top of relatively simple databases. If AI can interact with those databases directly — or if general-purpose models can replicate the core functionality — the specialized SaaS layer becomes redundant. High-profile examples abound: Cursor and GitHub Copilot have compressed the value of simple coding tools; AI writing assistants have pressured content workflow SaaS. The sophisticated version of this bear case does not claim AI destroys all SaaS — it claims AI destroys the least-defensible, most commoditizable layer of the SaaS stack: the thin-wrapper, workflow-automation products with shallow data moats and high substitutability. This version of the bear case is largely correct, and we share it. But it leads to a fundamentally bullish conclusion when applied to the SaaS companies we actually want to own.

// Why AI Supercharges SaaS

The companies we want to own in the AI era share a specific profile: they are the systems of record and workflow orchestration platforms for mission-critical enterprise processes. Think vertically integrated ERP for specialized industries, compliance management platforms with deep regulatory logic, or field-service management systems that touch every hour billed by a workforce. These businesses are not at risk of AI disruption for the same reason that early SaaS companies were not at risk from the next wave of SaaS: their value is not primarily in the software interface — it is in the accumulated data, the embedded workflow logic, and the switching costs that make them genuinely difficult to replace. For these companies, AI is not a threat — it is an expansion opportunity. The proprietary data exhaust they have accumulated over years becomes training material for models that make their workflows dramatically smarter. Copilot-style features become upsell vectors that expand ARR without displacing core licenses. Agentic automation of repetitive tasks — data entry, exception handling, report generation — can be layered on top of existing platforms, turning SaaS products into SaaS + agent suites. Salesforce's Einstein, HubSpot's Breeze, and ServiceNow's Now Assist are early commercial examples of this pattern generating measurable expansion revenue. Public cloud SaaS companies with deep data integration — Veeva, Procore, Toast — have seen their AI features correlate directly with higher net revenue retention in recent earnings cycles. The data on this is now clear: AI deepens the moat of data-rich SaaS, it does not dissolve it.

// The PE Opportunity

For private equity investors, "Long AI is Long SaaS" is not merely a thesis for observing public markets — it is an active acquisition strategy. The middle market contains hundreds of vertical SaaS businesses with the profile we have described: $25–$150M in revenue, profitable, deeply embedded in mission-critical workflows, with years of proprietary data accumulation, and currently trading at multiples that have not yet been repriced to reflect their AI optionality. These businesses were valued as software companies in a flat-to-modest-growth environment. The right framework values them as AI-native platforms in waiting — businesses that can be transformed, with the right operational investment, from legacy SaaS into systems of intelligence that command premium multiples and sustain premium NRR. The transformation program is specific and repeatable: audit the data assets, secure training rights, build the model fine-tuning infrastructure, design the AI feature roadmap, and re-launch the pricing architecture to capture the AI-generated value expansion. Covalent's investment thesis is built around identifying and executing this transformation in companies that the broader market has not yet recognized as AI inflection candidates. When the market catches up — as it did with every comparable technology wave — the delta between entry multiple and exit multiple will reflect the value we have created. The window to acquire these assets at pre-AI optionality pricing is closing. The thesis is not contrarian for much longer.