On May 28, Anthropic's $65 billion round grabbed the headlines. But the same day saw a string of smaller funding rounds that, taken together, tell a more telling story about where venture capital is heading.
Where the money is flowing: all the grimy, heavy-lifting jobs
Here are the deals from that day:
- Fonoa – $110 million Series C for tax compliance automation across 190+ tax jurisdictions.
- Garner Health – $100 million Series E for healthcare cost navigation, promising employers 12% savings on medical spending.
- Daloopa – $47 million Series C for traceable financial data fed to AI models like ChatGPT, Claude, and Perplexity.
- Saris – $28.8 million Series A for a banking lending process agent covering consumer, mortgage, and commercial loans.
- Triomics – $22 million Series B for oncology clinical AI used by MSK, MD Anderson, and Yale.
None of these companies claim to be “disrupting” an industry. Each points to a specific, expensive, high-stakes task and says, “We’ll handle this.” Tax, lending, insurance reimbursement, investment banking modeling, oncology — all are costly, tedious, error-prone processes. That is exactly the sweet spot for AI agents: they don’t need to be creative, just reliable, auditable, and able to work 24/7.
Saris is the most typical example
The company that looks most like an “AI agent startup” that day is Saris. It raised a $28.8 million Series A led by 8VC. Its pitch is unglamorous: automate the back-office processes in banks that are stuck on legacy systems. The CEO’s value proposition is blunt: “automate up to 70% of consumer, mortgage, and commercial-lending tasks.” Saris claims its agent can take over 70% of the work in consumer, mortgage, and commercial lending, cutting costs by up to 35%. And it doesn’t build from scratch — it plugs directly into the banks’ existing systems like Fiserv, Encompass, and MeridianLink.
Bank back offices are places where compliance is tight, processes are rigid, and a single mistake can mean real money plus regulatory fines. No one dared touch them before. Now agents can — because they leave a trace at every step, making it possible to audit errors. That is harder than building “a smarter chatbot,” and far more valuable.
The real signal from that day
String these deals together and a shift emerges: money is no longer paying for what AI can do in theory; it is paying for whether AI can take over a specific, expensive decision and still be auditable.
A few details are telling:
- Fonoa also acquired PwC’s Indirect Tax Edge platform. Even the Big Four’s own tax tools are being bought up and folded into agent workflows.
- Daloopa bets on “traceability”: every number fed to the AI must be traceable back to the original financial report. Otherwise, investment bankers won’t trust AI-generated numbers for decisions.
- Garner puts its value proposition in a hard number: saving employers 12% on healthcare costs. No vision talk, just savings.
Compare that to the previous wave of general-purpose chatbots, whose valuations relied on “imagination.” This wave of vertical agents is valued on “how much money saved, how low the error rate, and whether it can be audited.” The former sells stories; the latter sells bills.
What comes next
The last AI funding cycle was about how smart the model was, how many parameters it had, and how high it scored on benchmarks.
This cycle is about something else: can you point to a specific job and say, “I’ll handle this, and I’ll take responsibility if something goes wrong”?
Companies that can say that will find it easier to raise money. Those that still talk about “general intelligence changing the world” will find it harder. That line is being drawn more clearly every day in 2026.
Sources: Venture Capital & Startup Funding Roundup, May 28, 2026 (Tech Startups); CocoLoop; Latest AI Startup Funding News and VC Investment Deals - 2026 (Crescendo AI)