Goldman Sachs Just Bet $110 Million That AI Agents Can Make Financial Decisions. They're Not Wrong.
Last Tuesday, Goldman Sachs Alternatives led a $110 million Series A into a company called Taktile. Taktile builds autonomous AI agents that make real financial decisions. Credit approvals. Underwriting calls. Compliance checks. All handled by agents, not analysts.
When Goldman puts that kind of money behind AI agents handling financial decisions, the rest of us should probably take notice.
The numbers are hard to argue with
A Cambridge Centre for Alternative Finance survey found that 21% of financial services firms have already deployed AI agents to production. These aren't pilot programs or sandbox experiments. They're running on real transactions with real money.
Another 52% are either piloting or well into development.
The total market for agentic AI in finance hit $36.61 billion in 2026, up from $30 billion a year ago. That's 22% growth in twelve months, in an industry that normally moves like it's wading through concrete.
What Taktile actually builds
Picture a loan application. A human analyst reviews the file, pulls credit scores, checks bank statements, and runs it through underwriting rules. Taktile's AI agents do all of that on their own.
These agents run the full decision workflow from start to finish. No human sits between the data and the final call.
Goldman didn't invest just because the tech works. They've seen it operating inside financial firms, making faster and more consistent calls than the analysts it replaced.
Why this matters if you don't run a bank
Finance tends to be the last industry willing to hand decisions over to software. Too much at stake. Too many regulators watching. So when one in five financial firms already trusts AI agents with production-level decisions, that tells you something about where every other industry is headed.
The "we still need a human checking every output" argument is getting harder to defend. Goldman just made it $110 million harder.
What to take from this
You probably don't handle loan applications. But you do have repetitive decision workflows sitting in your business right now. Approving quotes. Routing support tickets. Qualifying leads. Processing invoices.
Every one of those follows the same pattern Taktile automates: gather data, apply rules, decide, execute.
Three things worth thinking about:
Start with one autonomous workflow. Pick something high-volume, rule-based, and low-risk. Lead qualification works well. So does appointment scheduling or invoice matching.
Define the guardrails first. Goldman didn't write Taktile a blank check and say "go make decisions." They set the rules, the boundaries, and the escalation paths before anything went live. Do the same.
Pay attention to the cost math. The reason 21% of financial firms moved to production isn't just about accuracy. An AI agent processing applications doesn't take lunch breaks, doesn't need retraining on new compliance rules, and doesn't leave after six months. The ROI is getting too clear to sit on the sidelines.
The takeaway
Goldman Sachs doesn't write $110 million checks for science projects. The firms running autonomous AI agents in production aren't experimenting. They're building a cost advantage that compounds over time.
If you're still debating whether AI agents are ready for real work, Goldman just answered that for you.
— Mark Garza, Laimen AI
