Kobalt Labs Co-Founders Ashi Agrawal and Kalyani Ramadurgam
Kobalt Labs
Kobalt Labs has raised a total funding of $12.7 million to automate one of the fintech sector’s most stubborn bottlenecks: compliance work that still runs on spreadsheets, screenshots, and humans combing through hundred-page documents.
The funding is another signal of where fintech is heading as AI agents move from novelty to infrastructure, and as regulators intensify scrutiny around third-party risk.
In recent weeks, that “agentic” shift has surfaced across the money lifecycle. FIS launched agentic commerce tools for banks. Alinea Invest introduced AI-native wealth guidance. Shopping startups like Phia are building agents that influence transactions before checkout. Anthropic is pushing deeper into financial services with enterprise AI tools aimed at wealth and investment workflows.
Kobalt’s $11 million in Series A, announced in December, led by First Harmonic with participation from Alloy Labs and Y Combinator, turns the spotlight to a less glamorous but critical layer of that transformation: risk and compliance.
Compliance Is The Next AI Fintech Battleground
Kobalt is building AI agents that automate deeply manual compliance workflows for financial institutions, starting with third-party risk management (TPRM), internal audit and marketing compliance, according to the founders. Its customers include fintech and banking names such as Chime, Bilt, Celtic Bank, Emprise Bank, American National Bank and Meriwest Credit Union.
The timing matters. AI is now showing up in the exact places financial institutions can’t afford to be sloppy. Anthropic, for example, recently rolled out new enterprise plug-ins for Claude targeting finance functions including investment banking and wealth management, underscoring how quickly agentic software is landing inside the financial stack.
If the agentic future is real — AI systems that can shop, negotiate, transact, invest and service customers — financial institutions face a hard constraint: you can’t scale agentic systems with compliance processes that still depend on humans reading 200-page PDFs.
“If you want to grow embedded finance safely, you can’t just hire your way through compliance,” CEO and cofounder Kalyani Ramadurgam told me. “You have to scale with technology.”
What Kobalt Is Building for Fintech
Kobalt’s wedge is TPRM — the due diligence banks conduct before onboarding a new vendor, platform or fintech partner. It’s the work behind the work: reviewing contracts, information security documents, policies, disclosures and ongoing monitoring requirements.
Banks trying to grow embedded finance programs and onboard multiple fintech partners in a year are hitting a wall, Ramadurgam said. “The only way they can do that…is if we scale with technology.” Without it, she added, institutions may need years and dozens of hires to keep up.
Kobalt’s pitch is that its AI performs the “first pass” across large volumes of documentation, consistently and quickly, surfacing issues for human review — so compliance teams spend their time on judgment, not scanning.
Co-founder and Chief Technology Officer Ashi Agrawal said the platform has produced major speed improvements in practice: marketing reviews taking roughly 75% less time and vendor reviews shrinking from hours to minutes in some client workflows, based on internal metrics shared by the company.
Agrawal also pointed to a structural problem: standardization. In TPRM, “the way someone learns…is from their boss,” she said. Kobalt’s value isn’t only automation — it’s consistency across processes that often vary by institution and by reviewer.
Why This Is A Fintech Story, Not Just RegTech
Kobalt isn’t the first company to tackle vendor risk or regulatory change management. Banks already buy vendor risk and compliance software — and many teams still default to Excel plus email threads when systems fail.
On the traditional side, vendor risk management and TPRM platforms from providers like ProcessUnity and Venminder have long supported third-party workflows. Separately, “regulatory intelligence” companies such as Compliance.ai and AscentAI focus on monitoring regulatory changes and mapping rules to policies and controls.
Kobalt’s claim is that those tools often organize the process but don’t eliminate the work — especially the part that breaks teams: reading, comparing, and interpreting mountains of documents at speed.
“That’s why most of them are not much better than a spreadsheet,” Ramadurgam said.
Whether that’s true across every use case will depend on how well Kobalt performs in messy, real-world conditions. But the market pull is obvious: compliance teams are overloaded, regulators are more attentive to third-party risk, and every bank is being asked to “move faster” while also proving it didn’t cut corners.
Why This Matters For Fintech’s Next Chapter
Zoom out and Kobalt’s raise looks less like a niche regtech story and more like a marker of fintech’s maturity.
The past decade rewarded fintech for building sleek front ends. The next decade may reward fintech for building trust infrastructure — the systems that let institutions scale without tripping regulatory wires.
That becomes more urgent as banks and fintechs add vendors, AI tools, embedded partnerships and cross-border exposure. Third-party risk multiplies with each integration. Every new relationship expands the surface area for failure.
And failures aren’t contained. A marketing disclosure error, a missing contractual clause or a weak vendor control can ripple into consumer harm — and into enforcement actions that slow growth for years.
Kobalt’s product choices reflect that reality. Beyond TPRM, it screens marketing materials against regulatory expectations, flagging missing disclosures or misleading “fee-free” claims, so teams can ship faster without bypassing consumer protections.
The Founders: Technical Depth Meets Lived Experience
Ramadurgam’s background aligns tightly with the problem. When she worked for Apple, her job was to block anyone on terrorism watch lists from using Apple Pay.
In a separate Forbes profile tied to her Forbes Under 30 recognition, she described how even at Apple the compliance process remained analog and document-heavy.
“Organizations were just throwing bodies at the problem,” she said.
Agrawal worked at Affirm, and the two met at Stanford before joining forces and going through Y Combinator.
In our interview, their fundraising advice was almost aggressively un-sexy: don’t game the process, build something customers love, and let references do the work.
“The best way to be in a good position for fundraising is to actually just build a good company,” Ramadurgam said — echoing the kind of “fundamentals-first” playbook more founders are returning to in a tighter market.
It’s tempting to talk about AI in finance as if the future is purely consumer-facing: agents shopping for you, agents investing for you, agents handling your financial life end to end.
But the reality is that AI agents will only scale in finance if the back office can keep up — if compliance, audit, and risk can move at the same speed as product.
That’s the shift Kobalt is signaling.
In an era where even major AI labs are building finance-facing plug-ins, the bottleneck is no longer imagination. It’s governance.
Anthropic’s push into finance workflows is a signal that platform providers are building the plumbing for agentic work at scale. Phia’s raise was a signal that “agentic commerce” will reshape the path to checkout. Kobalt’s raise is the signal that, as these systems proliferate, compliance becomes the gating function.
Compliance isn’t sexy — until it’s the reason your fintech survives.
And in a world where AI agents are poised to touch more regulated decisions, the next wave of fintech advantage may not come from who builds the flashiest agent. It may come from who builds the infrastructure that makes agents safe enough to deploy.