Meridian Raises $17 Million To Remake The Agentic Spreadsheet - Beritaja

The conflict to tame spreadsheets pinch AI isn’t complete yet. A caller institution called Meridian has emerged from stealth pinch a much broad IDE-based attack to agentic financial modeling — and plentifulness of backing to build it. On Wednesday, the institution announced $17 cardinal successful seed backing astatine a $100 cardinal post-money valuation.

“Our extremity is to make financial modeling and spreadsheets measurement much predictable and auditable,” CEO and co-founder John Ling told TechCrunch. “How could you return a process that traditionally mightiness person taken cool hours and condense it down into for illustration 10 minutes?”

The information was led by Andressen Horowitz and the General Partnership, pinch information from QED Investors, FPV Ventures and Litquidity Ventures. The institution says it is presently moving pinch teams astatine Decagon and OffDeal, and signed $5 cardinal of contracts successful December alone.

Excel agents person been a celebrated target for AI startups, owed successful portion for the precocious costs of human-led financial analysis. But wherever erstwhile Excel agents like Shortcut AI built agents into Excel, Meridian operates arsenic a stand-alone workspace, much akin to Cursor. This allows the app to run for illustration an IDE, integrating information sources and different extracurricular references that mightiness different create friction.

Based successful New York, the Meridian squad includes some alumni of AI firms for illustration Scale AI and Anthropic arsenic good arsenic financial veterans from firms for illustration Goldman Sachs.

As Ling describes it, Meridian’s biggest situation is the strict requirements of financial clients, which often conflict pinch the non-deterministic quality of AI models.

“if you spell to 10 different package engineers astatine Google, and you want to adhd immoderate caller characteristic into an app, you’ll about apt get like, 10 wholly different implementations. And that’s wholly fine,” Ling says. “But if you spell to 10 banking analysts astatine Goldman Sachs and you inquire for 10 valuation models for a company, you would about apt get 10 almost identical workbooks.”

As a result, the Meridian squad has done important activity to make their outputs much auditiable and deterministic, while maintaining the elasticity of LLM-based tools. The consequence is simply a substance of agentic AI and much accepted tooling, minimizing the hallucinations that slow down galore endeavor deployments.

“Our extremity is to really region the uncertainty furniture correct from the LLM process,” Li says. “You cognize precisely really the logic flows, and each of these assumptions aliases immoderate that spell into the model, you could spot precisely wherever they’re coming from.”