Meridian AI raises $17M for agentic spreadsheets

Meridian AI raises $17M for agentic spreadsheets

Meridian, a New York-based startup, has come out of stealth with $17 million in seed funding to build an “agentic spreadsheet” experience for financial modeling inside a standalone, IDE-like workspace. The company says the round values it at $100 million post-money and is focused on making spreadsheet-driven modeling faster, more predictable, and fully auditable for finance teams.

The funding round and what Meridian is building

Meridian announced a $17 million seed round at a $100 million post-money valuation, positioning the raise as fuel for rebuilding financial modeling workflows around AI agents rather than bolt-on spreadsheet plugins. The round was led by Andreessen Horowitz and The General Partnership, with participation from QED Investors, FPV Ventures, and Liquidity Ventures. Meridian is based in New York, and its leadership includes CEO and co-founder John Ling, alongside CTO George Fang and COO Zach Kirshner.

At a high level, Meridian is targeting one of the most entrenched parts of modern business: complex spreadsheet modeling that underpins investment decisions, budgeting, forecasting, valuation, and scenario planning. The product direction is not simply “add AI into Excel,” but to create a separate workspace that behaves more like a development environment, where models can be assembled with clearer logic, integrated data sources, and traceable assumptions. This matters because enterprise finance teams don’t just want speed—they need reliability, repeatability, and clear audit trails that stand up to internal review and regulatory scrutiny.

Why “agentic spreadsheets” are becoming the next battleground

Spreadsheets are still the default interface for financial work, even when the underlying business is powered by modern data stacks and cloud systems. That creates an obvious opening for AI: if models, tables, and scenarios live in spreadsheets, then AI agents that can research inputs, update assumptions, draft formulas, and reconcile inconsistencies could save hours of repetitive work. But the same characteristics that make finance a valuable use case—high stakes, compliance expectations, and the cost of errors—also make it a hard environment for probabilistic systems.

Meridian’s leadership frames the core issue as a mismatch between how AI systems behave and what finance teams expect. In many software tasks, different practitioners can arrive at multiple “good enough” implementations, but in institutional finance, analysts are often expected to produce near-identical models when using the same inputs and assumptions. If AI tools generate different structures or reasoning paths each time, teams lose confidence quickly—especially when decision-makers need to defend the model under review.

This is why Meridian is leaning into “determinism” and “auditability” as product pillars, rather than focusing only on flashy generation. The company’s thesis is that agents must be constrained by structured tooling, visible logic flows, and verifiable inputs, so users can trace where numbers came from and why the model behaved a certain way. In practice, that means building systems that keep the benefits of LLM-driven automation while reducing hallucinations and making model logic inspectable.

Meridian’s IDE-style approach: why it’s different from Excel add-ons

A number of AI startups have tried to tackle finance workflows by embedding agents directly inside Excel, largely because Excel is ubiquitous and the switching cost is real. Meridian’s bet is the opposite: operate as a standalone workspace, closer in feel to an IDE, so the environment can handle broader integrations and reduce friction created by trying to squeeze agent workflows into a classic spreadsheet UI. In the reporting around Meridian’s launch, the product is compared conceptually to developer tools like Cursor—an analogy that’s meant to signal a structured, multi-context workspace rather than a single grid with an assistant panel.

From a workflow standpoint, the IDE approach can help in three ways that matter to finance teams. First, it can pull in diverse datasets and external references more cleanly, which is critical when models depend on market data, internal systems, and third-party research. Second, it can enforce structure around assumptions and model logic, so reviews are faster and less dependent on tribal knowledge. Third, it can make provenance visible—helping analysts show what changed, when it changed, and what evidence drove the change.

Meridian is also positioning its team background as a credibility signal for this blend of AI automation and finance-grade rigor. The company is described as combining alumni from AI-native firms such as Scale AI and Anthropic with finance veterans from Goldman Sachs. That mix is relevant because the product has to satisfy both “AI systems thinking” (agent workflows, orchestration, model behavior) and “finance controls thinking” (repeatability, reviewability, compliance readiness).

The traction claims, while early, are also notable in the context of enterprise adoption cycles. Meridian says it is working with teams at Decagon and OffDeal and that it signed $5 million in contracts in December alone. If those numbers hold up over time, they suggest buyers are not only experimenting but paying for AI-native modeling infrastructure—an important signal in a market where many AI pilots stall at proof-of-concept.

What this means for finance, compliance, and enterprise AI

The biggest takeaway from Meridian’s positioning is that “AI in spreadsheets” is no longer just about making formula writing easier—it is turning into a broader redesign of how models are authored, reviewed, and governed. In finance, the real cost is often not building a first model, but validating it, updating it when assumptions change, and defending it in front of stakeholders. Meridian’s stated goal is to compress hours of modeling work into minutes while keeping the process transparent and reducing doubt about how the model was produced.

This emphasis on predictability is a direct response to the reality that regulated industries cannot treat AI output as a black box. If an AI agent updates a valuation model, teams need to understand exactly what inputs were used, how logic flowed through the model, and where each assumption originated. Meridian claims its approach is designed to make those assumptions and logic flows visible, which aligns with what compliance-oriented finance teams demand before adopting automation at scale.

There is also a broader market implication: enterprise AI is moving from “chat interfaces layered on top of work” to “systems that produce accountable artifacts.” In this framing, the spreadsheet is not just a grid—it is a contractual object inside organizations, where the model itself becomes evidence for decisions and an anchor for accountability. Agentic systems that can generate, revise, and reconcile models will only succeed if they preserve that accountability, and Meridian is explicitly building toward that constraint rather than avoiding it.

Still, the path is not simple. Even if a platform reduces hallucinations through structured tooling, enterprises will test it under edge cases: inconsistent inputs, missing data, ambiguous assumptions, and changes in how stakeholders define “correct.” Adoption may also vary by function—private equity and investment banking may value repeatability and auditability most, while FP&A may prioritize speed, collaboration, and integration with internal BI stacks. The winners in this category will likely be the vendors who can satisfy both worlds without forcing teams to abandon the spreadsheet mental model that still runs finance operations.

How AI World events can spotlight this shift

For the ai world organisation, stories like Meridian’s rise are useful because they show where “agentic AI” is moving beyond demos into enterprise-grade workflows that must withstand audit, compliance, and board-level scrutiny. Meridian’s focus on deterministic, reviewable outputs is exactly the kind of product principle that enterprise buyers ask about in real-world AI adoption, and it’s the kind of discussion that fits naturally into the ai world summit agenda tracks on enterprise AI, governance, and measurable ROI.

If you’re planning content around ai world summit 2025 / 2026, Meridian is a timely example to frame a larger theme: the next generation of productivity platforms won’t simply “assist” inside existing tools; they’ll redesign the workspace so agents can operate with guardrails, transparency, and traceability. That theme can also be tied to ai world organisation events programming focused on responsible AI, AI for business transformation, and AI’s operational impact across finance, HR, sales, and operations. And for SEO and community building, it provides a credible narrative bridge between daily AI news and the long-horizon value of attending ai conferences by ai world where leaders can compare approaches, validate vendors, and learn deployment patterns from peers.