Spreadsheets quietly run the world’s money, but almost nobody fully trusts them. A single hidden formula can move millions. That mistrust is exactly what Meridian is selling against: not smarter spreadsheets, but spreadsheets that can finally explain themselves.
In this piece, we’ll look at what Meridian is actually building, why investors are willing to value a seed‑stage “agentic spreadsheet” startup at $100 million, and what this says about the next decade of AI in finance and operations. We’ll also unpack what this means for European firms operating under some of the world’s strictest AI and data rules.
The news in brief
According to TechCrunch, New York–based startup Meridian has come out of stealth with a $17 million seed round at a $100 million post‑money valuation. The funding was led by Andreessen Horowitz and The General Partnership, with participation from QED Investors, FPV Ventures and Litquidity Ventures.
Meridian is not another Excel plug‑in. It’s a standalone, IDE‑like workspace for “agentic” financial modeling. Instead of asking an LLM to tweak a single sheet, users work in a dedicated environment where AI agents can build, refactor and maintain financial models while integrating external data sources.
The company is already working with teams at Decagon and OffDeal and claims to have signed $5 million worth of contracts in December alone. The team combines alumni from AI players like Scale AI and Anthropic with people who have done time inside big finance, including Goldman Sachs. Their selling point: outputs that are more predictable, auditable and deterministic than typical LLM tools.
Why this matters
If you strip away the hype, Meridian is going after a very specific pain point: high‑stakes decisions are still made in tools that behave like the Wild West.
Financial models inside banks, PE funds, SaaS finance teams and even scale‑ups are:
- built in Excel or Google Sheets
- maintained by overworked analysts
- passed around by email or SharePoint
- almost impossible to fully audit years later
AI has already arrived in this world — from Excel Copilot to a flood of “AI for sheets” plugins — but most of those tools optimise for convenience, not control. They help you write a formula faster; they don’t help a CFO sleep better before an audit.
Meridian’s bet is different: that the real opportunity is governed automation. In other words, AI agents that:
- can do the grunt work of structuring, cleaning and modeling data
- but leave behind a transparent, deterministic trail of logic
- and can be inspected, versioned and audited like code
The winners, if Meridian (or someone like it) succeeds:
- Finance leaders who get faster scenario analysis without sacrificing sign‑off discipline.
- AI‑native FP&A teams that can run more frequent, more complex models without hiring armies of analysts.
- Audit, risk and compliance groups that finally get traceable models instead of arcane spreadsheets.
The likely losers:
- Junior analysts whose value is mostly manual model wrangling.
- Legacy Excel add‑ins and scripting consultancies whose pitch is “we’ll automate this with macros.”
More subtly, this also shifts power away from Microsoft’s grip on financial modeling. A standalone IDE that feels like software development, not office work, will resonate with a younger generation of quantitatively minded operators who already live in VS Code and Jupyter.
The bigger picture: from copilots to accountable agents
Meridian slots neatly into two converging trends.
1. The rise of “agentic” AI for work.
We’re moving from autocomplete‑style copilots (finish this formula, rewrite this email) to agents that can:
- plan and execute multi‑step tasks
- call tools and APIs
- coordinate with other agents
OpenAI, Anthropic and others have been pushing this hard with coding agents and “agent teams.” Meridian is applying the same idea to finance: give an AI a modeling objective and a data universe, and let it propose, build and refine the structure.
2. The professionalisation of spreadsheets.
For two decades, there has been a slow drift from spreadsheets to more formal systems: BI platforms, FP&A suites, vertical SaaS like Anaplan, Pigment or Adaptive. Yet Excel never dies, because it’s the sandbox where new logic is born.
What Meridian is really doing is treating that sandbox like an IDE:
- models are entities with history, not random files
- assumptions are first‑class objects, not notes in hidden tabs
- logic is version‑controlled, comparable across scenarios, and re‑usable
Historically, the closest analog is the rise of software engineering practices in data science: notebooks giving way to production ML pipelines with tests, monitoring and governance. Financial modeling is going through the same maturation. The end state isn’t “no spreadsheets,” but spreadsheets that behave like well‑behaved software, with AI acting as a co‑maintainer.
Competitively, this puts Meridian somewhere between:
- Microsoft (Excel + Copilot + Fabric)
- modern FP&A platforms (Pigment, Anaplan, Causal, Cube)
- code‑centric tools (Cursor, Hex, Jupyter‑centric platforms)
The open question is whether enterprises want yet another standalone tool — or whether this approach becomes a feature baked into existing planning suites.
The European angle: AI agents meet hard regulation
For European firms, Meridian’s positioning around “predictable and auditable” is not just a nice‑to‑have, it’s existential.
EU regulation is converging on a clear message:
- The AI Act treats many financial use cases — credit scoring, risk modeling, underwriting — as high‑risk systems requiring documentation, transparency and human oversight.
- GDPR and Schrems II make transatlantic data transfers painful unless vendors can prove strong safeguards.
- Sectoral rules from the EBA, ESMA, ECB and national regulators increasingly ask: who owns this model, how was it built, who changed it, where is the evidence?
A black‑box “LLM built this spreadsheet, trust us” story simply won’t fly in this environment.
If Meridian can truly deliver deterministic, inspectable logic — where every assumption and transformation is visible and exportable — that becomes a regulatory feature, not just a technical one. You can imagine:
- automated model documentation for auditors and regulators
- reproducible back‑testing for risk teams
- clear segregation of personal vs non‑personal data to stay on the right side of GDPR
But there’s a catch: Meridian is US‑based. Large European banks, insurers and fintechs will ask hard questions about:
- data residency (EU data stored and processed in the EU?)
- use of US‑hosted foundation models
- contractual guarantees for audit access and retention
This opens space for European competitors to build “AI‑first financial IDEs” that are EU‑native in both tech stack and compliance posture.
Looking ahead: what to watch in the next 24 months
Several fault lines will determine whether Meridian becomes a category‑definer or just another interesting AI demo.
1. Can they prove real ROI over Excel + Copilot?
To justify a dedicated workspace, Meridian has to show order‑of‑magnitude gains:
- days of modeling compressed to minutes
- fewer manual errors
- faster deal cycles or planning iterations
If it’s merely “slightly nicer formulas plus a chat box,” enterprises will default back to Microsoft.
2. Governance and integrations will matter more than raw IQ.
CIOs and CFOs will ask:
- Can I run this in a private VPC or on‑prem?
- How does it log and version every change?
- Does it integrate cleanly with my data warehouse, ERP, CRM and existing FP&A tools?
The winners in this space won’t necessarily have the smartest model; they’ll have the least painful answers to risk, security and integration questions.
3. Expect consolidation or copycats.
If Meridian gains traction, there are obvious acquirers: established FP&A vendors, large cloud providers, even Big Four consultancies that want to productise their spreadsheet audit practices. At the same time, don’t be surprised when Microsoft, Google and maybe even ServiceNow ship their own “agentic modeling workspaces.”
Timeline‑wise, 2026–2027 will likely be the proof‑point years: pilots today, limited production next year, and only then broad rollouts. We’re early.
The bottom line
Meridian is interesting not because it puts AI in spreadsheets — that ship has sailed — but because it treats trust as the real product. In a world where LLMs can fabricate numbers with a straight face, the scarce resource is not intelligence, it’s accountability.
If your company is already experimenting with AI‑generated models, the pressing question is simple: two years from now, will you be able to explain to an auditor — or a regulator — exactly how those numbers were produced? Meridian is betting that you’ll want the answer to be “yes,” and that you’ll pay handsomely for it.
