How Meridian is building AI for the world's most-used programming language

Executive overview

Over a billion people use Excel as their primary tool for knowledge work, yet AI has barely touched this domain. John Ling, co-founder and CEO of Meridian, spent a thousand hours testing AI on financial models before anyone else did — and found a massive gap.

Spreadsheet work can be decomposed so AI handles each part of the workflow, enabling a 20x speed improvement. The bottleneck isn't the technology; it's that no one with deep domain expertise has seriously tried.

The biggest AI opportunity isn't coding — it's the spreadsheet.

Why Excel is a massive AI opportunity

• Excel is the most distributed programming language in the world
• Bankers and analysts default to doing work by hand rather than testing AI alternatives
• Coding tools were built by coders, so failure modes are obvious; spreadsheet AI has no equivalent expert builders yet
• Domain experts can tell a model output is wrong but can't articulate why — that gap is the product opportunity

How John developed his edge

• Spent ~1,000 hours testing AI on LBO models and financial workflows before founding Meridian
• At Scale AI, focused on data quality, benchmarks, and using LLMs to improve internal processes
• Immersing in a fast-moving AI ecosystem builds intuition about what is possible now and in 3–6 months
• That forward-looking intuition is a durable competitive advantage

The learning approach behind Meridian

• Treat every problem as a learning opportunity, not just a task to complete
• Read primary research rather than relying on execution momentum alone
• Prompting is a skill: explaining a task precisely to an LLM clarifies your own thinking
• Spending 10,000 hours doing finance with AI will be as valuable as any traditional credential

Building the company and culture

• $15M+ raised; seed led by Andreessen Horowitz general partnership
• Culture explicitly allows people to experiment and fail without penalty
• Hiring for first-principles thinkers who push back on constraints
• Don't assume outreach is pointless — reaching out is always worth trying