Goodbye Excel? AI Agents for Self-Driving Finance – Pigment CEO
Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
Podcast Summary
In this fascinating conversation, physicist-turned-entrepreneur Eleonore Crespo discusses how her company Pigment is revolutionizing enterprise performance management with AI agents. (00:30) The company has raised $400 million and serves Fortune 500 customers including Anthropic, Figma, Snowflake, Uber, and Coca-Cola with 500 employees and 60% of revenue coming from the US. (01:47) Crespo explains how today's accelerating world requires C-level executives to make faster decisions than ever before, and Pigment serves as an "AI enterprise performance management platform" that collects data from all systems to help companies make better, faster decisions based on the right data. The episode explores Pigment's three launched AI agents (analyst, modeler, and planner), the technical challenges of building error-proof AI for sensitive financial data, and the vision of autonomous planning systems that could fundamentally transform how enterprises operate.
- Main Theme: The transformation of enterprise decision-making through AI agents that can perform complex analytical tasks while maintaining accuracy in highly regulated environments like finance and planning.
Speakers
Eleonore Crespo
Physicist-turned-entrepreneur and CEO of Pigment, Eleonore studied fundamental physics and engineering with a parallel focus on entrepreneurship. She spent her career abroad working as a data scientist for the CFO of Google EMEA and Alphabet before joining Index Ventures as an investor, where she learned from some of the best founders on the planet including those from Figma and Datadog. This experience at Index was fundamental in understanding what great businesses look like and gave her the confidence to start Pigment, which she co-founded with Romain.
Matt Turk
Managing Director at Firstmark Capital and host of the MAD podcast. Matt was one of Pigment's first investors and has been instrumental in helping the company build its presence in the US market, including assistance with building their executive team and finding their first US customers.
Key Takeaways
Build AI Agents That Enable the Impossible, Not Just Automation
The most transformative AI agents don't just automate mundane tasks—they enable capabilities that were previously impossible for humans to achieve. (23:59) Crespo explains how large companies have combinations of products, countries, business units, and cost centers that make comprehensive analysis literally impossible to do at scale. For example, running 5,000 scenarios in parallel on infinite quantities of data is something no human could accomplish, but AI agents can. This approach goes beyond simple time-saving to unlock entirely new use cases and strategic capabilities that were previously out of reach.
Accuracy Over Speed: Constrain Agents to Eliminate Hallucinations
When dealing with sensitive financial data, accuracy must take precedence over speed, requiring a fundamentally different approach to AI implementation. (27:43) Pigment solves the hallucination problem by constraining their agents to work within the structured data environment of their platform, where every calculation is performed on the Pigment system rather than relying purely on generative AI outputs. This creates comprehensive audit trails and allows for human validation while maintaining the accuracy required for enterprise financial decisions. The key insight is that structured data environments allow for error-proof AI when properly constrained.
Reskill Teams to Become Strategic Decision Makers
AI agents transform job roles from data gathering and manipulation to strategic thinking and decision validation. (53:31) Rather than eliminating jobs, AI enables "everyone in a company to become a mini CFO" by providing them with analytical capabilities previously reserved for specialists. The new skill set involves learning how to give context and framework to agents, validate their outputs, set appropriate guardrails, and make strategic decisions based on comprehensive analysis. This represents a fundamental shift toward higher-value work focused on vision-setting and strategic thinking.
Design Collaborative Agent Systems with Human Supervision
Effective enterprise AI requires multiple specialized agents working together under human oversight, not a single all-purpose system. (22:22) Pigment's three agents (analyst, modeler, planner) each handle different aspects of the workflow—the analyst explains data and handles recurring analysis, the modeler adapts business models rapidly, and the planner generates multiple scenarios. These agents coordinate through a supervising system but always maintain human oversight for final decision-making, especially given the sensitive nature of financial data. This collaborative approach ensures both specialization and reliability.
Build Global Companies by Leveraging Technology for Distributed Teams
Modern technology enables building global category-defining companies without relocating to traditional tech hubs, but requires intentional strategy and adaptation. (59:43) Crespo built Pigment from Europe while capturing 60% of revenue in the US by leveraging COVID-era remote work adoption, working late hours to accommodate US time zones, building a US-based executive team, securing US investors, and maintaining frequent in-person touchpoints. The key is creating systems and relationships that make geographic boundaries irrelevant while maintaining the cultural and talent advantages of your home base.
Statistics & Facts
- Pigment has raised $400 million in venture capital and employs 500 people, with the company still growing 2x year-over-year. (03:30)
- 60% of Pigment's revenue comes from the US market, and they are growing 4x in the number of large Fortune 500 customers they serve. (03:54)
- Pigment typically becomes useful when companies reach approximately 1,000 employees, indicating their focus on large enterprise customers. (04:49)
