Via Every

“I think it’s the hardest part of AI, actually. And this is the part that has been so magical at this particular [private equity] firm that we’ve been working with. Our partner basically interviewed every single investor and every single team to really understand the nuances with which a team collectively thinks about every part of the investment memo. This work that we’ve been able to do together would not have been possible if it didn’t have such a high degree of tailoring. (This is like Savile Row sort of prompt tailoring.) It’s so so so specific. From the way that numbers show up, the way they express or think internally around this stuff . . . the prompts reflect that. And so the prompts end up being this [AI] analyst that does really high quality work that is dependable.”

— How We Built ‘Claudie,’ Our AI Project Manager (Full Walkthrough), at 14:53

---

Natalia Quintero, Head of AI consulting at Every, shares that the most powerful AI deployments aren’t just about using AI better — they’re about feeding it your organization’s unique knowledge and workflows.

At one private equity firm Every worked with, investors traditionally spent two to three weeks creating investment memos before presenting to the investment committee. With a carefully designed AI system, they now generate high-quality drafts in roughly 30 minutes.

The secret wasn’t just using AI; it was the meticulous groundwork:

• Detailed workflow mapping: Their internal champion interviewed every investor and team member to understand exactly how they approached each part of the investment memo process
• Proprietary context connection: The firm connected AI to a decade’s worth of accumulated knowledge stored in the company’s SharePoint server, which contained rich context about the firm’s investment thesis and sector expertise
• Highly tailored prompts: They created what Natalia calls “Savile Row prompt tailoring”—extremely detailed instructions about the firm’s specific way of presenting numbers, figures, and internal reasoning, to reflect how the team thinks internally about investment opportunities
• Team-specific customization: They accounted for variations between teams with different strategies

The result is an AI analyst that does dependable, high-quality work that reflects the firm’s unique perspective and standards—not generic output.

Connecting AI to your proprietary data sources is table stakes. The real value is unlocked by teaching AI how to interpret that data.

Almost all companies understand the need to connect AI to their data sources—their CRM, knowledge bases, internal documents, and operational systems. But connection alone isn’t enough. The differentiation comes from codifying how your organization thinks about and uses that data.

This was true in the private equity firm described above. It was also true for Every itself. Dan Shipper, Every’s CEO, says there are three places you could find their revenue data: in Stripe, in ChartMogul, or in PostHog. But Every’s Head of Growth has a specific way of defining Monthly Recurring Revenue (MRR) that’s specific to their business model.

Rather than expecting an AI agent to figure this out from scratch every time:

• Document exactly where to find each type of information
• Explain how your organization defines key metrics
• Provide the interpretive framework that makes the data meaningful
• Encode the nuances that distinguish your approach from generic industry practices

Taking the time to teach AI how to interpret data is especially critical as organizations move toward more autonomous AI agents that work independently for extended periods. The agent needs not just access to data, but the institutional knowledge about what that data means and how to use it properly in your specific context.