TL;DR: Crunched uses Claude Sonnet 4.5 and Opus 4.5 to deliver over 50% time savings on Excel modeling for private equity, investment banking, and consulting firms. One real estate investor cut a $100M transaction analysis from 10 hours to 1 hour—a 10x speed improvement. By solving firm-specific pain points one at a time with a sophisticated context searching system built by two McKinsey veterans, they're enabling economically useful work that was previously cost-prohibitive.

Picture this: You're a private equity associate with three days to build a full LBO model from a 200-page information memorandum. Your VP expects granular drivers, verifiable assumptions, and zero linking errors. Meanwhile, the investment banking analyst down the hall just submitted their third version of a model because the VP keeps catching formula mistakes that kill credibility in client meetings.

Welcome to the reality of high-stakes Excel modeling. Crunched founders Philip Borge and Michael Sakowski spent 10,000+ hours in Excel at McKinsey's private equity practice experiencing these exact problems. Models with 50+ tabs and thousands of formulas. Tight deadlines. VPs catching linking errors and logical flaws in first-pass reviews, forcing iteration cycles that eat deal timelines.

The math is brutal: consultants need fully-linked market sizing models with granular, verifiable assumptions but lack time to gather and structure data from multiple sources. Senior dealmakers constantly identify analyses that would strengthen deals but remain too expensive given finite resources. Transaction sizes often don't justify allocating more associates, so valuable work simply doesn't happen.

Borge and Sakowski built Crunched as an AI Excel analyst specifically for the top 1% of Excel users worldwide—the professionals where models are either the main deliverable or the backbone of every presentation.

But here's the thing: they didn't build another generic AI tool. Crunched takes a forward-deployed approach, working on-site with customers to solve firm-specific pain points one at a time.

The platform handles three core workflows:

• Error-checking: Analyzes models of any size, identifying linking mistakes, logical flaws, and methodology errors specific to finance. Serves as a review layer between analysts and VPs—Crunched handles the first-pass, saving iteration cycles and giving VPs higher confidence.
• Model building and editing: Fills out templates, links into input sheets, and builds valuation models, business cases, and analyses from scratch. Need to link a "management scenario" into an LBO model? Crunched handles it. Need a fully-linked market sizing model? Crunched builds it with granular assumptions, verifiable rationales, and clickable links.
• Data gathering: Pulls information from PDFs, images, web sources, and customer-specific databases. Private equity associates can upload an information memorandum and build an LBO with granular drivers tailored to the specific case.

The genius part? Crunched built a sophisticated context searching and management system—their core differentiator—that gathers precisely the right information for each task, enabling accurate operations on workbooks of any size.

Crunched evaluated multiple LLMs before choosing Claude for its combination of speed, accuracy on quantitative tasks, and reliable performance on complex, multi-step financial modeling workflows. They implemented the Claude integration in less than a day.

The Stack:

• Claude Sonnet 4.5: Parallel tool calling gathers context from multiple workbook locations simultaneously, while the extended context window processes more information at once, reducing the risk of missing critical dependencies. Low latency enables the real-time iteration speed that power users demand.
• Claude Opus 4.5: Handles complex workflows requiring deep analysis across multiple financial statements and scenarios.
• Proprietary context searching system: Core differentiator that gathers precisely the right information for each task, augmented with deep domain expertise in Excel modeling best practices, finance methodologies, and consulting frameworks.

Sonnet 4.5 dramatically extended what their context searching system could do, transforming the platform's capabilities for handling massive workbooks with complex dependencies.

Since deploying Claude Sonnet 4.5 and Opus 4.5, the numbers tell a compelling story:

• Over 50% time savings across Excel modeling tasks
• Financial spreads: 8+ hours reduced to 1 hour
• Company write-ups: 8 hours reduced to 20 minutes
• $100M real estate transaction analysis: 10 hours reduced to 1 hour (including tenant research, valuation model building, and analysis error validation)
• First-pass model reviews: Minutes instead of VP oversight
• Market sizing models: Days reduced to hours
• Iteration cycles reduced and deal timelines accelerated

Real-world wins: An investment banking team now completes first-pass model reviews in minutes instead of requiring VP oversight, reducing iteration cycles and accelerating deal timelines. A management consulting firm builds fully-linked market sizing models in hours instead of days, arriving at client meetings with granular, verifiable assumptions that strengthen recommendations. For one real estate investor, a $100M transaction analysis including tenant research, valuation model building, and analysis error validation dropped from 10 hours to 1 hour.

The broader impact extends across project and deal types. Crunched enables economically useful work that was previously cost-prohibitive. Private equity teams now bring full analytical rigor to early-stage deals that previously only received deep analysis at later stages. Investment bankers run additional scenarios that strengthen client recommendations.

Three golden rules from Crunched's AI journey:

• Solve specific pain points, not generic problems. Professionals don't need another DCF template—they already have those. They need help with firm-specific workflows, company-specific formatting, and industry-specific methodologies. Crunched's forward-deployed model tailors solutions one pain point at a time.
• Domain expertise on top of AI foundation wins. Crunched built financial and Excel-specific knowledge on Claude's foundation, augmenting the model with deep expertise in Excel modeling best practices, finance methodologies, and consulting frameworks. The sophisticated context searching system gathers precisely the right information for each task.
• Build the tool you needed yourself. When founders deeply understand the problem from lived experience, they build solutions that actually work. Crunched serves the top 1% of Excel users because the founders are those users.

🐯 Tiger Takeaway

Crunched proved that 10x improvements come from solving real problems with precision. Instead of building a generic AI tool, they built a forward-deployed solution that solves firm-specific pain points one at a time. By layering deep domain expertise on Claude Sonnet 4.5 and Opus 4.5, they're delivering over 50% time savings and enabling economically useful work that was previously cost-prohibitive. The lesson? AI doesn't replace expertise—it amplifies it. When you combine world-class models with sophisticated context systems and deep domain knowledge, you don't just make work faster. You make previously impossible work possible.

🤖

AI-Powered Newsletter