Shortcut launches as an AI Excel agent that beats Goldman and McKinsey first-year analysts 89% of the time

Summary

Shortcut, a Menlo Park-based startup, launched publicly on July 28 with an AI agent purpose-built for Excel financial modeling. The product's headline claim is striking: in a blinded benchmark, Shortcut outperformed incoming first-year analysts from Goldman Sachs, McKinsey, BCG, and JP Morgan 89% of the time, as judged by managers from those same firms. Analysts were given over 90 minutes per task across five categories including op models, M&A models, LBO, DCF, and product dashboards. Shortcut completed comparable work in a fraction of the time.

The benchmark methodology has an honest caveat built in. The founder acknowledges that first-year analysts are, bluntly, not especially skilled at Excel modeling, and that giving humans a full week rather than 90 minutes might have tightened the gap. Still, the directional result is consistent with what coding tools like Cursor demonstrated for software engineering around August 2024, which is the explicit analogy Shortcut is leaning into.

Where the Product Works and Where It Doesn't

Shortcut's strongest fit is in private equity and corporate real estate, where analysts receive confidential investment memos (CIMs), typically multi-document PDF packages, and need to rapidly build or populate LBO and DCF templates. The workflow now lets a PE analyst attach 100 CIMs and request 100 models back, shifting the bottleneck from model construction to data review.

The product is weakest at editing large, complex existing templates with new data, which the founder identifies as the critical capability threshold it has not yet fully crossed. On a metaphorical timeline, if August 2024 represents coding-agent maturity, Shortcut currently sits in June. Hedge funds relying on real-time data feeds and proprietary low-level systems are explicitly not the near-term target market.

FP&A at small and mid-sized businesses, solo operators, and anyone building financial models from scratch represent the more immediately accessible market. Writing Excel formulas, the founder argues, is structurally easier for AI than writing complex code, which raises the question of why a capable product didn't exist sooner.

The Observability Problem

A recurring theme is that accuracy alone is not the bar. Drawing on the Cursor analogy, the founder argues that what unlocked AI coding tools for enterprise use was not raw accuracy but observability, specifically the ability to see exactly what changed and why. Shortcut's launch includes a diff UI inspired directly by Cursor's apply-diff function, with source traceability down to the specific page of a 10-K or web search that produced a hardcoded value. The framing is that perfect traceability matters more than 99.9% accuracy, because human analysts also make errors and the real need is the ability to audit outputs quickly.

This connects to a broader concern raised about the Metr study, which found that experienced software engineers using AI coding tools on complex open-source bugs were approximately 20% slower, not faster. The Shortcut founder's response is that using AI is itself a skill, and that even veteran practitioners eventually find high-leverage patterns. For finance, the product burden is greater because Excel users are less technically sophisticated than engineers, meaning the product must abstract away failure modes rather than expecting users to work around them.

Competitive Positioning

On hyperscalers, the view is essentially that Microsoft and Google could build this but haven't because of organizational inertia and misaligned priorities, not capability. On OpenAI specifically, the assessment is more cautious: competition would be bitter, but OpenAI is currently spread across phones, browsers, coding environments, and research priorities, placing a dedicated Excel agent low on its stack.

The founder's core competitive argument is philosophical: frontier lab researchers largely believe the industry is moving toward pure input-output models where structured formats like Excel become irrelevant. Shortcut is betting there are billions of dollars of value to capture in the transition period before that happens, if it happens at all.

Go-to-Market

The company is pursuing a bottoms-up, prosumer-led distribution strategy, consistent with how Cursor grew, while leaving the door open to an enterprise motion. The founder reports that CIOs at major banks have expressed immediate interest. The current plan is to let product traction dictate resource allocation between the two paths, with the founder acknowledging his comparative advantage is on the bottoms-up side.