February 2026

The Excel AI Revolution Has Started

How to Use Shortcut

Nico Christie

Nico Christie

February 19, 2026

Introduction

Shortcut is the best way for professionals to do their spreadsheet work. It is built by Fundamental Research Labs, a research lab that spun out of MIT in late 2023. Fundamental is led by Dr. Robert Yang — a computational neuroscientist and former MIT EECS and BCS faculty assistant professor — and Nico Christie, co-founder, chief product officer, and creator of Shortcut.

The product was born out of our team's computer-use agent research, which doubled the state of the art on OSWorld, the premier benchmark for measuring agent computer use for knowledge work. Among OSWorld's many categories, LibreOffice Calc stood out: by the end of 2024, our agent scored 68% against a 62% human baseline — the first time any AI system had surpassed human performance on spreadsheet tasks in the benchmark. We took those research learnings, raised $40M from Andreessen Horowitz, Eric Schmidt's deep tech fund First Spark, Prosus Ventures, Patrick Collison, Balaji Srinivasan, and others, and launched Shortcut six months later as the first superhuman Excel AI agent.

Since launch, the product has improved significantly and has evolved toward being an irreplaceable tool in the hands of spreadsheet professionals — one that dramatically uplevels both work quality and speed.

Setting Up and Getting Started

  1. Your enterprise team admin will add you to your team plan, and you will receive an invite to accept and join Shortcut.
  2. Add the Shortcut add-in from the Microsoft 365 store, the Google Workspace Marketplace, or use the web version directly.
  3. Once added, open Shortcut from the Add-ins button in any Excel file or Google Sheet.
  4. You will be prompted to log in. We support SSO login for enterprises.

Here are some sample tasks you should run to quickly learn about Shortcut's capabilities:

  • Open an empty Excel file with the Shortcut add-in and ask it to build a three-statement model for any public company. You will immediately see what it means for an agent to create tabs, write formulas, pull data, and structure a model end to end.
  • Take a real model you are already working on — something unfinished or something you need to QA — and ask Shortcut to review it. It should become immediately clear that agents are far superior to humans at auditing and reviewing Excel work.

Getting the Most Out of Shortcut

There is a quiet spreadsheet revolution currently underway. Shortcut experts are turning days of work into hours, but there is a genuine learning curve. It is our goal to guide you through it. We genuinely believe we are at a point where not using Excel AI means you are at risk of falling behind — but you cannot use it for everything yet with equal success.

There are three areas where using Shortcut is already vastly superior to doing work manually:

  • Auditing and reviewing spreadsheets. Hand Shortcut a 50-tab workbook and ask it to check every formula, flag circular references, and surface risks. It reads every cell, catches errors humans skip, and does it in minutes instead of hours. Use it to QA models before they go out the door, verify assumptions after a handoff, or stress-test a model you inherited.
  • Data extraction into your spreadsheet. Upload 10-Ks, CIMs, pitch decks, scanned PDFs, or any document with structured data. Shortcut's custom extraction pipeline is dramatically more accurate than general-purpose AI on real-world large and messy PDFs and attachments with images. It parses tables, footnotes, and even embedded chart images directly into your model. Every extracted cell includes a source citation you can click through to the exact page and highlighted passage in the original document. Stop copying numbers by hand and start auditing agent-extracted data instead. Shortcut supports up to 5GB of attachments of any type — 150x larger than the nearest competitor.
  • Turning repeated workflows into routine executable Skills. Once you have a process that works — a monthly reporting template, a standard valuation setup, a data normalization pipeline — turn it into a Skill. Upload examples of before-and-after work, ask Shortcut to build the Skill for you, and have it interview you until it has a perfect understanding of the workflow. Trigger any Skill by sending /skills in the chat. This is how teams go from one analyst doing the work to the entire group running the same playbook at the same quality level.

Five Key Features to Use

Once you have the basics, a few practices and features make the biggest difference:

  1. Plan Mode. Use Plan Mode when your task feels sufficiently difficult, or when you are not absolutely certain what the right prompt is. Plan Mode will explore your sheet for necessary context and interview you before writing the main prompt for you. It typically educates people on the ambiguities of their prompts, and often points to better solutions than what they had initially had in mind.
  2. Skills. Upload examples of before-and-after work for any routine process, then ask Shortcut to build a Skill for you. Tell it to interview you until it has a perfect understanding of the workflow, and it will create the Skill. Then have Shortcut execute it once, share feedback on how the Skill should be improved or flag any issues it ran into, and it will refine and further improve it. Your whole team can trigger any Skill by typing /skills into the chat. You can also download hundreds of community-made Skills or share your own with your team.
  3. Parallel agents. Spin up multiple chats in the same file in parallel to get the most out of Shortcut. Only one agent should be taking actions on your workbook at a time, but you can have several others planning and reviewing in parallel. A frontier technique is to use two to three agents ahead of work to plan and explore, and another two to three to independently grade or review agent work. With enough planning, you can always arrive at the ideal action plan, and with enough verification, you should be able to catch any red flag or error.
  4. Auditing attachment extraction accuracy. You should always use Shortcut to extract data from attachments, but you should also use Shortcut to audit the outputs. Click on any cell with a source citation to open Source Preview and verify the extracted value against the exact page and highlighted passage in the original document.
  5. Working memory. Tell Shortcut your formatting preferences, formula conventions, and naming standards early. It remembers these across sessions, which means the more you use it, the less you need to repeat yourself.

Where Shortcut Excel AI Is Going in 2026

A year ago, Shortcut could hardly sum columns together. Now, it is solving some of the most sophisticated spreadsheet work in the world in the hands of top analysts at leading financial institutions. The rate of improvement is not slowing down. You should expect that Shortcut will be better than any human at any manual Excel work by the end of 2026.

The second and third order effects of this are still unfolding, but we believe the most immediate impact is clear: teams will get back the thousands of hours they currently spend manually wiring and updating models, and redirect that time toward what actually matters — pressure testing assumptions, developing new investment theses, stress testing existing models under different scenarios, and spending more time on the decisions that move outcomes. The bottleneck in finance has never been the ability to build a model. It has been the time it takes to build one well enough to trust. That bottleneck is disappearing, and what comes after it — when building and verifying models is essentially free — is something the industry has never had to think about before.

Try Shortcut Today

Experience the most powerful Excel AI agent available. Start free or talk to our team about enterprise deployment.