Copilot vs Others in Excel: Surprising

Key insights

• Test setup
  Video compares four AI assistants in Excel across five real-world tasks.
  Scoring focused on speed, accuracy, and presentation quality.
• Scenario types
  Tasks included PDF import and extraction, workbook comparison, forecasting, error detection in financial models, and pivot-table analysis.
  These reflect typical finance and data-cleaning workflows.
• Results summary
  Claude generally led for speed and usable output, while Copilot—despite being native to Excel—often trailed on complex tasks.
  ChatGPT performed weakest on full workbook work and missed some key items. Tracelight was mentioned but appears unclear or mixed with other tools in the test.
• Tool strengths
  Claude handled cross-workbook tasks and planned steps before building outputs.
  Copilot used native Excel formulas and showed step transparency.
  ChatGPT offered flexible language help but needed more structure for models.
• Limitations
  All tools need human review: rate limits, missing items, and methodology errors appeared across tests.
  Presentation looks vary and some outputs required cleanup or correction.
• Practical advice
  Choose a tool based on the task: use Claude for complex cross-file builds, Copilot for quick native-formula work, and treat ChatGPT as a drafting assistant.
  Always verify formulas and assumptions before publishing or using results in decisions.

Video Overview and Context

Kenji Farré (Kenji Explains) [MVP] recently published a hands-on comparison that puts four leading AI tools to work inside Excel. In the video, he tests Claude, Copilot, ChatGPT, and Tracelight across five practical scenarios to evaluate speed, accuracy, and presentation quality. Consequently, the piece aims to show how these assistants handle typical finance and data tasks rather than delivering lab-style benchmarks. Overall, the approach centers on real spreadsheets and real problems to reflect everyday user needs.

Methodology: Five Real-World Scenarios

Kenji structures the test around five scenarios: importing and extracting data from PDFs, comparing two workbooks to spot differences, forecasting in a financial model, finding errors in models, and manipulating data with pivot tables. He scores outcomes on three axes—speed, accuracy, and aesthetics—so viewers can weigh practical tradeoffs rather than a single metric. Moreover, he narrates each step and shows the tools’ actions inside Excel so the audience can judge traceability and edits. This hands-on format emphasizes what professionals actually face when automating spreadsheet work.

Key Findings from the Head-to-Head

Across the scenarios, Claude consistently ranked high for speed and for producing ready-to-use outputs, often offering a clear plan before making changes. Meanwhile, Copilot surprised some viewers by lagging behind despite being native to Microsoft 365; it performed reliably on basic formulas but sometimes produced simplistic approaches when models needed nuance. In contrast, ChatGPT delivered mixed results: it can generate useful formulas and explanations but occasionally omits critical items or creates structures that need significant edits. Finally, Tracelight showed competitive strengths in specific tasks, with the video noting it handled several workbook-level operations smoothly, though Kenji points out limits when tasks required heavy cross-file logic.

Tradeoffs: Speed, Accuracy, and Usability

The test highlights clear tradeoffs. For instance, a faster result can be less auditable, so speed without transparency risks hidden errors, whereas a slower tool that explains its steps improves trust but costs time. Furthermore, native integration like that offered by Copilot brings familiar Excel functions, which helps with adoption, yet it does not guarantee superior modeling outcomes. Therefore, users must balance immediate productivity gains against the need for reliable audit trails and domain-specific accuracy.

Challenges Identified in Practical Use

Kenji’s demonstrations underscore several recurring challenges. Large PDFs and multi-tab models expose limits in document parsing and cross-workbook logic, and some tools hit rate limits or struggle with very complex forecasting scenarios. Additionally, the video shows that automated fixes can introduce subtle formula errors if the model’s intent is not clearly expressed, so manual review remains essential. As a result, the real risk is over-reliance on a single AI step without human oversight.

Implications for Finance and Data Teams

For finance and analytics teams deciding which tool to adopt, the lesson is to match the tool to the task rather than assuming one assistant fits all needs. In practice, teams might use Claude for rapid model construction and exploratory analysis, rely on Copilot for tasks tightly tied to Excel functions and Microsoft workflows, and reserve ChatGPT or Tracelight for targeted data prep or prototype work while verifying results carefully. Moreover, combining tools and maintaining clear review steps often yields the best balance of speed and accuracy.

What Kenji’s Video Adds to the Conversation

Unlike abstract benchmarks, Kenji’s video provides narrated, timestamped demos of each scenario, which helps practitioners see how outputs evolve and where manual fixes are needed. Consequently, the presentation quality and the step-by-step playback help viewers judge whether a suggested solution fits their internal controls and report standards. The video therefore acts as a practical guide rather than an endorsement, and it underlines that human expertise still plays a central role. Finally, viewers are encouraged to test these tools on their own data before changing established workflows.

Bottom Line

In summary, Kenji Farré’s comparison offers a useful, work-focused snapshot of AI assistants in Excel. While Claude often leads on speed and usable output, Copilot brings integration benefits but may underperform on complex modeling, and ChatGPT and Tracelight each present situational strengths and limits. Therefore, organizations should evaluate tools against specific needs, enforce review practices, and avoid blind trust in any single automated approach. Ultimately, these AIs can boost productivity, but they work best as partners to experienced users rather than replacements.

Keywords

Claude vs Copilot vs ChatGPT vs Tracelight Excel comparison, Best AI for Excel, Excel automation with AI, AI assisted Excel formulas, Excel data analysis with AI, AI productivity tools for Excel, ChatGPT Copilot Claude Tracelight comparison, Top AI Excel assistants