"When NOT to use the COPILOT function
COPILOT uses AI and can give incorrect responses.
To ensure reliability and to use it responsibly, avoid using COPILOT for:
- Numerical calculations: Use native Excel formulas (e.g., SUM, AVERAGE, IF) for any task requiring accuracy or reproducibility.
- Responses that require context other than the ranges provided: The COPILOT function only has access to the prompt and context provided to or referenced by the function. It does not have access to other data from your workbook, data from other files or enterprise information.
- Lookups based on data in your workbook: Use XLOOKUP to look up data based on a table or range.
- Tasks with legal, regulatory or compliance implications: Avoid using AI-generated outputs for financial reporting, legal documents, or other high-stakes scenarios.
- Recent or real-time data: The function is non-deterministic and may return different results on recalculation. Currently, the model's knowledge is limited to information before June 2024."[0]
Manager: "What's the rate of return according to our financial model?"
Analyst: "Let me vibe the answer for you. Just a sec."
Amazing
What really is lacking here and what AI does not solve, is the problem of communication and the problem of work as time wasting anti-productivity.
It's so easy to mess up on Excel, there are no crystal clear diffs like on Git. And even if that on Git we make mistakes.
Very often you have data that affects other data etc.
Also, we have DBs and other resources (even MS Access...) because we know how error-prone are Excel sheets.
If using AI adds another layer of potential problems with my Excel sheets, I'm not sure how using it would improve my overall work.
I believe they'd need to change the Excel quite a lot in order to make using AI 'excel'.
Maybe the new Excel is a new Excel-esque application with a new layer of tools to enable it to work properly with AI.
Like something that allows you to validate against a set of values described by prompt etc.
There are so many possibilities now with LLMs and I feel those integrations are very shallow, more of the same thing.
I don't need a Clippy on steroids. I want a new experience.
There is SOMETHING there though, I think if you could show the flow of data in the same space as the data itself, you'd have a worthy Excel-like thing that does well with an AI agent jumping in to either fix something or attempt to irreparably break everything. Rather than looking at a diff grid of cells with hidden expressions, or a spaghetti mess of nodes with lines, I think a structured and shallow-nested tree would be easier to understand... maybe. Rather than expressions, you'd have features like in CAD that act on some base table, and you could jump back to the table as it was at a state where that feature was applied, make a tweak, and jump forward again.
Eg. Adding an item-specific tax calculation is a single operation in between the items + sum feature, rather than two operations where I have to hook up items => itemsWithTaxes, then tweak sum to read from each itemsWithTaxes. If I'm ever confused about what sum is doing while setting this up, I can go "back" in time to the state the sum feature works on, and I only see the data it has to work with.
https://www.theverge.com/2023/10/21/23926585/microsoft-excel...