Claude in Excel Is Insane
Directionally Correct Newsletter, The #1 People Analytics Substack
By: Cole Napper
I don’t normally post articles in back-to-back weeks, but this was too good to not share.
If you’ve been following along with Anthropic’s releases lately, you know things have been moving fast. With the release of Opus 4.6 (and now Sonnet 4.6) alongside the Claude in Excel integration, something pretty significant just happened for people analytics, and data analysis in general.
The “Analysis” phase of people analytics just collapsed.
Let me show you what I mean.
I started with a simple dataset in Excel. Nothing fancy, fake data with columns for gender, salary, and job satisfaction. The kind of thing you’d throw together for a quick workforce analysis.
Then I opened the Claude sidebar and typed ONE prompt. One. I didn’t specify formulas. I didn’t tell it which statistical tests to run. I asked it to build a dashboard. I just described what I wanted to understand about the data.
Claude iterated for probably five minutes. I watched it think, plan, create tabs, write formulas, build charts. And it just worked.
It came back with a full dashboard on a brand new tab with all the results. Every single one was correct. You can hover over the cells and see the Excel formulas actually working underneath. It’s a living, breathing Excel workbook with real formulas driving real results.
But it didn’t stop there. It also created a scenario planning tool that allows you to model out what happens when you adjust variables. Want to see the impact of a 5% salary increase for a specific group? Slide the inputs and watch the model update. This is the kind of thing that would have taken an analyst hours to build, and would be completely out of the reach of the lay-person.
And most importantly, it created a key findings tab. A clean, structured dataset summarizing the critical insights from the analysis. Not a wall of text. Not a generic summary. An actual findings dataset you could hand to a CHRO and have a real conversation about.
Let that sink in for a second.
One prompt. Five minutes. A complete, accurate, interactive analytical workbook with dashboards, scenario planning, and summarized findings. All in Excel. All with working formulas you can audit.
The value of “analysis” in people analytics is collapsing before our eyes.
I wrote about this back in 2023 when I introduced the concepts of The Inquisitor and The Change Agent. The argument was simple: As generative AI matures, the ability to run an analysis in people analytics stops being the bottleneck. The bottleneck shifts to asking the right questions (The Inquisitor) and driving action on the answers (The Change Agent).
That future isn’t coming. It’s here.
Welcome to a new, democratized world of data analysis. The people who thrive won’t be the ones who know the most complex Excel formulas or the fanciest statistical methods. They’ll be the ones who know which questions matter, how to interpret what the data is telling them, and how to turn insight into organizational change.
These are exactly the kinds of skills you’ll learn in the Data Driven HR Academy. Check it out now.
I hope you like this article. If so, I have a few more articles coming out soon. Stay tuned. If you are interested in learning more directly from me, please connect with me on LinkedIn.
Cole’s recent articles
For access to all of Cole’s previous articles, go here.
Great demo, Cole! And I'd gently complicate one assumption underneath it.
In The Practical Guide to HR Analytics, my co-authors and I argued that data literacy wasn't just about producing analyses, it was about being a savvy enough consumer of analytics to distinguish between robust findings and noise. That caution matters more, not less, when the execution barrier drops. When analysis is hard to produce, there's a natural filter: the effort involved slows things down and invites scrutiny. When one prompt can generate a full dashboard in five minutes, the risk isn't that people won't have analysis. It's that they'll have too much of it, produced too easily, with too little critical examination of what's actually being measured and what the findings actually mean.
Gender=1 earns 59% more on average is a finding. Whether that reflects discrimination, selection, job level, or industry composition requires exactly the kind of judgment your Change Agent has always needed. The tool doesn't tell you what question to ask, and it doesn't tell you when the answer is misleading. I'm genuinely excited about what you're showing here. I just think the literacy argument from the book becomes more urgent as the execution argument goes away — not less. And maybe just a stronger argument for what you're offering in the Data Driven HR Academy!
This also fundamentally changes how we can (and should?) teach People Analytics....