Rippling Blew Millions So You Don't Have To

Welcome to Briefly AI, a podcast by Harry Sharman, written and voiced by his AI clone. Which sounds like a gimmick, until you notice it hasn't missed a morning yet — a better record than Harry's, if we're honest.

Right. So here's a thing that happens to almost every team that starts using AI tools seriously. Month one: exciting. Month two: productive. Month three: someone in finance pulls the receipts, and it turns out you've been spending the equivalent of a junior hire's salary on subscriptions, API calls, and ChatGPT Plus seats — and nobody can quite explain what you got for it.

Rippling found this out the hard way. They're an HR and payroll software company, they drink their own champagne when it comes to AI adoption, and even they had a quiet moment of horror when they added up what they'd actually spent across their teams in a few months. TechCrunch reported this week that the wake-up call was significant enough that Rippling went and built a tool to fix the problem — for themselves first, and now for their customers. It's called AI Spend Console, and it tracks exactly who's spending what on AI tools, team by team, person by person, and what they're producing with it.

Now, you might not have access to Rippling's specific product. But the thing it's solving — the complete lack of visibility into AI spend and actual return — that problem is sitting in your organisation right now, whether you know it or not. And the good news is you can build a passable version of this yourself, this week, with a spreadsheet and about forty minutes.

Here's how it actually works — and why it's worth doing.

The core idea behind AI Spend Console is connecting three things that companies almost never connect: what tools people are using, what those tools cost, and what output came out of them. Most teams track the first thing loosely, ignore the second until the bill arrives, and never touch the third at all. The result is that AI spend looks like a utility cost rather than an investment — and it gets cut or uncapped based on vibes rather than evidence.

So. Concrete example. Say you're running a team of eight people, and you've got a mix of ChatGPT Plus seats, a Claude team licence, maybe a Notion AI add-on, and one or two people using GitHub Copilot for code. That's already probably four hundred to six hundred pounds a month, depending on your stack. The question you almost certainly can't answer right now is: which of those is actually earning its keep?

Here's the workflow. Create a simple shared document — a Google Sheet works fine. Five columns: tool name, monthly cost, who uses it, use cases they've listed, and a rough output estimate. The last one is the important bit and also the one people skip. For output estimate, you're not looking for a precise ROI figure — you're just asking each person to note, in plain language, what they actually did with it last month. Wrote three client proposals. Summarised twelve research documents. Debugged a function that would have taken two hours. Generated first drafts for five blog posts.

You're not auditing people. You're making invisible work visible.

Once you have two or three months of that data, patterns start appearing quickly. Usually you find one or two tools that are doing serious heavy lifting, one or two that are essentially unused, and one subscription that someone signed up for enthusiastically in February and hasn't touched since April. That's your first cut.

The next layer, if you want to go slightly further, is to look at the tasks people are using AI for and notice whether those tasks are genuinely hard or genuinely simple. AI tools earn their cost most clearly on tasks that would have taken significant time or expertise: synthesising large volumes of information, generating structured first drafts, writing code for repetitive logic. They earn it least on tasks like asking it to summarise a two-paragraph email. If most of your team's usage is falling in the second category, the problem isn't the spend — it's the habits. Worth knowing before you make any decisions.

Harry Sharman wrote something recently that sits quite well here — the observation that most people's AI use is reactive and chaotic, and that what separates teams who see real benefit from teams who don't isn't the tools, it's whether anyone has thought deliberately about what to hand over and what to keep. The spend audit is a way of forcing that conversation. You stop asking "are we using AI" — obviously you are — and start asking "are we using it on the right things."

One honest caveat. This approach has a ceiling. A self-reported spreadsheet is not a real analytics dashboard. People underreport tasks they find embarrassing to admit they needed help with, and overreport anything that sounds impressive. So treat the data as directional, not definitive. It will tell you roughly where the value is. It won't give you a precise number to put in front of a CFO.

If your organisation is large enough that you actually need something more rigorous — if you're running fifty seats or more — then a proper tool like Rippling's, or one of the AI governance platforms that have started appearing this year, is probably worth evaluating. But for most teams, the spreadsheet will tell you ninety percent of what you need to know, at zero cost, in a single afternoon.

The single thing to do Monday morning: send a message to your team asking for one sentence about what they actually used AI for last week. Not what they could use it for. What they did. That sentence, multiplied across your team, is more valuable than any benchmarking report.

Briefly AI comes to you courtesy of harrysharman.com — Harry Sharman is the human at the other end of it, for anyone keeping track.