Your Team Loves Their AI Tools — But Who’s Watching the Meter?
You’ve probably told your team to use AI to get more done. Maybe you’ve even paid for a few premium subscriptions. But here’s the awkward question: do you actually know what your employees are doing with it? Not just how much they use, but whether the output is any good?
Rippling, a US HR software company, thought it had AI under control. Then in March, finance pulled up a number that left executives “incredulous.” The company was on track to burn resources equal to 40% of its entire engineering headcount budget on AI tokens. Usage was growing 80% month-over-month. In other words, the more employees used AI, the more it consumed — and the value they were getting in return was unclear.
The scary part is that this could easily happen to a business your size. You don’t need thousands of employees to hit the same problem. You need just a handful of eager teammates who love to ask ChatGPT for advice, or a developer who keeps the newest model open all day. That’s the setup.
TL;DR: AI usage can explode quietly. A few people will drive most of the burn, and simple fixes like choosing the right model for the right task can cut usage dramatically. Here’s how to keep your team productive without letting the AI meter run wild.
What This Means: “Tokenmaxxing” Explained
The article introduces the concept of “tokenmaxxing” — the activity of using AI tools as much as possible, often without any consideration for how much it’s costing the business. A token is the unit of text that AI models read and generate. Every prompt you type and every response you read uses tokens. And not all tokens are equal: asking a frontier model to draft a one-line email is like driving a sports car to the grocery store.
Rippling discovered its engineers were defaulting to the newest, most resource-heavy models for everything. The company also found that 10–15% of employees were driving about 60% of total AI usage. One engineer alone was consuming a month’s worth of tokens that a small team would use in a year. That single person can throw your entire AI environment out of balance.
Rippling’s response was twofold. First, they built an “AI gateway” that routes each prompt to the most effective model for the task. Second, they started tracking how many prompts translate into actual work output like pull requests and lines of code. After these changes, they kept using about 600 billion tokens per month — but their monthly AI bill dropped to just 37% of what it had been at the peak.
“The inference providers, like Anthropic and OpenAI, have absolutely no incentives to help you control your spend. They have every incentive for it to be a runaway expense.” — Matt MacInnis, Rippling
How This Applies to Malaysian SMEs
If you run a smaller business — say, a 30-person agency or a 10-person startup — you might think this story is only for tech giants with R&D departments. But look closer: Rippling’s problem isn’t scale, it’s visibility. And that problem is likely in your business right now.
Think about your own team. You’ve got a marketing executive who writes social media captions with ChatGPT. You’ve got a salesperson who uses AI to improve their emails. Your accountant might use AI to summarize invoices. Each person has a separate subscription, or maybe they share one login. Have you ever looked at the monthly usage report of an AI tool? Many Malaysian SMEs don’t even know those reports exist. The first step is to access them, even for a single tool.
The Rippling data tells us that roughly 60% of AI usage comes from just 10–15% of the team. That means the opportunity is not to police everyone; it’s to find the heavy users and check whether their output is worth the extra resources. In a small team, you can do this in a casual way. Sit down with your developer who’s using AI to generate code, and ask them to show you what they ship. Look at the code review comments. Has the code been accepted, or are you seeing “please fix this” notes from peers?
The same applies to your marketing team. If they’re using AI to draft blog posts, are you seeing quality output or a fair amount of “AI slop” that needs heavy editing? “AI slop” means content that looks polished but is empty or repetitive. It might be better to have a human write it from the start. You don’t need a fancy dashboard for that kind of assessment — just a few minutes of inspection each week.
For Malaysian SMEs, the practical fix is about process, not software. You can create a simple rule: use free AI tools for internal drafts, and reserve premium tools for client-facing work. Set up a short weekly huddle where the team shares the AI output they’ve used. Most importantly, make sure you have one “AI captain” who is responsible for keeping everyone updated on the best tools and practices.
Practical Takeaways
- List every AI tool your team uses. Include free accounts and trials you’ve forgotten about.
- Check the usage dashboards. Even the free tier of most AI tools shows consumption.
- Identify your top 3 heavy users. Are they producing more output? Is it good output?
- Create simple prompts/rules: “For a quick reply, use the standard model. For a key project, use the advanced one.”
- Review sample AI output once a week. Does it sound like your company, or does it sound like a robot?
- Appoint an “AI captain” who shares learnings and helps your team use AI responsibly.
| Number from Rippling | What it tells you |
|---|---|
| 80% month-over-month growth | AI usage can spike fast. If you don’t check, it’ll become a bigger problem. |
| 10–15% of employees responsible for 60% of usage | Heavy usage is concentrated. Talk to these people first. |
| 605 billion tokens in a single month | AI is doing a lot of work. You just need to know what you’re getting back. |
| 37% of original spend after routing to better models | Simple changes in how you choose tools can shrink your AI usage dramatically. |
The Bigger Picture
The Rippling story is a bookmark moment. AI is no longer a shiny new thing; it’s a utility. But utilities require metering. The companies that will succeed over the next five years won’t be the ones with the most AI subscriptions. They’ll be the ones whose AI usage is tied directly to outcomes — like “we onboarded more customers” or “our code reviews are cleaner.”
That doesn’t mean you have to build an AI gateway in your SME. It simply means you should get into the habit of asking: what did this AI tool produce for my business today? If you can answer that, you’re ahead of most companies three times your size.
Rippling’s CEO Parker Conrad has also suggested that access to AI might not be a blanket privilege for all employees — it could be based on demonstrated productivity. If your company gets to that point, it would be a big shift. But you can start shaping that future now with a simple guideline: AI usage is a privilege, not a free-for-all.
The best lesson for Malaysian SME owners: you don’t need to be a tech company to have a healthy relationship with AI. You just need to watch the meter, find your power users, and make sure the output makes sense. It’s not about control; it’s about keeping the edge that AI gives you.
Ready to Streamline Your Operations?
Your business should run itself. AutoRunBiz deploys AI agents to automate your daily operations — WhatsApp orders, invoicing, customer follow-ups, and accounting. Book a free 15-min ops audit to see where automation fits your business →