Paying for AI? Make Sure You’re Measuring Spend AND Results
By now the question isn’t if your team is using AI, but where, how much, and how much it might be costing you.
Many businesses, even ones that consider themselves relatively low-tech, have already graduated from basic free AI use (like casual, occasional questions to ChatGPT). Someone, somewhere is already paying for credits or tokens in one ecosystem or another.
Chances are, it’s more than one someone and more than one ecosystem.
We’re not here to suggest this is bad or even avoidable. AI is here, and it makes good sense for many businesses to use it wisely.
But we are suggesting that a DIY or laissez-faire approach to AI spend is a bad strategic move.
Does your business know how much it’s paying for AI, across which platforms and ecosystems? Do you have an effective way to measure that spend? What about comparing that spend against actual business results?
Distinguishing Between “Free,” Bundled, and Paid
Before we get into those questions, it’s worth distinguishing the various models, because you don’t want to treat them all the same.
Free
First is actually-free use. This is when you go to ChatGPT or Google search or Gemini or Grok and ask it a question, plain and simple. Each of these platforms will answer basic questions for free, even if you don’t have any credits. Most work even if you aren’t logged into an account at all, to a point.
This kind of free usage is generally not a concern, with two caveats:
- Stick to well-known, major brands and platforms.
- Don’t feed free models anything sensitive, proprietary, or restricted.
The thing is, “free” has a pretty short runway. Business use cases tend to hit the end of free pretty quick.
Bundled
Bundled AI is the stuff you’re paying for as a part of other tools you can’t live without. Microsoft 365, Google Workspace, and many other paid software and SaaS solutions now build in some level of AI capability. We’ve talked about Copilot a fair amount on this blog — some of Copilot is baked in, without meaningful limits. Other elements are walled off, requiring credits or tokens.
Bundled AI also isn’t a strong spend concern. You may want to inventory who’s using it in what ways to make sure people aren’t getting fooled, thinking they’re saving time when they’re not. But as far as budgeting goes, don’t worry about it.
Paid
True paid services are what happens when you hit the end of free or bundled capability. Other apps lock all their AI tools behind a paywall, and some image and video models don’t have a meaningful free tier at all (generating visual content is still compute-heavy).
This is the category to watch out for. Different teams and team members may be paying for whichever app or platform best suits them, and this can get out of hand in a hurry.
Measuring the True Cost
As you work to understand your AI spend, make sure to understand the true cost. There’s the monthly subscription and the tokens, sure, but you also need to think through:
- Setup and maintenance
- Training around a new tool
- Security
- Quality control (how much time is your copywriter spending fixing the email AI wrote for her?)
That last point is a big one, too: every AI tool you pay for should be delivering real value, not just moving work around. If it takes your copywriter 20 minutes to write an email and 20 minutes to fix an AI-generated one, you’re spending more to do it with AI than without.
So instead of just looking at raw capabilities, look for what problems are getting solved.
Maybe AI is helping your customer service/support team handle 30% more calls an hour. That’s worth a pretty penny!
Maybe teams are accomplishing more thanks to AI-powered meeting summaries: instead of listening through a missed meeting, the summary can get them straight to the part relevant to them.
But maybe AI is just trading one kind of repetitive work for another, without saving time.
Not sure how to measure AI ROI or rein in spend-happy departments? We can help. Reach out to our team anytime.
Comments
Post a Comment