In case you didn’t notice, it’s getting more expensive to ride the AI bandwagon. Web professionals are using the technology to do more than write code. We’re also employing it for just about every other task you can think of – even the most mundane ones. Tools like OpenClaw are taking our usage to new heights, literally. I mean, did you look at your latest bill?
Seriously, some of us are spending something close to a BMW’s (or perhaps a Ferrari’s) monthly payment on AI. What started as a small monthly fee to use ChatGPT, Claude, or Copilot has turned into something completely different. We have become the proverbial kids in the candy store.
It’s a bad case of consumerism fueled by our desire to experiment with new toys. It doesn’t help that AI companies are making it easy to spend significant money without a second thought. Integrating with your favorite SaaS or even WordPress is easy. Yet, the virtual cash register silently rings with each prompt and automation we create.
Perhaps all of this is happening too fast (maybe that’s the intention). We may be falling into an AI-powered trap that will be difficult to escape.
The Grumpy Designer likes to keep things rational. So, let’s hit the brakes, take a step back, and analyze what we’re doing. We’re the only ones who can stop the insanity, after all.
Your Monthly AI Plan Isn’t All-Encompassing
The early days of AI (a whole 2-3 years ago) were exciting. I had fun playing around with ChatGPT and seeing what silly or productive things I could accomplish.
Its coding abilities may pale compared to what we have today. But it was still incredibly handy for troubleshooting and building basic WordPress plugins. I decided that their $20-per-month Plus plan was a relative bargain. It offered access to the latest models and improved performance.
And it’s still a good deal, if you use it in your browser or their app. An individual performing occasional coding, image generation, or file analysis will get a lot for their money. It’s once you step outside of those boundaries that things become costly and confusing.
Such monthly plans typically don’t provide you with API access. For that, you’ll need to purchase tokens. It’s a common practice across the big AI companies.
I won’t argue with the premise of token-based usage. AI requires a lot of computing power to do its job. It must provide quick and accurate answers to user queries. Then there is the cost associated with training and improving each model.
The real problem lies in educating consumers on the differences. It’s not until you need an API key that you realize your monthly plan won’t cover you.
How Much Will That Task Cost?
Generating an API key and connecting your favorite AI model to a third-party service is simple. Figuring out how many tokens you’ll need isn’t so clear. That’s the downside of not having a good/better/best pricing model.
Reputable AI companies let you set spending caps and disable auto-purchasing once you hit your limit. This will save you from a massive bill. It’s also a good security measure, should your API key fall into the wrong hands.
However, experimentation is the only way to understand the costs of using AI tokens. Figure out what you want to do, ask AI to do it, and check your account dashboard to see how much you spent. The trick is to extrapolate that cost based on how many times you’ll perform a given task.
The other challenge is that some tasks will use more tokens than others. For example, generating a headline for your blog post will likely cost less than generating a featured image. Thus, you’ll want to watch your token usage carefully.
Setting a small spending limit may be fine for light-duty tasks. That strategy won’t work for organizations with multiple users who are highly dependent on AI, though. You’ll quickly run out of tokens and will have to go without the service or set a higher limit.
It appears that companies are beginning to realize that willy-nilly usage of AI is costly. Some will narrow the scope of what these tools do for them. It’s also an argument for developing local AI models. Maybe we still need humans after all?
Put AI in Its Place
Sure, AI can do just about everything. But the decision on when and how to use it is ours to make. The increasing and uncertain token costs are reason enough to scrutinize every task. Some of what we’re doing may not be worth the price.
There are other benefits to being less reliant on AI. Fewer tokens spent means less electricity (and less pollution). The environmental impact is real and only getting worse as more of us employ these models.
It’s also an opportunity to stay connected to our humanity. Imagine all the knowledge we might lose if we cede every task to machines. There is also the potential to lose personal and professional relationships. Do you want to trade that for efficiency?
All of this makes me want to treat tokens as something to spend wisely. From now on, I’ll ask myself: Is this task tokenworthy?
Top
