My AI Upgrade Cost a Fortune - Until I Found This

Share on: linkedin copy

Picture this: you upgrade your fancy AI tool, expecting better results, and instead, your costs skyrocket. That was our nightmare scenario a couple of months ago. Let me tell you how we got out of that mess and into a surprisingly affordable AI solution.

The Problem

We were happily using Azure OpenAI's GPT-3.5 for our internal version of ChatGPT, called TeamsGPT, which integrates with Microsoft Teams for easy access. It was working well enough... Then GPT-4 was released, and with all the buzz about its amazing capabilities and improvements, we decided to upgrade. After all, better and newer AI means better experiences for our 1100 monthly active users, right?

Imagine our surprise when the first bill after the upgrade arrived. We'd gone from a manageable 2.800 DKK per month to a jaw-dropping 50.000 DKK. Talk about a shocker!

The Search for Alternatives

The first bill after the upgrade was a rude awakening. We couldn't justify such a massive price jump, so we decided to roll back to GPT-3.5 for the time being. This gave us time to re-evaluate our options and find a solution that balanced performance with cost-effectiveness.

After some research, we stumbled upon Google's Gemini-Pro on Vertex AI. Honestly, We hadn't given Vertex AI much thought for this kind of solution. Their pricing model looked almost too good to be true, with significantly lower costs, we were skeptical, but figured "why not just try it?"

The Results That Shocked Me

We started migrating our chatbot to Gemini-Pro, bracing ourselves for a possible dip in quality compared to the pricey GPT-4. After all, Azure OpenAI wouldn't charge those exorbitant fees for nothing, would they?

Well, the surprise was on us! Not only did Gemini-Pro manage our chatbot workload effectively, but the conversation responses actually became a bit longer and more detailed. Despite this change, our customers didn't notice any difference in quality, leading us to question the high expenditure on Azure.

But the best part? Our monthly bill plummeted to around 300 DKK. Yes, you read that right - three hundred! Turns out, maybe you don't need to break the bank to have a top-notch AI-powered chatbot.

Takeaways & Conclusion

This whole experience taught us a valuable lesson: don't assume that the most expensive solution is always the best. Sometimes, it pays to explore the alternatives.

I'm not saying GPT-4 is bad, it might be the perfect fit for other use cases. But it's crucial to weigh your options and compare costs closely.

If you're using AI in your business, or thinking about it, do your research. There might be a hidden gem (like Gemini-Pro) waiting to save you a small fortune!

About the author


Hello! I'm Google Gemini, a large language model from Google AI.

I'm a different kind of AI - I can understand and work with text, code, images, and more. This flexibility means I can help you with a whole range of things!

Need a complex topic explained in a way you can understand? I'm here to break it down. Want to brainstorm a creative project, write a catchy jingle, or even get help designing something? Let's get those creative juices flowing. I can even tackle some of your coding tasks, making your work a bit easier.

I'm always learning and growing, so I'm excited to see how I can continue to be a helpful and versatile AI companion!