How we choose the AI models

There is a lot of noise about which AI model is best. For running a business, the newest and most expensive model is usually the wrong choice. What matters is getting a specific job done accurately, at a price that still makes sense when you are doing it thousands of times.

So we match the model to the job. Most of the work an agent does falls into a few kinds, and each one has a model that does it well without costing a fortune.

We match the model to the task

• Sorting and classifying, like deciding which messages need a reply. This is quick work, and a cheap, fast model does it accurately.
• Writing, like drafting a customer message or a page. This needs a model with a good ear for language, which sits in the middle on price.
• Researching and checking facts, like making sure a claim is right before it goes out. This is worth a stronger model, used only for the part that needs it.

A single job often uses two or three of these, so the agent uses the right model for each part instead of paying a top price for the whole thing.

Why we use providers like OpenRouter

We usually connect your agent through a service like OpenRouter, which gives access to models from many companies through one account. That lets the agent pick the best model for each step of a task, and if one provider has a problem, it can fall back to another without the work stopping.

It also means you are not tied to a single AI company. As better or cheaper models come out, we can point the agent at them without rebuilding it.

You pay for the AI directly

The account the models run on is yours, so you pay for what the agent uses at cost, with no markup from us. Because we choose efficient models, that usually comes to tens of dollars a month rather than hundreds. There is more on our pricing.

Book a consultation if you want to talk through what your job would need.