Integrating AI While Protecting Your Data AND Controlling Cost

Avantalytics ensures your data stays private and costs remain under your control with our AI integration.

The largest concerns for customers and Avantalytics alike concerning AI revolve around the privacy of customer data and the controlling of costs when using the major LLM providers.  When possible we utilize local models to ensure that data never leaves our premises for the analysis, but due to customer curiosity we've rolled out integrations with Anthropic's Claude LLM models (and Grok for audio).

But as we integrate the analysis capabilities into our service offerings, we insisted that each customer's account interact with their own corporate AI account.  What does this mean?  Simply put - we don't make requests to Anthropic under a shared key model.  The reason we don't do this is because while the data would be secure within our corporate account with Anthropic, according to their data privacy policy, it wouldn't give us good confidence that our customers data would not or could not ever become co-mingled by the LLM.  

So, we instituted a "bring your own key" policy where each must provide an Anthropic key and we will make requests on their behalf to ensure that each customer's data is completely separated.  This is sufficient to allow data privacy, but doesn't move to solve the other biggest concern of our customers -- COST.

Its alarmingly easy to over-use or misuse LLM resources in a way that can create a nasty bill if your not careful.  In a data analytics platform where customer's often default to pulling the most granular data they can in hopes of rolling up an analysis, this type of "feed it all to Claude" mentality is not only not the right way to utilize LLMs, but also very expensive from a token usage standpoint.

Understanding this we absolutely design integration pathways to try to limit data-blasting at Claude when the results wouldn't be improved by doing so.  A conservation approach is prudent, but it left the "shared key" across Avantalytics platform users in the same account problem intact -- how can we ensure one user doesn't blow through tokens and drive up a big bill for their company?  

The answer was to institute AI cost management and controls within the platform.  So we did this.  Now customer's can set daily / monthly limits on each individual user, they can limit which users have access to the AI tools, and the absence of setting limits defaults to small sensible limits anyway.  We also provide transparency at an incredibly granular level to allow you to inspect who did what, and what the cost of each interaction was for proper auditing.

And lastly, we included user feedback capability to determine if the results our customers get are helpful or not - because the worst of all worlds is one where you're spending money on AI and getting nothing useful in return.

If you'd like to see how all this works in practice, just reach out to us - we'd be happy to show you it all, get you set up so you can explore with us how we can redefine AI like we redefine retail analytics.

See all your retail accounts in one place

A 30-minute walkthrough built around the retailers you sell to.

Book a demo