Salesforce’s chief data officer says customers can run a frontier, open or custom model for each AI task, including Koa, the company’s own CRM reasoning model.
What happened
In an October 13 Salesforce Newsroom interview, Chief Data Officer Michael Andrew described how Salesforce lets customers choose the model behind each use case. The newsroom piece says some tasks, such as routing a case or updating an opportunity, need precise multistep reasoning, while others need to run quickly or cheaply. Salesforce says customers can use a frontier, open or custom model like Koa.
Salesforce describes Koa as its first CRM reasoning model, built with NVIDIA, running inside the Salesforce trust boundary and trained in simulations drawn from 27 years of Salesforce experience running CRM workflows. The company says Koa is orchestrated alongside frontier models rather than in place of them. Andrew said, “We believe in giving our customers choice, that you can pick the model that’s right for your particular use case,” and added that customers should be able to change models over time based on speed of resolution and cost.
Andrew also described multiplayer AI in Slack, where people and agents work in the open on shared context instead of one person talking to one agent in a private tab.
Why it matters
Model choice moves a decision from the vendor to the buyer. We looked at the model itself in our report on CRM software training its own AI models, and at the evidence Salesforce offered for it in an opinion column on its benchmark. This interview adds the deployment side: Salesforce is describing a stack where several models run at once and the customer picks among them.
In our reading, a sales leader buying on that model should ask for two things in writing: which model handles which step, and who can change the assignment after launch. Speed and cost are the two criteria Andrew named, and both show up on the invoice and in the rep’s wait time.
Source: Salesforce Newsroom
