Artificial intelligence news.
Straight from the original announcements.

Salesforce Lets Customers Choose Their AI Model, Including Its New Koa

October 13, 2026, 17:00By the AI News Bites Editorial TeamSource: Salesforce

Salesforce says customers should be able to choose the AI model that fits each job, whether it is a frontier, open or custom model. The company points to Koa, its first CRM reasoning model, which it says is orchestrated alongside frontier models rather than replacing them.

Key points
  • Salesforce says customers can use frontier, open or custom models, depending on the task.
  • Koa is described as Salesforce's first CRM reasoning model, built with NVIDIA and running inside the Salesforce trust boundary.
  • According to Salesforce, Koa is trained in simulations drawn from 27 years of its CRM workflow experience.
  • Chief Data Officer Michael Andrew says model choice should weigh use case, cost, value and speed of resolution.
  • In Slack, agents work with teams on shared context, with security and permissions, regardless of the underlying model.

Different tasks, different models

According to Salesforce, not every job calls for the same AI model. Tasks such as routing a case or updating an opportunity need precise multistep reasoning, while others may need to be done particularly quickly or cost-effectively. Rather than tying customers to a single tool, the company says it lets them use whichever model works best for a given use case.

Salesforce describes Koa as its first CRM reasoning model, purpose-built for the work enterprises do. It was built in collaboration with NVIDIA and runs inside the Salesforce trust boundary. According to the company, it is trained in simulations drawn from 27 years of Salesforce experience running CRM workflows.

Chief Data Officer Michael Andrew explained the thinking in an interview with the Salesforce Newsroom. "We believe in giving our customers choice, that you can pick the model that's right for your particular use case," he said, adding that customers should have "the flexibility to keep evolving and changing what model you use."

Deterministic software, probabilistic agents

Andrew drew a distinction between two kinds of systems. Traditional software is deterministic: if you press the button to open a case, you get a case every time. Businesses, he said, have always run on such systems, which guarantee the same result everywhere in the world.

AI models are different. Andrew compared them to a new employee who might make mistakes, calling nondeterministic AI systems "basically big artificial brains" that can be trained and guided but are not always reliable.

That is part of his argument for connecting a company's deterministic systems to any model. Models have different strengths and weaknesses and are changing all the time, he said. Options include frontier foundation models from partners such as Anthropic and OpenAI, open-source models, and custom models like the one Salesforce released with NVIDIA. He framed the choice around use case, cost, value, speed of resolution and data security, and described AI as a fuel that the company wants to let come from anywhere, including Anthropic and Claude, OpenAI and ChatGPT, or other partners announced for AIforce.

Agents working in Slack

The interview also covered multiplayer AI in Slack. Salesforce describes a setup in which, instead of one person, one screen and one agent in a private tab, people and agents work in the open on shared context, so the whole team learns from every exchange.

Andrew said about 85,000 employees use Slack at Salesforce to talk with colleagues about projects and accounts. Users can ask an agent for help directly in the chat, where it acts as a digital colleague in the flow of work. He said Slack also brings security and permissions, so the agent is limited to the channels and actions the user allows.

In his words, the agent is "working with your team together, and it has all the shared context." It can know what was discussed about a deal or customer issue, giving it the same context as team members. Salesforce says this holds no matter which model the agent is based on.

What it means for customers

For enterprise customers, the message from Salesforce is that model choice stays open as the market shifts. Companies can match a model to a task based on accuracy, speed or cost, and switch later, while keeping their data secure and their processes running inside Salesforce. The source does not give pricing or availability details for Koa or the AIforce partner announcements.

Related news

OpenAI and Databricks bring governed enterprise data to AI tools
BusinessNovember 13, 2026, 16:30

OpenAI and Databricks bring governed enterprise data to AI tools

OpenAI and Databricks are expanding their partnership to help enterprises use ChatGPT, Codex and GPT models with governed business data. The companies describe three…

Stage 2 Capital closes $100M fundraise to bring AI to startup GTM support
BusinessNovember 11, 2026, 21:47

Stage 2 Capital closes $100M fundraise to bring AI to startup GTM support

Stage 2 Capital has closed $100 million across two funds and is adding an AI agent to help portfolio founders access go-to-market expertise from its network of more than…

Business
BusinessNovember 5, 2026, 08:01

Sinch Q3 sales rise 5% organically as Americas drives growth and AI ties deepen

Swedish cloud communications company Sinch reported third quarter 2026 net sales of SEK 7,085m, an organic increase of 5 percent, with the Americas region cited as the…