Salesforce’s new reasoning model and the question pharma teams should ask next

Salesforce’s new reasoning model and the question pharma teams should ask next
PUBLISHED
November 17, 2026
CATEGORY
Other

On 15 September 2026, Salesforce and NVIDIA announced Koa, Salesforce’s first CRM reasoning model for Agentforce, built on NVIDIA Nemotron and designed to help agents reason through complex, multistep workflows and use the right tools. Salesforce says Koa was post-trained on synthetic scenarios across more than 14 industries, including healthcare, with no customer data, and offers it as a specialised, Salesforce-hosted option for Agentforce use cases. Koa is available to select pilot customers now, with general availability expected in Winter 2026 in U.S. regions.

What Does It Mean for Pharma Teams?

Our view: Koa is a strong signal of where enterprise AI is heading. Models are becoming more capable and more specialized for the work they support. That’s good news for commercial teams. It also puts the spotlight on the part each company brings to the table.

A model that brings reasoning

Salesforce has said no customer data was used to train Koa, and that it performs post-training and inference entirely within its own trust boundary. Salesforce describes this as a commitment to data protection. It also means that Koa works with the knowledge each company provides: your brands, your approved claims, your audiences, and the rules of each market you operate in.

Whatever an agent does for your teams, it does it with the content and knowledge it’s given. The model brings reasoning. You provide the material it reasons over.

Teach the agent before you put it to work

We think of it this way: an agent has to be taught before it’s used. In pharma, the teaching material is content: approved, structured, tagged, current, and localized.

The better prepared that content is, the more value teams can get from whichever model they choose. Modules that aren’t tagged are hard to find. Approval status that sits in a separate system is hard to check. An expired eDetailer that’s still in the library can still be picked up. That’s why we see content readiness as the first step of agent deployment, not a clean-up job for afterwards.

Six signs your content is agent-ready

These are the criteria we use with pharma teams when we assess content for agent use:

  • Metadata and tagging. Every asset carries consistent tags for brand, indication, audience, market, channel, and key message. An agent finds content by how it’s labelled, not by what a brand manager remembers.
  • Modularity. Content is built from approved modules (claims, visuals, references) instead of being locked inside finished PDFs and detail aids, so the parts can be reused within approved rules.
  • MLR status carried in the asset. Approval status, approval code, and any usage conditions travel with the asset itself, so any system that reads it can tell approved content from a draft.
  • Multichannel formats. The same approved message exists in the formats your channels need, such as rep-triggered email, remote detailing, and web, without a separate rebuild for each one.
  • Versioning and expiry. There’s one current version of each asset, with an expiry date, and retired material is actually withdrawn. If an agent can still read outdated content, that’s a compliance risk.
  • Localization. Local variants stay linked to their global source, with market-level approvals recorded, so each market adapts content instead of starting again.
Agentforce Viseven

A better question to ask

The announcement of Koa is a good moment to ask a new question. Instead of only “which model?”, ask “is our content ready for the model we choose?” Models will keep getting better. The part that stays in your hands is whether your content is structured, tagged, and approved in a form an agent can use.

Viseven joined the Salesforce Agentforce 360 for Life Sciences ecosystem as a Content Provider in June 2026. Our work there is helping pharma teams prepare content so agents can use it.

Check Your Readiness

Our CRM and Content Migration Readiness diagnostic takes about 5 minutes and covers 12 questions. It shows where your content foundations are real and where readiness is only assumed.

*Salesforce, Agentforce, NVIDIA, and Nemotron are trademarks of their respective owners, mentioned for identification only. No endorsement is implied.

Models will keep improving. Will your content keep up?

A 5-minute diagnostic shows the gaps between your content library and agent deployment, before an agent finds them.

Assess my readiness