Salesforce's Koa Decisions model routes requests and flags slipping deals fast
Summary
On November 6, Salesforce introduced Koa Decisions, the next version of its HyperClassifier routing model inside Agentforce. Built on Google's Gemma, it picks from categories a team defines in a single step. Salesforce measured routing at roughly 26 milliseconds versus about 1.4 seconds for a general-purpose LLM. It plans to extend the model into flows, Apex, prompt templates and an API, and cites flagging slipping deals as a use.
TopCloserR Perspective
The takeaway
A lot of what an AI agent does in a sales org is make small calls: which team gets this, is this deal at risk. Salesforce is betting that a fast, narrow model can make those calls, so the right person hears about a slipping deal before they open the record.
Who gets the most from this
- In the role
- Salesforce says routing and deal-risk alerts could move from custom models that take months to build to categories a team adds itself.
- Aspiring
- Sales analysts and admins moving toward RevOps. — Knowing when a fast decision model fits and when a full LLM is needed is a useful skill to build now.
- In the role
- Deal-slip alerts that also name the team to step in could change how fast managers hear about at-risk deals.
- Aspiring
- Senior reps stepping into first-line management. — Software will make more of the 'who should act' calls. Your value is what you do once the alert lands.
Put it to work
- Try the Koa Decisions demo on TrailheadRevenue Operations Manager
Sign up for a free Trailhead account and run the demo to see how category-based routing behaves before a vendor pitches it to you.
- List the yes-or-no calls in your sales processRevenue Operations Manager
Write down the routing, lead assignment and deal-risk decisions that wait on a person today, then mark which ones a decision model could make.
- Define what a slipping deal looks likeSales Manager
Agree with your team on the signals and on who steps in, so any automated flag points to a real next step.
Recommendations
More AI in Sales news