Archive · 16 Sep 2026
Salesforce Unveils Koa and AI Partnerships at Dreamforce
Salesforce held its Dreamforce conference in San Francisco.
It announced a new AI model called Koa for sales, marketing, and customer-support work.
Koa was trained using made-up data rather than real customer data.
Salesforce also introduced AIforce, which lets people use Salesforce information through other AI tools.
An upgraded Claudeforce lets customers use Claude to complete Salesforce tasks.
Salesforce said it is working more closely with Anthropic, Google Cloud, and Amazon Web Services.
Air India will use Salesforce AI agents to handle some refunds, baggage claims, loyalty claims, and ticket changes.
Salesforce says these tools should be secure and responsible.
Some technology leaders are also warning that powerful AI agents need stronger safety measures.
Salesforce introduced Koa, its first CRM reasoning model, built on Nvidia’s open-weight Nemotron family and post-trained with synthetic data.
The company launched AIforce, allowing customers to use Salesforce data, workflows, permissions, and security through external AI interfaces.
An upgraded Claudeforce lets companies use Anthropic’s Claude to perform Salesforce tasks while their data remains in Salesforce’s system of record.
Salesforce announced an Air India partnership to automate refunds, baggage and loyalty claims, eligibility checks, and ticket name changes through Agentforce.
The announcements come as AI agents reshape enterprise software and raise concerns about safety, responsibility, and changing SaaS contracts.
- Who
- Salesforce, its CEO Marc Benioff, Anthropic, Nvidia, Google Cloud, Amazon Web Services, and Air India are involved.
- What
- Salesforce announced Koa, AIforce, an upgraded Claudeforce, expanded technology partnerships, and an Air India collaboration using Agentforce.
- When
- The announcements were made on Tuesday, September 15, 2026.
- Where
- The announcements were made at Dreamforce in San Francisco, California, United States.
- Why
- Salesforce is developing AI-agent products as enterprise software changes and is seeking to keep customers connected to its ecosystem.
AI Expansion
AI Caution
Use of AI agents
AI Expansion
Salesforce says AI agents can take responsibility for tasks, automate enterprise work, and provide intelligent interfaces for customers and employees.
AI Caution
Concerns about increasingly autonomous AI agents have intensified, and Anthropic CEO Dario Amodei has called for an intentional slowdown of frontier AI progress so safety can catch up.
Enterprise software strategy
AI Expansion
Salesforce is expanding access to its data, workflows, and applications through AIforce, Claudeforce, and partnerships with major technology companies.
AI Caution
The rise of AI agents is prompting SaaS companies to reconsider how customers interact with their software and what customers will pay, with reports pointing to revised contracts and smaller software footprints.
AI responsibility
AI Expansion
Salesforce CEO Marc Benioff said AI should handle actions ethically and be held responsible for them, while Salesforce emphasizes security, governance, and zero data retention.
AI Caution
The article notes broader concerns about whether powerful AI systems can be deployed safely and ethically as their capabilities increase.
- Event
- Dreamforce 2026
- Location
- San Francisco, California, United States
- Reasoning model
- Koa, Salesforce’s first CRM reasoning model
- Model foundation
- Nvidia’s open-weight Nemotron family
- External AI access
- AIforce brings Salesforce data and workflows to other AI interfaces
- Air India use cases
- Refunds, baggage and loyalty claims, eligibility checks, and ticket name changes
- Claudeforce availability
- Open beta with more than 37 pre-built sales skills