Skip to content

AI Safety Debate Meets Reality At Salesforce’s Dreamforce

Businesses are also becoming more selective about AI spending

Pic via X

The AI industry is debating how quickly frontier models should advance, but many businesses say they are still catching up with technology that already exists, according to CNBC.

At Salesforce’s Dreamforce conference, executives and customers said older, cheaper AI models are sufficient for many sales, customer service and software tasks.

💡
G2 executive Tim Sanders said most AI agents do not require frontier-level capabilities, while Nice technology chief Kevin Lee said even models a generation behind remain highly effective.

Businesses are also becoming more selective about AI spending. Docusign uses both frontier and open-weight models, reserving more powerful systems for complex reasoning tasks.

Companies such as Docusign and Nice are adopting model-routing systems to send different requests to the most cost-effective models.

However, some firms continue to use cutting-edge AI. Databricks made OpenAI’s GPT-6 Astra available to its 3,500 developers for complex tasks.

The contrast comes amid growing concerns over AI safety and calls from industry leaders to slow frontier-model development.

Related Tweet:

Also Read:

Microsoft AI Chief Calls OpenAI’s Latest Safety Finding A ‘Serious Situation’
Suleyman called that episode significant enough to prompt broader discussion among AI leaders.

Comments

Latest