Google quietly rewrites the enterprise AI playbook

Pivot 5: 5 stories. 5 minutes a day. 5 days a week.

1. Google quietly rewrites the enterprise AI playbook

• Google has expanded its Gemini AI into Workspace and Cloud while landing deals with major enterprises like Bayer, HCA Healthcare, and General Motors. Its strategy emphasizes integration with existing business tools and sector-specific models.

• While competitors focus on flashy releases, Google is embedding AI where day-to-day business actually happens. It’s winning not on noise, but on depth.

• The real advantage isn’t the model—it’s being inside the tools employees already use every day.

Read the full story here

2. Perplexity becomes default on Samsung and Motorola phones

• Perplexity AI will now come pre-installed on Samsung and Motorola smartphones, appearing directly on the home screen. This partnership turns Perplexity into a baked-in behavior, not just another app.

• Search is shifting away from typing queries into a browser—it’s becoming conversational, contextual, and mobile-first. The competition for the homepage is now the competition for the customer.

• This move puts Perplexity in a position Google once held—being the default means winning by inertia.

Read the full story here

3. Salesforce’s AI now writes 20% of its code—and developers are embracing it

• Salesforce says its Einstein Copilot now produces a fifth of the company’s code. Developers aren’t resisting—many are building their own copilots to boost team productivity.

• This isn’t about replacement—it’s about freeing up engineers to focus on harder, higher-value problems. AI isn’t disrupting the workflow; it’s enhancing it.

• The story here isn’t just AI writing code—it’s developers taking the lead on how AI gets used.

Read the full story here

4. New technique lets AI answer sensitive questions without tripping filters

• A new training method called DPO-Chat, developed by DeepSeek and Peking University, helps AI answer sensitive prompts without triggering safety refusals. It retains factual quality while bypassing standard ethical filters.

• This opens new use cases in fields like healthcare and law, where AI must respond to complex or regulated questions. It also reveals how current safety systems can be worked around with training tweaks.

• What counts as “unsafe” is becoming less about model limits and more about how the model is trained.

Read the full story here

5. OpenAI begins international buildout for Stargate AI infrastructure

• OpenAI is planning to build Stargate data centers internationally, with negotiations underway across Europe, the Middle East, and Asia-Pacific. This signals a global scale-up in AI infrastructure far beyond U.S. borders.

• Regional hosting enables better latency, local compliance, and access for global enterprises. AI infrastructure is becoming a national asset, not just a corporate one.

• The next phase of AI competition won’t be about models—it’ll be about who controls the compute.

Read the full story here