Welcome to this week's edition of Overclocked!

In this issue, Salesforce CEO claims that AI is already doing 30-50% of internal tasks at the company. Plus, Meta successfully poaches AI researchers from OpenAI. Let’s dive in ⬇️

In today’s newsletter ↓
🧮 Salesforce claims AI handles half its workload
🚀 Meta hires three top minds from OpenAI
🛠️ Google rolls out free Gemini CLI
⚖️ Meta beats authors in copyright case
💡 Weekly Challenge: Work faster with AI

🧮 Inside the AI Takeover at Salesforce

Salesforce CEO Marc Benioff recently stunned investors by announcing that artificial intelligence now performs between 30-50% of the company’s internal tasks. Speaking with CNBC, Benioff said generative models streamline “everything from code generation to marketing copy,” trimming project timelines and freeing staff for higher-value work.

Check out the full interview below:

🧰 What Counts as “Half the Work”?

Benioff outlined three AI-heavy arenas:

• Auto-generated sales emails: Einstein GPT drafts outreach in seconds, cutting reps’ prep time by 70%.
  

• No-code dev: Low-tier admins prompt Flow AI to build automations once handled by certified developers.
  

• Self-healing software: Predictive AIOps detects and patches minor platform bugs before humans notice.
  

Fox Business confirmed that Salesforce’s internal metrics log ~14 million monthly AI-generated actions—everything from slide titles to Slack summaries.

📈 The Productivity Math

Salesforce’s 23-percent operating-margin guidance for FY 2026 hinges on these gains. Benioff argued that AI helps “grow revenue per employee” without mass layoffs. Yet Fast Company notes that Salesforce trimmed around 10 percent of staff in 2024, suggesting efficiency already has a human cost.

🛡️ Trust and Hallucination

To reduce errors, Salesforce uses RAG pipelines that pull from vetted CRM records, not the open web. Sensitive drafts route through human approval queues. Still, legal teams worry about AI summarizing contracts—one mis-parsed clause could void a deal.

🌐 Industry Echo

Microsoft claims 30 percent of its code is AI-written, and SAP just demoed an assistant that configures ERP fields. The pattern is clear: enterprise software vendors see generative AI less as a product and more as the new operating layer.

Takeaway: If Salesforce can truly offload half its workload without sacrificing quality, Wall Street will cheer—and every rival will race to replicate the playbook.

🚀 Meta Lures Three OpenAI Researchers 

Meta’s AI division just scored a high-profile talent heist, hiring Tim Rocktäschel, Jackie Cheung, and Susan Zhang—all senior researchers on OpenAI’s reasoning team. Reuters reports Meta dangled compensation packages worth $15 million to $25 million apiece, including stock units that vest in two years.

💰 Anatomy of a Poach

OpenAI CEO Sam Altman told CNBC last week that Meta has floated bonuses “north of $100 million” to select staff. While those numbers shocked Silicon Valley, insiders say the trio left primarily for greater open-source freedom and direct access to Meta’s 350 million-GPU training cluster.

🧠 Why Meta Needed Them

Rocktäschel pioneered reinforcement-learning agents at DeepMind; Cheung authored key papers on multilingual LLMs; Zhang co-designed scaling laws for GPT-3. Meta’s Llama roadmap now prioritizes tool-use agents and multilingual robustness—exactly these researchers’ specialties.

🔥 The Talent Arms Race

The Decoder notes this isn’t Meta’s first raid: last year it hired vision architect Tal Hassner from Apple and ethics lead Aleksandra Przygonska from Anthropic. Meanwhile, OpenAI lured ex-DeepMind star Geoffrey Irving in April. With limited senior talent worldwide, six-figure sign-on checks are becoming eight-figure.

🌎 Broader Context

• McKinsey predicts global AI demand could outstrip qualified researchers by 1 million roles in 2026.
  

• VC funds now pre-pay bonuses for portfolio startups to snag ex-Big-Tech scientists.
  

• Regulators eye non-compete clauses—California may tighten rules after employees claimed “golden handcuffs” limit mobility.
  

Bottom line: In AI, code matters—but people matter more. Meta’s latest hires signal that the talent bidding war is just heating up.

The Weekly Scoop 🍦

🔧 Google offers free Gemini CLI for local model testing

🏫 ChatGPT-style tutor approved for limited classroom pilots 

⚖️ US judge sides with Meta & Anthropic in authors’ AI copyright case

🖼️ Getty Images v. Stability AI begin landmark copyright trial

💼 Harvard survey: only 20% of HR leaders shape AI strategy

🎞️ Google’s AI video tool stirs misinformation worries

⏱️ Weekly Challenge: Work Faster With AI

By now, most of us have integrated AI into just about every corner of our lives. But are we actually saving time, improving our work, and using it to our benefit? This week, we work on using AI to help us get ahead.

📅 Monday – Auto-note meetings

Record your next Zoom or Teams call, paste the transcript into Claude Sonnet (or NotebookLM), and ask for bullet-point action items.

✉️ Tuesday – Rewrite routine emails

Drop your three most-sent templates into ChatGPT 4o and request shorter, friendlier versions with clear CTAs.

📝 Wednesday – Craft performance feedback 

Feed raw notes about a colleague into Gemini 2.5 Pro and ask for a balanced, motivational review.

🔍 Thursday – Sprint through research

Pose a complex “How do we …?” query in Perplexity and ask for a three-step plan with sources.

📊 Friday – Auto-build a weekly recap

Tell NotebookLM to scan the week’s docs and list unfinished tasks, key wins, and looming risks.

🔐 Security tip: Strip out confidential data unless you’re using a company-approved private model. Track your saved minutes—hit 60 by week’s end and make your favorite hack a habit!

From half-automated enterprises to eye-watering hiring wars, AI keeps rewriting the playbook. See you next Friday—stay informed, stay curious, stay Overclocked! ⚡

Zoe from Overclocked