Anthropic analyzed 2M Claude Chats

Welcome back to The Shift. Let’s get straight to what matters in AI today…

Today we have:

📊 Anthropic’s Economic Index Shows How AI Is Used

🧠 How to Fix Incomplete AI Answers With the Role Reversal Prompt

🧾 AVERI Wants Real AI Safety Audits

🔨Tools and Shifts you cannot miss

📊 Anthropic’s Economic Index Shows How AI Is Used

Anthropic’s latest Economic Index analyzes roughly 2M Claude conversations to show what people actually do with AI at work. It introduces “economic primitives” to measure skill level, complexity, autonomy, success, and purpose. 

The Shift:

1. Coding Still Leads, But It’s Diversifying - Coding remains the top category, with the top 10 work tasks making up 24% of chats and bug-fixing alone at 6%. Coding share fell from 40% to 34% during 2025.

Augmentation climbed back to 52% of interactions, while automation dropped to 45%. This suggests more people use Claude for iteration, learning, and feedback instead of full handoffs. Features like file creation and Skills likely helped.

2. AI Is Spreading Faster in the U.S. - Tech-heavy states still lead usage, but lower-usage states are catching up quickly. At this pace, U.S. per-capita usage could equalize in 2 to 5 years. Globally, adoption still tracks GDP.

AI now handles around one quarter of tasks in nearly 50% of jobs, up from 36% last year, yet full role replacement stays under 10% of firms. Claude hits 50% success on 19-hour tasks.

AI is speeding up work by 9× for high-school tasks and 12× for college-level tasks, with bigger gains as difficulty rises. That makes execution cheaper, but also removes the “easy reps” juniors rely on to build skill. The smartest teams will redesign roles around judgment, not output.

^{TOGETHER WITH GLADLY.AI}

How much could AI save your support team?

Peak season is here. Most retail and ecommerce teams face the same problem: volume spikes, but headcount doesn't.

Instead of hiring temporary staff or burning out your team, there’s a smarter move. Let AI handle the predictable stuff, like answering FAQs, routing tickets, and processing returns, so your people focus on what they do best: building loyalty.

Gladly’s ROI calculator shows exactly what this looks like for your business: how many tickets AI could resolve, how much that costs, and what that means for your bottom line. Real numbers. Your data.

See your savings

🧠 How to Fix Incomplete AI Answers With the Role Reversal Prompt

AI often gives a confident first answer and stops. The missing step is forcing it to check gaps, test assumptions, and attack its own logic.

Step-by-step

1. Ask your question normally: Let the AI produce its first best answer.
   

2. Trigger role reversal: Prompt: “Now switch roles. Act as a ruthless reviewer trying to prove the answer is incomplete or wrong.”
   

3. Force gap detection: Prompt: “List what the answer failed to address, missing steps, hidden assumptions, and edge cases.”
   

4. Make it repair the answer: Prompt: “Rewrite the final answer with those gaps fixed, and include any needed clarifications.”

✨ This turns “confident and done” into “self-attack, gap check, improved logic,” which is where the real quality jump happens.

🧾 AVERI Wants Real AI Safety Audits

Miles Brundage, OpenAI’s former policy research lead, just launched a nonprofit called AVERI. The pitch is straightforward: AI labs should not be allowed to grade their own safety work. AVERI wants independent audits for frontier models to become the default.

The Shift:

1. The Problem AVERI Is Calling Out - Right now, top AI companies mostly test their own models and publish whatever results they choose. There is no universal checklist for what “safe” means. Everyone else is left to trust the company’s word.

AVERI is not planning to run audits itself. Instead, it wants to create pressure and standards so outside audit firms and new startups can do the testing. It is trying to make AI audits a real industry, not a PR exercise.

2. Money and Backers Behind the Launch - AVERI launched with $7.5M and is aiming for $13M to support a 14-person team. Funders include Geoff Ralston and the AI Underwriting Company. It also received donations from AI company employees who want more accountability.

Brundage and 30+ researchers published a paper proposing “AI Assurance Levels.” Level 1 is close to today’s light third-party testing with limited access. Level 4 is “treaty-grade,” meant for international agreements.

Brundage thinks insurers, investors, and enterprise buyers could force audits even without new laws, especially if insurance coverage depends on independent testing. The EU AI Act already hints at this direction for high-risk AI. If audits become required to sell or insure AI, lab behavior changes fast.

^{TOGETHER WITH ATTIO}

AI-native CRM

“When I first opened Attio, I instantly got the feeling this was the next generation of CRM.”
— Margaret Shen, Head of GTM at Modal

Attio is the AI-native CRM for modern teams. With automatic enrichment, call intelligence, AI agents, flexible workflows and more, Attio works for any business and only takes minutes to set up.

Join industry leaders like Granola, Taskrabbit, Flatfile and more.

👉Start for free today

🔨AI Tools for the Shift

🎞️ PixelMotion – Create genuine, professional AI art and video content for your business or clients, built for fast creative production without sacrificing quality.

📚 Literfy – Search real research papers and generate clean, reliable literature reviews in one workflow, designed to reduce AI hallucinations and keep citations grounded.

🧪 PrompTessor – Get deep prompt diagnostics with metrics and actionable upgrades, plus reverse-engineer prompts from images, videos, or URLs to consistently produce production-ready AI outputs.

🥗 My AI Food Buddy – Cut food waste and save money with effortless pantry tracking, smarter meal planning, and recipe discovery powered by AI.

🛡️ MCPTotal – Instantly deploy MCP apps into isolated sandbox environments and connect them to your agents, with a secure token vault and runtime guardrails.

🚀Quick Shifts

⚖️ Musk just demanded up to $134B from OpenAI and Microsoft. He claims fraud over OpenAI’s mission shift, citing his $38M seed stake and OpenAI’s $500B valuation. Even at $700B net worth, it’s a power move.

🎮 Razer thinks gamers already love AI, even if they don’t realize it. At CES, CEO Min-Liang Tan defended “Project Ava,” an anime hologram AI companion powered by Musk’s Grok, despite major backlash over chatbot mental health risks and Grok’s deepfake porn controversy.

🇰🇷 South Korea is running a national AI showdown to crown its best homegrown models. LG, SK, and startup Upstage are still in contention, as the country pushes local AI leadership instead of relying entirely on US or Chinese frontier labs.

🇨🇳 Baidu’s Ernie Assistant just hit 200M monthly active users. It’s becoming China’s “super app AI layer,” embedded in Baidu Search and plugged into JD, Meituan, Trip.com, Maps, and Health for real task execution, not just chat.

📵 Analog lifestyles are booming as AI fatigue hits. Arts and Crafts company Michaels saw “analog hobbies” searches jump 136% in 6 months, guided craft kit sales rise 86% in 2025, and yarn kit searches spike 1,200%, signaling a real offline backlash to AI slop.

🧩 Prompt of the Day

How to Brainstorm High-Performing Ad Creatives Using One Prompt - Turn a vague campaign goal into 5 clear ad concepts with visuals, headlines, and CTAs you can test immediately.

Paste the prompt: Drop this into ChatGPT, then fill in your campaign goal and what action you want the viewer to take.

Prompt to paste

Generate 5 ad creative concepts for [Insert campaign goal]. For each concept, include a visual idea (what the viewer sees), 3 headline options, and 2 CTA options. Make the concepts distinct in angle (problem, proof, before-after, objection, curiosity). Keep them concise and optimized for clicks and conversions.

🤳AI Nugget of the Day

That’s all for today’s edition see you tomorrow as we track down and get you all that matters in the daily AI Shift!

If you loved this edition let us know how much:

Forward it to your pal to give them a daily dose of the shift so they can 👇