Key Takeaways From AI Interactions

HOW we work with AI matters. Emerging modes of interaction are reshaping roles. Most people are stuck at method 1 or 2. Here are 4 key AI interaction types—and when to use each: 1️⃣ AI as a Microtasker One-shot problem solver. Ideal for quick, contained tasks: rewriting a sentence, generating a one-off image, answering a data question, or fixing a bit of code. High precision, low overhead. 2️⃣ AI as a Copilot Persistent, live support for extended tasks. It stays with you in pairing mode—watching your screen, listening, coding, brainstorming. A back-and-forth partner for creative or technical work in real time. Human in the loop, always. 3️⃣ AI as a Delegate Assign it a goal and let it work autonomously, for minutes or days. Great for complex, long-form tasks like research—no human in the loop. It self-directs, self-checks, and reports back after/while completing tasks. Think: Manus AI, autonomous agents. 4️⃣ AI as a Teammate A presence across your team or org. It joins meetings, takes notes, surfaces insights, runs simulations, offers opinions. Can even be in a manager role. Not just assisting YOU but enhancing the collective. An ambient, participatory AI system. And roles 3 and 4 mean the AI can work in a completely different way than our human systems. Knowing which role to use—and when—is the new AI literacy.

One of the single most important issues is the impact of AI on human thought. This extensive and very interesting paper dives deep. I fully agree with its thesis that “Ultimately, harmonious coexistence with AIs will depend on revaluing cognitive diversity, designing interfaces that foster reflection, and making AI an augmentative partner of human thought, not its replacement.” Some key insights: ⚠️ Cognitive shortcuts weaken reasoning. Heavy reliance on AI showed a strong negative correlation with critical thinking, with cognitive offloading as the key driver. 🌍 Standardization narrows cultural and cognitive horizons. Generative systems trained on Anglo-American corpora nudged writers worldwide toward Western norms, reducing local nuance and expression. Algorithmic personalization reinforced echo chambers, creating “closed-circuit thinking” where diversity of perspective is dulled. 🎭 Manipulation risks bypass human reasoning. AI systems can exploit biases, tailor hypernudges, and generate synthetic personas—shaping decisions without awareness or consent. 🛡️ Safeguards must protect autonomy. The paper highlights transparency through internal logs, bans on subliminal techniques, neurorights for cognitive privacy, and “cognitive hygiene” education. These measures aim to secure epistemic plurality before opacity and automation erode mental sovereignty. 🚀 Design AI as a copilot, not a pilot. Positive potential emerges when AI is built to extend human cognition rather than replace it. Keeping humans “in the loop” ensures that AI serves as an augmentation tool instead of a substitute for thought. 🧑🏫 Pedagogy keeps humans thinking. Thoughtful integration in education—where AI outputs are paired with active reasoning exercises—preserves critical faculties. Training users to engage, verify, and question helps prevent erosion of independent judgment. 🤝 Interfaces should invite reflection. Instead of providing instant answers, AI can be designed to pose questions back to the user, prompting active engagement. This preserves cognitive effort while still supporting exploration and discovery. 🌱 Flourishing requires cognitive diversity. A healthy AI–human partnership means valuing diverse perspectives, fostering reflection, and designing systems that amplify—not homogenize—human creativity and judgment. ⚖️ Human–AI balance redefines collaboration. Individuals using AI performed at the same level as human-only teams, but AI-enabled teams dramatically outperformed both—showing that the deepest gains come from synergy, not substitution. 🌟 Augmentation as the true measure of success. The future of AI will not be decided by raw efficiency but by whether it strengthens or weakens human autonomy. Systems that expand reasoning, preserve diversity, and nurture reflection will be the ones that truly advance human flourishing.

Sam Altman has been on a podcast blitz this week. 3 appearances in 5 days, each one a post-Dev Day sermon about the future of intelligence. I went through them all (fine, I read the transcripts) partly out of curiosity, partly out of professional obligation. When the person architecting the next platform shift narrates his thought process in public, you pay attention. Takeaways: ▪️The Verticalization of Intelligence → “I was always against vertical integration, and now I think I was wrong about that.” OpenAI’s biggest pivot since its founding: the lab is now an empire - building chips, models, and end-user interfaces in one continuous loop. In the intelligence economy, whoever controls compute and energy controls cognition. ▪️ Strategy as Evolution →“Let tactics become a strategy.” OpenAI’s R&D is Darwinian. Ship chaos, observe order, scale the mutation. Memory wasn’t conceived as a moat - users made it one. Altman’s genius isn’t foresight; it’s feedback. ▪️AI Scientists →“For the first time with GPT-5, we’re seeing little examples where models are doing science, making discoveries.” Altman’s AGI test is novel scientific discovery. Within two years, he predicts AIs will generate publishable research - and soon after, it’ll feel routine. Civilization’s next compounding force: automated invention. ▪️ Customization Is the New UX →“It would be unusual to think you can make something that would talk to billions of people and everybody wants to talk to the same person.” ChatGPT’s uniformity was naïve. The future: AIs that adapt tone, personality, and worldview to each user - an identity layer that mirrors your cognitive and emotional style. ▪️Post-Interface Computing →“You talk to your device and it does exactly what you want - then gets out of your way.” Voice is the natural endpoint of human-AI interaction - ambient, context-aware, invisible. The rumored io device is his post-screen bet: a computer that listens, reasons, acts. He is betting on the disappearance of interfaces. ▪️ Distribution Moves Inside the Assistant →“There will be a new distribution mechanic developers figure out… we’ll learn together.” Future startups will live or die by whether ChatGPT mentions them. It’s not SEO anymore; it’s AIO - Assistant Optimization. ▪️ The Democratization of Creation →“In the first few days, ~30% of users were active creators...” Altman sees creativity as universal, just bottlenecked by friction. Sora removes it, turning everyone into a micro-studio. The economics will follow: per-generation pricing for heavy users, rev-share for cameos, maybe ads if it tilts social. Compute is the new canvas: 1M downloads in <5 days, faster than ChatGPT. Altman’s worldview in one loop: Build → Release → Observe → Scale → Moralize Later. He’s a capitalist empiricist, not a philosopher. He summarizes: “AGI will come; it will go whooshing by… the world will not change as much as you’d think in a big-bang sense.”

My biggest takeaways from Fei-Fei Li: 1. Just nine years ago, calling yourself an AI company was considered bad for business. Nobody believed the technology would work back in 2016. By 2017, companies started embracing the term. Today, virtually every company calls itself an AI company. 2. The modern AI revolution started with a simple but overlooked insight from Fei Fei: AI models needed large amounts of labeled data. While researchers focused on sophisticated mathematical models and algorithms, she realized the missing ingredient was data. Her team spent three years working with tens of thousands of people across more than 100 countries to label 15 million images, creating ImageNet. This dataset became the foundation for today’s AI systems. 3. The human brain’s efficiency vastly exceeds current AI systems. Humans operate on about 20 watts of power—less than any lightbulb—yet accomplish tasks that require AI systems to use massive computing resources. Current AI still can’t do things elementary school children find easy. 4. Simply scaling current approaches won’t be enough. While adding more data, computing power, and bigger models will continue advancing AI, fundamental innovations are still needed. Throughout AI history, simpler approaches combined with enormous datasets consistently outperformed sophisticated algorithms with limited data. 5. Breakthrough technologies often start as toys or fun experiments before changing the world. ChatGPT was tweeted by Sam Altman as “Here’s a cool thing we’re playing with” and became the fastest-growing product in history. What seems like play today might transform civilization tomorrow. 6. Spatial intelligence is as crucial as language for real-world applications. In emergency situations like fires or natural disasters, first responders organize rescue efforts through spatial awareness, movement coordination, and understanding physical environments—not primarily through language. This is why world models that understand three-dimensional space represent the next frontier beyond text-based chatbots. 7. Physical robots face much harder challenges than self-driving cars, which took 20 years from prototype to street deployment and still aren’t finished. Self-driving cars are metal boxes moving on flat surfaces, trying not to touch anything. Robots are three-dimensional objects moving in three-dimensional spaces, specifically trying to touch and manipulate things. This makes robotics far harder than creating chatbots. 8. Everyone has a role in AI’s future, regardless of profession. Whether you’re an artist using AI tools to tell unique stories, a farmer participating in community decisions about AI deployment, or a nurse who could benefit from AI assistance in an overworked health-care system, you can and should engage with this technology. AI should augment human dignity and agency, not replace it—which means both using AI as a tool and having a voice in how it’s governed.

The inaugural LinkedIn Skills on the Rise list is out, and here's what's fascinating: AI Literacy isn't just #1— AI can actually help you develop the #2 and #3 skills on the list! (Conflict Mitigation and Adaptability) This synergy creates a powerful opportunity for professional growth in 2025. 1️⃣ AI LITERACY What's the best way to build your—and your team's—AI skills fast? Here are tips from my new book, "How to Think with AI": 1.    Take the leap now: Don't wait for perfect clarity or "when you have time." The opportunity cost of waiting rises every quarter as AI advances. 2.    Try using AI as a thought partner: Instead of basic requests, challenge AI with sophisticated problems—a conflict with a colleague, a market opportunity analysis, or a strategic decision. Higher expectations lead to more valuable results. Keep the dialogue going with follow-up questions and feedback—AI improves through conversation. 3.    Make it a habit: If you are struggling to fit AI into your life, apply the five-minute rule. Start with just five minutes of AI interaction daily, and do that for a month—small enough not to feel like work but consistent enough to build the habit. Every day, ask yourself "how could AI help me today?" to expand your thinking of where AI can deliver value to you. 2️⃣ CONFLICT MITIGATION Try using AI as a "neutral" perspective: AI can serve as an impartial "third party" to evaluate different sides of a conflict. Have it role-play various stakeholders to simulate negotiations before difficult conversations, helping you anticipate objections and prepare responses. This preparation can significantly reduce tension when addressing real conflicts. 3️⃣ ADAPTABILITY AI supercharges your ability to navigate change. For example, try this: Future scenario planning: AI excels at exploring multiple possible futures and their implications. Challenge AI to generate diverse scenarios for upcoming changes—from market shifts to organizational restructuring—and work through potential responses for each. Perspective expansion: AI can help you view situations through different lenses —customers, competitors, regulators, different generations, diverse cultural viewpoints—revealing blind spots in your thinking. ____ 👋 Hi, I'm Alison McCauley, and focus on how to leverage AI to do better at what we humans do best. I'll be sharing more about how to Think with AI to boost your brainpower. Follow me for more, and share your thoughts below! https://lnkd.in/gQgA6sGi

𝐓𝐚𝐥𝐤𝐢𝐧𝐠 𝐭𝐨 𝐀𝐈 𝐌𝐚𝐝𝐞 𝐌𝐞 𝐚 𝐁𝐞𝐭𝐭𝐞𝐫 𝐂𝐨𝐦𝐦𝐮𝐧𝐢𝐜𝐚𝐭𝐨𝐫. 𝐒𝐮𝐫𝐩𝐫𝐢𝐬𝐞𝐝? Like many in the Data & AI space, working extensively with Generative AI has offered unexpected lessons. Beyond the tech itself, I was surprised to find that mastering AI prompting also sharpens crucial human communication skills. It's become a practical, everyday training ground. Here’s how interacting with AI translates directly to better communication with people: 1. 𝐂𝐨𝐧𝐭𝐞𝐱𝐭 𝐢𝐬 𝐂𝐫𝐮𝐜𝐢𝐚𝐥: GenAI needs sufficient background to avoid errors or "hallucinations." Give it too little context, and the output is often off-target. Sound familiar? It's the same when we communicate with colleagues. Providing clear context before delegating or explaining prevents misunderstandings and ensures everyone starts on the same page. Clear context drives effective action. 2. 𝐂𝐥𝐚𝐫𝐢𝐭𝐲 & 𝐂𝐨𝐧𝐜𝐢𝐬𝐞𝐧𝐞𝐬𝐬 𝐌𝐚𝐭𝐭𝐞𝐫: Vague or wordy prompts confuse AI and waste resources (think of it like mental energy – for both AI and humans!). Precise, concrete language gets better results, faster. This directly mirrors human interaction. Clear, concise messages cut through the noise, respect others' time and focus, and ensure the core point lands effectively. Less fluff, more impact. 3. 𝐈𝐭𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐋𝐞𝐚𝐝𝐬 𝐭𝐨 𝐈𝐦𝐩𝐫𝐨𝐯𝐞𝐦𝐞𝐧𝐭: Often, the first AI prompt isn't perfect. We refine it, tweak the wording, and adjust based on the output. This iterative process teaches patience and precision. Similarly, effective human communication isn't always a one-shot effort. It often involves clarifying, rephrasing, and adapting our message based on feedback (spoken or unspoken) to achieve true understanding. Mastering GenAI pushes us to be better communicators – clearer, more contextual, and more adaptable. These aren't just "AI skills"; they are fundamental professional skills, amplified by our interaction with technology. What communication lessons have you learned from working with AI? #AI #GenAI #CommunicationSkills #FutureOfWork

Have you ever imagined what #AIAgents really think? AI Agents don’t “just answer.” They sense, reason, act, and learn in a continuous loop. Here’s how the 4 stages work: 1️⃣ Perception The agent collects inputs from the world — text, images, audio, even IoT sensors. This is multimodal fusion — combining all signals so the AI can see, read, and listen at the same time. 2️⃣ Reasoning Once inputs are captured, the agent doesn’t jump blindly to action. It taps into its knowledge base + memory, weighs options, and applies decision-making logic to choose the best path forward. This is where many a time RAG (Retrieval-Augmented Generation) come in. It supercharges reasoning - pulling the right context at the right time so decisions are grounded in truth, not guesses. 3️⃣ Action Action is the execution layer of an AI Agent — where cognition turns into real outcomes. It’s the moment decisions are translated into tangible results. But execution doesn’t happen in isolation. AI Agents need tools, interfaces, and systems to act — APIs, databases, applications, or physical devices. Without tools, reasoning stays theoretical. With them, actions become measurable impact. Key Points: - Perception understands. - #Reasoning decides. - Action delivers through the tools it’s connected to. 4️⃣ Learning Learning is what makes an AI Agent more than a one-time executor — it’s what makes it adaptive. Every perception, decision, and action creates feedback. The agent doesn’t discard it - it stores it in memory, finds patterns, and updates how it thinks for the next cycle. This continuous feedback loop means: - Mistakes become data points. - Repeated tasks get faster and more accurate. - Context builds over time, so responses feel less generic and more tailored. Without learning, an AI Agent is static. With learning, it evolves — turning short-term execution into long-term intelligence. LEARN FROM PAST: This is the real difference between a #chatbot and a true AI Agent. One just replies. The other perceives, reasons, acts, and learns — creating outcomes, not noise.

This week is the first time that I attended Google Cloud Next as an ex-Googler AND a partner! Earlier this week, we announced an exciting partnership between Google Cloud and Cresta. This makes it significantly easier for GCP customers to adopt Cresta for their contact center AI transformation initiatives. Big thank you to Alexandra D'Amico Williams, Alex Sharp and Darren Mowry for the incredible collaboration in bringing this to life. Yesterday, I had the honor of joining a panel on AI agent deployment, alongside Matt from Glean and Lukas Biewald from Weights & Biases, masterfully moderated by Ritika and meticulously coordinated by Sabrina. Here are a few key takeaways from the discussion (ALL 𝐞𝐫𝐫𝐨𝐫𝐬 𝐚𝐫𝐞 𝐦𝐢𝐧𝐞!): • Use Cases: AI agents today excel at “vaguely right” tasks like content generation and deep research. However, they require careful handling for “precisely wrong” workflows that demand multi-step precision where the cost of error is high. • KPIs: Define clear KPIs. One great example from Matt was using open rates to measure AI SDR email campaigns. In contact centers, measuring success is often more straightforward since outcomes are tied to each interaction. • Data: AI agents are powerful, but they’re not a panacea. Data silos remain a major blocker. Enterprises need to unlock their internal knowledge. For a great perspective, check out Philip Kolterman and his recent post about their investment in building a robust knowledge base to support their AI agent journey. https://lnkd.in/g7YHszqZ • Actions: AI agents need to take real actions, therefore API infrastructure matters. Protocols like MCP and agent-to-agent integrations are moving in that direction as well. Personally, I believe that AI agents need to adapt to operate in environments optimized for humans i.e., seeing, clicking, and typing on GUIs. • Evaluation: LLMs are probabilistic by nature. We discussed how to design deterministic workflows, probabilistic guardrails, LLM judges and test case management systems to achieve precision and predictability in real-world deployments. The most exciting part? This space is evolving incredibly fast. I fully expect all of this to look quite different in 3, 6, or 12 months. We’ll keep sharing what we learn on our journey at Cresta.

Google Cloud Next: Key Insights for AI Devs 🚀 Just wrapped up an inspiring Google Cloud Next, and wanted to share the highlights that I think are particularly relevant for those of us building the future of AI. A major takeaway was the focus on infrastructure built for the next wave of AI. 👉The new TPU v7 "Ironwood" is a beast, offering the power and memory bandwidth needed for the increasingly complex models we're working with. This isn't just about training; it's about having the horsepower to continuously run sophisticated AI. What really stood out to me was Google's strong push into making agent development a reality. This shift is huge for how we'll be building AI going forward. Key elements for developers include: 🟢 Agent2Agent (A2A) Protocol: This shared language will be crucial for building systems where different AI agents can communicate and collaborate effectively across models and tools. 🟢 Vertex AI Agent Builder: This new tool looks incredibly promising for streamlining the process of creating agents with integrated tools, memory, and reasoning capabilities. 🟢 Gemini Code Assist: Having more powerful AI-powered copilots directly integrated into the development workflow will be a game-changer for productivity. It's clear that Vertex AI is evolving into a comprehensive platform designed specifically for building and deploying these intelligent agents – going beyond just model training. We're seeing a move towards thinking in terms of context management, tool orchestration, and understanding the long-term behavior of AI systems. Ultimately, the future of AI development is pointing towards building coordinated, persistent systems that can learn, plan, and interact with their environment in real-time. This means focusing on things like long-term memory, multi-step decision-making, and seamless integration with various tools and other agents. Link to a more detailed overview in the comments Richard Seroter Karl Weinmeister Jeff Dean Thomas Kurian Oriol Vinyals Ivan 🥁 Nardini (Another highlight from the week was @arizeAI being announced in the keynote!)

6 key takeaways from my "AI Search" episode of Ahrefs Podcast with Ryan Law: (...where we discussed insights from over a dozen data studies our team ran this year...) ▪️ 1. Off-page SEO is back (but not backlinks). The strongest correlation we found for appearing in AI Overviews? - Branded web mentions. Our research showed a 0.67 correlation between how often a brand is mentioned across the web and its visibility in AI search. AI models learn what your brand is about based on how others talk about you. Mentions on Reddit, Quora, G2, and industry blogs is what drives your visibility in AI search. ▪️ 2. AI chatbots don't just "Google" your question. Only 12% of the links cited by ChatGPT, Gemini, and Copilot appear in the top 10 Google results for the same query. That’s because AI chatbots use “fan-out queries” — they break your prompt into multiple, more specific searches. This means you no longer have to mirror the consensus of top-ranking pages all the time. Having a unique point of view can actually get you cited by AI. ▪️ 3. AI prefers fresh content (way more than Google). Our research on 17 million citations showed that ChatGPT, Copilot, and Gemini heavily favor newer pages. Likely because users ask about topics not yet in the training data, forcing AIs to fetch newer sources. ▪️ 4. Watch out for hallucinated URLs. When we checked our own analytics, we found almost 4% of all visits from LLMs went to pages that don't exist. Things like "ahrefs.com/keywords" — a URL that seems logical but isn't real. You should monitor for these hallucinated 404s (we have a one-click filter in Ahrefs' Web Analytics) and redirect them to the nearest alternative. The chances are, you might get some highly relevant traffic out of it. ▪️ 5. Find your "entity gaps". LLMs understand your brand through “co-mentions” — what topics and products you’re mentioned alongside arond the web. If your competitors are frequently mentioned in discussions about “school backpacks” but you’re not, AI assumes you’re irrelevant. Use tools like Ahrefs' Brand Radar to find and fill those missing associations. ▪️ 6. Stop obsessing over "AI-formatted" content. We're already seeing people publish 20,000-word, markdown-formatted, AI-generated pages with no images or links, purely for LLM ingestion. And that can actually get you cited. For now. But if a user clicks through to that spammy, unreadable mess, what have you gained? Good AEO tactics should also be good for users. If a tactic only works for robots, it's probably not a good long-term strategy. ... Check out the full episode: https://lnkd.in/gREaW6Fg All 46 minutes of it are well worth your time. I promise!