Piush Vaish

AI Agents are just the next step in autonomous evolution Here's how the agentic market is currently evolving.... Agentic AI is rapidly evolving, powered by breakthroughs in foundation models like GPT-5, Claude Code, and other cutting-edge agentic releases. To identify the ever-changing progression, Me + the team of researchers and authors, wrote a framework about autonomous progress. Here are the five 5-levels of Agentic AI progression that we are currently looking at: 📌 Level 1: Rule-Based Automation – Think of early chatbots following simple if-then rules to respond to FAQs. - They execute predefined actions with no learning capability, like a basic customer service bot routing queries. - For example, Google Sheets, Microsoft Word, etc. 📌 Level 2: Intelligent Automation – Here, systems use basic AI/ML models to process semi-structured data and make decisions. - For example, an email sorting tool, RPA workflow tools like UI Path and others. 📌 Level 3: Agentic Workflows – This changed the entire autonomous workflow starting from 2024! Systems integrate foundation models with natural language understanding, reasoning, and self-aware tool use. - Imagine a virtual assistant that not only schedules meetings but also learns from feedback to suggest optimal times. - These workflows are now supercharged by state-of-the-art models like GPT-5, Claude Opus 4.1, and Claude Code, enabling unprecedented reasoning and execution. 📌 Level 4: Semi-Autonomous Agents – These agents handle multi-modal tasks with complex goal interpretation. - Picture an AI project manager that autonomously plans tasks, understands context from past projects, and adjusts timelines—all while keeping you in the loop for final approvals. - For example, Autonomous robots like Figure and even generalist agents like Manus AI. 📌 Level 5: Fully Autonomous Agents – This is the future of autonomous systems. These systems exhibit complete awareness, independent goal-setting, and advanced reasoning. - These are still theoretical, but from the examples of Level 4, you can guess that this means a fully autonomous software entity which can not only plan but also act accordingly with very minimal supervision. 🔗 You can learn these concepts more in-depth from my latest book: https://amzn.to/4irx6nI Save 💾 ➞ React 👍 ➞ Share ♻️ & follow for everything related to AI Agents

223

41 Comments

After 200+ hours of trial, error, and binge-watching  These 8 videos are what you need to understand data science: 1. 𝐒𝐭𝐚𝐭𝐢𝐬𝐭𝐢𝐜𝐬 𝐅𝐮𝐧𝐝𝐚𝐦𝐞𝐧𝐭𝐚𝐥𝐬 (Josh Starmer - StatQuest): https://lnkd.in/duGqwSp2 2. 𝐏𝐲𝐭𝐡𝐨𝐧 𝐟𝐨𝐫 𝐃𝐚𝐭𝐚 𝐒𝐜𝐢𝐞𝐧𝐜𝐞 (CampusX): https://lnkd.in/d9G8gAm7 3. 𝐌𝐚𝐜𝐡𝐢𝐧𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐏𝐥𝐚𝐲𝐥𝐢𝐬𝐭 (Krish Naik): https://lnkd.in/dcpAS5gA 4. 𝐃𝐞𝐞𝐩 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐒𝐩𝐞𝐜𝐢𝐚𝐥𝐢𝐳𝐚𝐭𝐢𝐨𝐧 (Andrew Ng): Available on Coursera 5. 𝐒𝐐𝐋 𝐟𝐨𝐫 𝐃𝐚𝐭𝐚 𝐀𝐧𝐚𝐥𝐲𝐬𝐢𝐬 (Alex the Analyst): https://lnkd.in/dAWvbVMV 6. 𝐃𝐚𝐭𝐚 𝐒𝐜𝐢𝐞𝐧𝐜𝐞 𝐂𝐨𝐮𝐫𝐬𝐞 (CampusX): https://lnkd.in/ddf8pMJt 7. 𝐏𝐚𝐧𝐝𝐚𝐬 𝐌𝐚𝐬𝐭𝐞𝐫𝐜𝐥𝐚𝐬𝐬 (CampusX): https://lnkd.in/dSpy4YNN 8. 𝐃𝐚𝐭𝐚 𝐕𝐢𝐬𝐮𝐚𝐥𝐢𝐳𝐚𝐭𝐢𝐨𝐧 (CampusX): https://lnkd.in/d2HqsNgv - 𝐒𝐭𝐚𝐭𝐐𝐮𝐞𝐬𝐭 turns intimidating formulas into crystal-clear explanations. - 𝐊𝐫𝐢𝐬𝐡 𝐍𝐚𝐢𝐤 covers every ML algorithm with real examples. - 𝐂𝐚𝐦𝐩𝐮𝐬𝐗 gives you the complete roadmap from zero to job-ready. Most courses overwhelm you with theory. These creators focus on what actually works in practice. 𝐘𝐨𝐮𝐫 𝐚𝐜𝐭𝐢𝐨𝐧 𝐩𝐥𝐚𝐧: Pick one playlist. Follow it completely. Build projects along the way. That focused learning beats jumping between 20 different courses. ♻️ Repost to help someone break into data science 𝐏.𝐒. I share job search tips and insights on data analytics & data science in my free newsletter. Join 16,000+ readers here → https://lnkd.in/dUfe4Ac6

231

43 Comments

ChatGPT 5 vs Claude vs Gemini vs Grok vs DeepSeek The world of AI Models are moving fast, but not all models are built for the same purpose. Some are good at multimodal creativity, others work flawless in research, cost-efficiency, or real-time data. Knowing the right tool for the right job can make you an AI experts. Here’s the breakdown : ChatGPT 5 -> The all-rounder model which is great for text, audio, video, code, and automation-heavy tasks. Gemini -> Perfect if you work in Google Workspace, it comes deeply integrated with Gmail, Docs, Sheets, and Drive. Grok -> This is a real-time social AI for X/Twitter which provides Fun- witty answers and is perfect for breaking news. DeepSeek -> The open-source star, It is Cost-effective, technically strong, and built for scale. Claude -> The AI that is built for safety has Long-context reasoning (200k+ tokens) and works precisely with legal/policy-heavy work. Key takeaway: Mastering AI is not just about picking “the best” model… it’s about choosing the right model for the right task. [Explore more in the post] If you found this helpful don’t forget to save this for later and comment your thoughts. Follow Denis Panjuta on Linkedin : https://lnkd.in/eUHjTBUi

957

131 Comments

𝐇𝐨𝐰 𝐏𝐞𝐨𝐩𝐥𝐞 𝐔𝐬𝐞 𝐂𝐡𝐚𝐭𝐆𝐏𝐓 A groundbreaking NBER study analyzing 18 billion weekly messages reveals surprising insights that should reshape how companies approach AI implementation: 📊 Only 27% of ChatGPT usage is work-related (down from 47% in 2024) 80% of conversations fall into just 3 categories: Practical Guidance, Writing, and Information Seeking. "𝐀𝐬𝐤𝐢𝐧𝐠" 𝐟𝐨𝐫 𝐚𝐝𝐯𝐢𝐜𝐞 𝐜𝐨𝐧𝐬𝐢𝐬𝐭𝐞𝐧𝐭𝐥𝐲 𝐨𝐮𝐭𝐩𝐞𝐫𝐟𝐨𝐫𝐦𝐬 "𝐃𝐨𝐢𝐧𝐠" 𝐭𝐚𝐬𝐤𝐬 𝐢𝐧 𝐮𝐬𝐞𝐫 𝐬𝐚𝐭𝐢𝐬𝐟𝐚𝐜𝐭𝐢𝐨𝐧. 💡 Key Takeaways for Business Leaders: ✅ DO: Focus on decision support over task automation. Users value AI as an advisor more than an executor. ✅ DO: Prioritize writing assistance - it dominates work usage at 40% of professional interactions. ✅ DO: Design for consultation and guidance rather than complex workflows. ❌ AVOID: Over-investing in programming use cases (only 4.2% of usage) unless you're in tech. ❌ AVOID: Ignoring personal use cases - they drive 70% of engagement and adoption. 🎯 Bottom Line: Companies succeeding with AI align with natural user behavior - prioritizing advisory functions, writing support, and information synthesis over pure automation.

55

2 Comments

𝗛𝗼𝘄 𝘁𝗼 𝗖𝗮𝗹𝗰𝘂𝗹𝗮𝘁𝗲 𝗡𝗲𝘄 𝘃𝘀. 𝗥𝗲𝗽𝗲𝗮𝘁 𝗣𝘂𝗿𝗰𝗵𝗮𝘀𝗲𝘀 𝗶𝗻 𝗕𝗶𝗴𝗤𝘂𝗲𝗿𝘆 (𝗪𝗶𝘁𝗵𝗼𝘂𝘁 𝗟𝗼𝘀𝗶𝗻𝗴 𝗙𝗮𝗶𝘁𝗵 𝗶𝗻 𝗛𝘂𝗺𝗮𝗻𝗶𝘁𝘆) #GA4’s interface makes everything look beautiful. You open a report, see a neat metric like first_purchase, smile, and move on. But in hashtag #BigQuery, things are a little more honest — and, as usual, a bit more complicated. But hey, we’re not here for the easy way out, are we? So, what counts as a new purchase? It’s when a user buys from you for the very first time. Not the second, not the third—the first. That magical moment when a marketer’s heart skips a beat. And a repeat purchase? That’s any transaction that comes after the first one. Maybe less exciting, but way more reliable. And let’s be honest—reliability is everything. How do you calculate this in hashtag #BigQuery? Let’s say you’ve got your purchase events. Then: 1. Group all purchases by user_pseudo_id 2. Find the first purchase date for each user That’s it. A little window function magic, and boom — you’ve got a clear split between new customers and your regulars. Why bother? • To know how many people actually bought for the first time • To measure retention in revenue, not just in pretty percentages • So your sales team doesn’t mistake 1 customer with 3 orders for 3 new clients Bottom line: GA4 will show you a “first purchase” metric — but it won’t tell you the story. BigQuery gives you the full picture: who bought, when, how often, and how much revenue they’ve brought in. SQL isn’t pain — it’s just the truth, plain and simple. Who notice a mistake or has a less complicated solution? Drop it in the comments! Want to finally understand where your customers are really coming from? Check out my free lesson — we’ll break down traffic sources and figure out which ones actually drive results: https://lnkd.in/gSz_kywp

86

5 Comments

Everyone’s racing to build the ‘next-gen’ of AI - new models, new tools, new capabilities. But the ‘baseline’ for what AI even is hasn’t been set yet. This space is a moving target, and every leap forward creates new opportunities and new ways for bad actors to exploit it. It’s mind-blowing to see the innovation happening right now. At ActiveFence, we’ve always believed that if you understand how bad actors operate, you can train the best models to handle the real world and today, our infrastructure is already protecting millions of conversations every day. Innovation without safety is a risk. Innovation with the right guardrails changes the game. Let's keep going! Thanks, Mo Sadek, for an awesome conversation and tour!

75

3 Comments

The AI Paradox Report – 2025 Early adopters of AI show a striking contradiction: They benefit personally from AI every day, yet they worry about its broader impact on society. In July 2025, I surveyed early adopters to capture this tension. The results reveal what I call the Pragmatic Paradox: adoption is rising, but trust is falling. This report presents three key frameworks: The AI Journey Map – optimism in personal use vs. concern for society The Tension Web – eight dimensions of risk, with truth and identity at the core The Paradox Quadrant – most respondents benefit individually while doubting collective outcomes You can explore the full report here on LinkedIn. Where do you see yourself in this paradox?

40

4 Comments