Daniel Lemire's Linkedin Analytics

Have you ever thought about the mindset shift required for mastering AI tools? When you take the time to learn by making room for AI MISTAKES, that's where the muscle is built. Instead of just learning how to use them, it's crucial to understand the underlying principles that drive their functionality. This just takes time and practice. So don't wait until you know everything to get started. Find ways to do a deep dive into the algorithms and structures that power these tools. Read the latest papers and take a look through the github repos of popular software. There are all kinds of way to dig in. It's not enough to press buttons; we need to cultivate an agentic approach. This means taking initiative and responsibility for how AI is applied in our work. How are you ensuring that your engagement with AI extends beyond surface-level operation? 📝𝗜 𝘄𝗿𝗶𝘁𝗲 𝗱𝗮𝗶𝗹𝘆 𝗮𝗯𝗼𝘂𝘁 𝗔𝗜 𝗮𝗻𝗱 𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 𝗺𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁 🔔𝗥𝗶𝗻𝗴 𝘁𝗵𝗲 𝗯𝗲𝗹𝗹 𝗼𝗻 𝗺𝘆 𝗽𝗿𝗼𝗳𝗶𝗹𝗲 𝘁𝗼 𝗯𝗲 𝗻𝗼𝘁𝗶𝗳𝗶𝗲𝗱 𝘄𝗵𝗲𝗻 𝗜 𝗽𝗼𝘀𝘁 📰 𝗦𝗶𝗴𝗻-𝘂𝗽 𝗳𝗼𝗿 𝗺𝘆 𝗳𝗿𝗲𝗲 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿 📺 𝗦𝘂𝗯𝘀𝗰𝗿𝗶𝗯𝗲 𝘁𝗼 𝘁𝗵𝗲 𝗖𝗵𝗮𝗻𝗻𝗲𝗹 ☎️ 𝗚𝗿𝗮𝗯 𝗮 𝗰𝗮𝗹𝗹 𝘄𝗶𝘁𝗵 𝗺𝗲 👉 𝗦𝗲𝗲 𝗺𝘆 𝗽𝗿𝗼𝗳𝗶𝗹𝗲 𝗳𝗼𝗿 𝗱𝗲𝘁𝗮𝗶𝗹𝘀 𝗮𝗻𝗱 𝗹𝗶𝗻𝗸𝘀!

I imagine there's someone out there, rolling their eyes at my posts. "Oh, it's that AI MISTAKES guy again." 🙄 Well, if that's you, and you're annoyed, apparently, it's only going to happen more from here on out, because now that's basically how AI Agents work. Let me break it down! MISTAKES are stepping stones. It's how we know we're off track. It's how we identify opportunities for improvement. The trick is, we have to do it in a way that enables us to fail in a noncatastrophic way. That's why I built the MISTAKES framework. It helps you toward success, but in a way that doesn't create unnecessary risk. This helps you get to success, without undoing the good progress you've already made. But here's the really fun part—This is cutting-edge stuff now! The most widely released and cutting-edge agentic workflows use testing for MISTAKES to improve and learn. Let's say you give a task to an agent: 1.) you let it try out reasoning through the problem, identify likely steps and then have it propose a solution. (human-in-the-loop implementation... you are using human-in-the-loop right?!) 2.) When it proposes a right solution, you let it proceed with the plan, but when it makes a mistake (and you give it a correction), something amazing happens! 3.) It remembers the mistake (long-term memory) so it doesn't repeat it next time! These new frameworks include mechanisms for reasoning, tools, and both long and short-term memory. It's the long-term memory systems that are dramatically improving outcomes and reducing hallucinations. It's not just prompts and one-shot anymore. Now get out there and use AI to make some MISTAKES and learn! (and come back here to tell everyone what you've discovered!)

The AI revolution waits for no one. ServiceNow's own CEO, Bill McDermott, puts it: “The world has never changed this fast and will never move this slow again.” We are indeed seeing an increasing pace of change. I've even shared posts that chart this out (see my recent post regarding computing power growth and AGI predictions). But knowing about the increasing pace of change and doing something about it are quite different. In the Forbes piece linked here, Bill shares some of his views and our collective work in helping organizations fully leverage the new paradigm. http://spr.ly/6043LQwRB

Today is the day! ServiceNow Yokohama release is here. I’m super excited about AI Agent Studio and building custom agents, but you also must know that there might not be a huge need to build your own, because on day one we have preconfigured AI agents across IT, CRM, HR, industries, and data and analytics. Read on to meet some of the new ServiceNow AI Agents: http://spr.ly/60470OtLV

AI disappears when it works well, but how will you ensure it’s working well? As soon as organizations decide to start using AI, they need to understand the risks (and mitigations) involved. I’ve talked about data management in the past, but with generative AI, it’s not just the data used. Increasingly, it’s necessary to understand the performance of the AI models. This means you need the same rigor that has been part of DevOps for the last 10+ years. I’m not suggesting that you need DevOps for AI, but I am suggesting that you need a way to automate, manage, govern, and continuously improve the use of AI in your organization. This is possible today, and it only gets better from here. If you need help to sort out how to do this for yourself or your organization, let’s connect!

AI adoption isn’t just about the technology; it’s about the people. You have to consider both to be successful.

Agentic AI is here to stay, but what does that really mean? Agentic AI, offers autonomy and decision-making capabilities and represents a significant shift from conventional computing paradigms. Unlike traditional deterministic automation systems that rely on predetermined execution logic, these intelligent agents are designed to assess and act with minimal intervention. The word minimal here I think is where all the challenges will materialize. The methods, tools, processes, and capabilities are relatively new and far from well understood. Even the most well known foundation model builders don't fully understand what is possible and realistic, we're all building, testing and learning together. If you have traditional workflows and automation already working, the last thing you should do is trade it all in for AI agents. Instead, be thinking about AI Agents as the anti-workflow. Where do decisions need to be made, but you don't have an easily codified pathway already established. The truth is that combining AI Agents with existing tools and processes can be incredibly powerful. This is what I am most excited about for what's next. We've yet to really discover the use cases that will be the most impactful by combining deterministic and non-deterministic flows and processes. What excites you most about Agentic AI? Do you see more opportunities or challenges in your field?

In 1999, Ray Kurzweil predicted that AI would achieve human-level intelligence (AGI) in 2029, or 30 years. Is he right? It may not even matter, because what we can do already today is incredible. Just look at the chart below. This price-performance of computation chart tells you everything you need to know. (also provided by Ray in his book published last year The Singularity is Nearer: When We Merge with AI) Think of it this way: if you can imagine it, there will be a computer that can calculate it. Whether it's a simulation, an immersive experience, a complicated experiment or understanding your chemistry homework. The world is changing fast, and it's all based on what you see in this chart.

Many people see Generative AI primarily as a tool for producing content quickly. More often lately so much of what I see is about writing code. That’s the micro view, what’s the macro view? It won’t happen instantly, but there is a coming reduction in the cost and time associated with trial, error, and iteration. In the past, experimentation required considerable resources, making failure expensive. Today, with Generative AI it’s possible to conduct almost cost-free exploration of ideas. This is part of my premise behind AI MISTAKES. it is this allowing countless creative “failures” to happening rapidly, that leads to driving faster innovation. It’s not just how fast the AI capabilities are evolving, it’s this additional acceleration that is leading to rapidly evolving technology in other areas. When understood in this light, the true value of Generative AI emerges clearly: It is not merely further automation of existing processes, but empowering rapid learning cycles that is changing the world. This is ultimately why I believe it is vitally important for as many people as possible to use AI in hands-on ways toward the things they already know and understand. We do not fully know where this accelerated learning cycle will have the most significant impact. What’s one thing you have thought about in using AI, but you’re hesitating? I’d love to capture some of those insights. Perhaps we can solve it together. Let me hear from you in the comments or DM. Let’s Learn (together) from MISTAKES with AI.

Are you making the most out of AI, or simply adding more complexity? Let's dive into the nuances of AI adoption beyond the buzzwords. For many, it's not about having the shiniest new tool, but rather understanding how these technologies can be strategically aligned with your unique needs. It’s akin to possessing a Swiss Army knife; knowing when and how to use each component is key. Consider companies thriving right now—not because they have all AI tools at their disposal—but because they've mastered using the right ones at pivotal moments. I also see the opposite: companies that focus on having the Swiss Army knife with every possible tool available, but it's not clear when to use each tool. As we navigate this latest tech wave, ask yourself: Are you leveraging AI as a force multiplier in your strategy? Or are just adding complexity by adding more tools? Let's make sure we're part of shaping this future, not just chasing shiny new tools.

Google wins again! What happens when you combine Google's latest models with their new determination? A winning tool for image generation and editing!

What is the potential of Agentic AI? In a world bustling with automation and machine learning, Agentic AI stands out as a game-changer. Unlike traditional AI systems, Agentic AI adapts, learns, and makes decisions autonomously within defined constraints. This isn't just programming; it's evolving intelligence that mimics human decision-making processes. I've spent this whole week building and deploying agents. These new agent systems not only process data but understand context and evolve from interactions—a true partner in completing tasks! As we venture deeper into this field, I'm curious: How do you see Agentic AI transforming your work? Do you see more potential challenges or more breakthroughs and advantages ahead? What questions do you have?

Are you ready to face the future of artificial intelligence? The rise of AI has sparked heated debates about its potential impact on humanity. Will AI control your future, or will you be in control? Let's explore the possibilities and implications of AI on our daily lives, jobs, and society as a whole, and then get practical by examining the actions you can take for your future. Join the conversation and discover the answers to the questions that will shape our collective future.

If you spend time on AI Agents, this is an excellent representation of what has been happening over the last 27 months. Some of these boxes have greater implications than others, but they are all transformative in the application of Generative AI to creating value.

Last week, I had the privilege of attending and leading a workshop at the North Texas ServiceNow User Group (SNUG) at the Equinix InfoMart in Dallas. I had a fantastic time at this event, and I especially enjoyed the opportunity to learn from and with other professionals who use and manage their organization's ServiceNow platform. This was my very first SNUG event, and it was such a great opportunity to learn and share with a larger community. Thanks to everyone who helped make this a great event!

Last week was amazing at GTC. I'm not quite to the point where I get a press pass for these kinds of events (but that's also because I'm trying to learn and understand and neglect posting to social media). Besides the record-setting number of attendees (25,000 in-person/300,000 virtually), there were some fantastic announcements. I'm proud to highlight the work that ServiceNow and NVIDIA are doing together, advancing agentic AI to redefine enterprise intelligence. ServiceNow isn't slowing down on developing (and using) the latest developments in AI. The way that agents learn is an essential development, and these new AI Agent evaluation tools and integrations with NVIDIA’s Llama Nemotron Reason models are raising the standards for AI. 2025 is going to be an amazing year for AI agents, and we're only just getting started. Check out the full release at: http://spr.ly/60400CNQQ