Sasthi Sarma

When building an AI chatbot to retrieve order history in an e-commerce setting, consistency matters. This is why we set the model temperature low — to reduce randomness and deliver reliable, repeatable answers. LLMs are inherently probabilistic. Temperature controls how likely the model is to choose less probable outputs. Lower temperature implies more determinism; higher temperature allows for more creativity. It’s a key parameter when designing AI agents for enterprise applications. For example, consider this order status query: User: “What’s the status of my last order?” Response 1: “Your last order for a wireless mouse was delivered on July 28.” Response 2: “Looks like your recent order arrived a few days ago. Let us know if you need anything else!” Both are correct — but in enterprise workflows, variation like this can cause confusion. That’s why tuning model temperature is critical. A few seats are still available in Building AI Agents in Enterprise course that Ajay Shenoy and Abhijith Neerkaje are teaching. Link to register is in the comments below.

٦٩

١ تعليق واحد

A lot of AI news out of Walmart yesterday. Two new EVP level dedicated AI roles (AI Acceleration and AI Platfroms) elevates AI to the ELT at Walmart. They also announced 4 new super agents targeted at their 4 core audiences: New version of "Sparky" for shoppers;"Marty" for suppliers and advertisors; a new agents for Walmart's 1.6M associates; an agent for it's 10K+ in-house engineers. https://buff.ly/3MqIdWP

١١٢

٢ تعليق

What a read—AI is flipping every shopping aisle on its head. Key takeaways from Andreessen Horowitz article “AI x Commerce”: 1️⃣ Search isn’t dead—yet. AI is eating low-intent queries, but high-intent commerce searches are next. 2️⃣ Commerce isn’t one-size-fits-all. In their article, Justine and Alex maps five types of purchases—from impulse to life-defining—and shows how AI will reshape each: • Impulse → smarter targeting • Routine → agent-driven auto-purchase • Lifestyle → personalized curation • Functional → expert AI consultations • Life purchases → guided by AI coaches 3️⃣ AI commerce needs new infrastructure. For agents to work, we need: • Clean, real-time product data • Unified APIs for search, cart, and checkout • Smart identity & preference systems • Embedded capture of customer interactions 💡 We’re moving from human browsing to being guided. AI will first transform routine, lifestyle, and functional commerce—and fast. 🔗 The full great a16z article by Justine Moore and Alastair (Alex) Rampell: https://lnkd.in/eWVBW7eK 👉 That’s exactly why we’re building Mirakl Nexus: the agentic commerce OS powering this next wave of agentic commerce. Read more about Nexus: https://lnkd.in/efvSX2HN #Retail #AgenticCommerce #AI #DigitalTransformation

٧٢

My takeaway from last week’s CXO CDO conference was this: Data is pure logic. Empathy is pure biology. Will we ever truly bridge the two with AI? This thought has been stuck in my head all week. Look at the picture in this post. An AI can logically decompose it with incredible detail: "a large sailboat docked at a pier under a dark, star-filled sky." It sees the pixels, the objects and the data points; it gets the lighting by the shore mixed up with stars. But it can't feel the warm sea breeze, the gentle rock of the boat, or the sense of wonder for the beauty of the world we live in. It can describe the moment, but it can't be moved by it. 🧠 🧠 🧠 This is an empathy gap and can AI ever bridge it? AI will get exceptionally good at simulating empathy ("I see you're frustrated. How can I help?"). But if it can never genuinely feel, is a perfect simulation good enough for the future of business and customer experience? How much is customer experience about feeling? If human empathy is rooted in our physical experience, flooded with neurotransmitters, hormones and other biological responses, can AI ever really have empathy? 🤖 🤖 🤖 We spent a lot of time discussing what it could be like "managing" AI agents alongside humans in teams: what would it be like to line manage an AI? This moves the empathy gap from a technical problem to a leadership one. What new skills do leaders need to manage a hybrid team of humans and AI agents? How do you inspire, mentor, or performance-manage an algorithm? 🌟 🌟 🌟 Finally, I was struck by how often neurodiversity came up in our conversations, even without any formal sessions on it. From stories of neurodivergent individuals in our teams, to individuals being open about their own neurodivergence, it was the subplot to all the other conversations. It struck me that the neurodivergent brain may be one jigsaw puzzle that we need to bridge this empathy gap. If the empathy gap gets bridged in the future, is it catalysed by individuals who can sometimes feel more empathy and in different ways to others (the double empathy problem and hyper empathy?), but are hard wired to think very logically and in patterns? ❓ ❓ ❓ So many questions here! I'd love to know what others think, in the comments below! The key to a good professional development event is that you leave with as many new questions than answers to existing ones, as the content, the people and the surroundings force you to think differently and reflect. Thanks to Andrew Bahadoor and Siobhan Cooper and the whole CXO Group Ltd team for such an inspiring event and to Johanna Hutchinson for her inspirational leadership of the community. And, if I may, one build for next year could be to have an official table tennis tournament, so Nadim Antar or Jake Bengtson can beat me “officially” and claim a small trophy.

٩٠

٧ تعليق