Tobias Zwingmann's Linkedin Analytics

AI Agents are clearly overhyped, esp. in broad applications. Agents that "do sales" or "run your business" fall short. But in well-defined, high-value niches, AI agents are game-changers. One such area is the semantic layer: What’s a semantic layer? A semantic layer acts as the glue between your (often fragmented and complex) data sources and your (often therefore frustrated) business users who simply want fast, reliable insights. Building a semantic layer from scratch involves mapping data fields across disparate sources. A "customer" might have 12 different definitions depending on which system you ask. This manual mapping process has been extremely time-consuming, and expensive. This is where AI-powered Semantic Agents come in. These intelligent systems make it much easier to build and maintain a semantic layer by reducing the manual effort, cost, and risk involved. I recently collaborated together with Prashanth H Southekal, PhD, MBA, ICD.D and Arun Marar, Ph.D. on a new D2A2 report to explore how Semantic Agents can help organizations break down data silos and accelerate time to insight-driven decision-making. Interested? Check out the report – it's free! Download link in the comments. ⤵️

Those who start their AI journey with AI agents will most likely fail. Why? Because agents by definition require you to build AI solutions that are both highly integrated as well as highly automated (otherwise it wouldn't really be an agent). Doing one of these is hard. Tackling both at once is near impossible. If you've never done it, here's what typically happens: 1) The agent looks good in a demo 2) The agent utterly fails in production 3) You spend some time "fixing the data" 4) You spend a lot more time to "fixing the data" 5) You strip away some tools/data (less integration) 6) The agent thing still doesn't work 7) You put a human in the loop (less automation) 8) You rebrand your agent into an "assistant" 9) You don't want to hear anymore about "AI"  10) Your AI journey ends right there So why all the hassle? Instead of going through hell of building complex AI agents, start with simple, augmented AI workflows (AI assistants or AI copilots) that you can integrate-automate more as you go. Gives you a ton of opportunities to build a great roadmap and unlock AI agents where it makes most sense for you. Augmentation shouldn't be the end of your AI journey. Augmentation should be the beginning. — PS: Join my newsletter for more insights like this: https://lnkd.in/eFzzQrMJ

Most companies approach AI by boiling the ocean. Here's what usually happens: 1. Hold endless innovation workshops 2. Generate 100+ possible use cases 3. Create pretty PowerPoint slides 4. Pick projects that sound cool 5. Watch them die a slow death Want to skip all that and find real AI opportunities in under 3 hours? Process mapping will reveal where AI actually makes sense for your business. Here's how I do it: 1/ Start by mapping 2-3 key business flows you're responsible for Examples: Lead generation to closed sale. Customer onboarding to renewal. Support ticket to resolution. 2/ For each flows, outline 4-6 critical steps Don't go overboard - if you have 3 flows with 6 steps each, that's already 18 tasks to analyze. 3/ Now mark the specific tasks where people waste time, make mistakes, or get stuck in bottlenecks. Things that hurt you now or soon. ✅ Check 1: You've just prepped your AI opportunity map. From here, it's much easier to prioritize problems and fit them to an AI skill that can actually help. (Which becomes your AI solution.) You're not trying to "do AI" anymore. You're solving real business problems with AI where AI is the right tool. I ran this exercise in a workshop few weeks ago. Took 90 minutes total. We identified 7 high-impact opportunities that the business actually needed to have solved. "We just did in 90 minutes what others do in 90 hours." Don't start with AI. Start with your business processes. The right AI opportunities will reveal themselves. —— PS: Join my newsletter for more insights like this: https://lnkd.in/eFzzQrMJ