Jason Hood

Jason HoodAug 12, 2026

Same script. Same agent. One toggle changed how believable it sounded. We built a voice agent from scratch and ran it through a test call with Expressive Mode off. It answered correctly. It just sounded like it was reading. Then we turned Expressive Mode on and ran the exact same call again. Nothing about the logic changed. The words were identical. What changed was timing, emphasis, and tone — the agent paused where a person would pause, and leaned into the parts of the sentence that actually mattered. That's the gap most teams miss when they evaluate voice AI. They test for accuracy and stop there. But in a real support call or sales conversation, delivery is doing as much work as the answer itself. A correct response delivered flatly still reads as "clearly a bot" — and that perception costs trust before the conversation even gets to the point. We recorded both versions back to back so the difference is easy to hear rather than take our word for it. It's a one-minute watch. If you're building or evaluating voice agents, this is worth testing on your own use case before you assume "good enough" accuracy is enough. Try ElevenAgents: https://lnkd.in/geh-xYPS #AIAgents #VoiceAI

Jason HoodAug 13, 2026

Everyone assumes shipping a production AI agent takes a dev team and a few sprints. Here's what it actually took. We signed up, uploaded a knowledge base, connected the tools it needed to actually do something (not just talk), and tested the agent live — all in one sitting. No custom backend. No weeks of integration work. The agent had context (from the knowledge base) and capability (from the connected tools) before we'd even finished the coffee. That combination is the part worth paying attention to. A lot of "AI agent" demos stop at conversation — it sounds smart, it answers questions. But an agent that can only talk is a chatbot with a nicer voice. The moment it can look something up, check a status, or trigger an action through a connected tool, it's doing actual work instead of just describing it. The gap between "impressive demo" and "something you'd trust in production" usually isn't the model. It's whether the agent has the knowledge and the access to be useful once the conversation gets specific. Full setup is in the video — a little over a minute, start to finish. Try ElevenAgents: https://lnkd.in/geh-xYPS #AIAgents #FutureOfWork

Jason HoodAug 11, 2026

Enterprise AI gets far more interesting when companies stop just renting intelligence and start owning what actually makes them different. Access to the same frontier models won’t be the moat forever. The real advantage will come from proprietary data, specialized workflows, continuous feedback loops and intelligence that compounds inside the business over time. That’s why this perspective from Oumi stood out to me. The next phase of enterprise AI isn’t just about using better models. It’s about building intelligence that becomes uniquely valuable to your organization.