Building autonomous agents? Here are the arguments I keep bumping into—speaking from the trenches: Vectorize everything vs. giant context windows You say “vectorize your data.” they say “why bother… I’ll just load the raw docs at runtime with a jumbo context window.” RAG for grounding vs. let-the-model-read-it-all You say “RAG + ranked chunks keeps answers …
As we advance Vizerto to take on Agentic Tasks; here are a few first-hand observations: Tokens Blow Up Fast – The moment you let an AI agent do heavy research or reflections, the token count can shoot through the roof. This has direct $ implication. Run Times Can Be All Over the Place – Sometimes …
With autonomous agents able to navigate across CRM, Order-to-Cash, Customer Support, and more… do we still need centralized systems of record? The case for disruption: Agents dynamically query multiple systems to create task-specific data views No need for rigid pipelines—data is pulled on demand We shift from “Where’s the official data?” to “What’s needed for …
Is the “System of Record” still relevant in the age of AI Agents? Read More »
As we develop AI agents, here are a few important considerations we’ve discovered: Current APIs are designed for humans driven workflows AI agents need APIs to navigate and access application data, but many existing APIs are built with human-assisted workflows in mind. This limits the agent’s performance potential because the APIs may not expose direct, …
Most of us tackle a task—be it responding to customer inquiry, writing business requirements, test and validation, conducting market analysis, or developing a sales pitch—through our personal mental model: Past experience bias (“I’ve always done it this way.”) Effort-vs-time trade-off (“How long will this take?”) Risk calculus (“If I try something crazy, will it blow …
The Mental-Model Ceiling—Why GenAI & Agents Matter Read More »
2022 © Vizerto. All rights reserved.
