An opinionated guide to which AI to use to do stuff
AI is moving from simple chat interfaces to agentic systems that can perform multi-step tasks autonomously.
This guide explains how these systems work by giving AI access to “computers” or specific tools. Instead of just answering questions, agents can plan a project, research the web, and organize files in one go. The author suggests using basic models for simple questions, like writing a letter or finding a recipe. For high-stakes work involving legal or complex data, he recommends using the most advanced models because they have lower error rates. He also describes two ways to give AI more power: using a virtual environment provided by the AI company or giving it access to your own machine to handle large, multi-file projects.
For those working in the public sector or large organizations, this shift changes how we think about productivity. It moves the focus from “writing prompts” to “managing workflows.” Instead of doing one step at a time, you can delegate several hours of human work to an agent. However, this introduces new requirements for governance and security. Because these agents can take actions like sending emails or modifying files, setting correct permissions is vital. You must decide what the AI can do alone and where it must always ask for human approval first. This is a key consideration for maintaining data security and preventing errors in professional environments.
Are you already using AI agents to handle multi-step workflows, or are you sticking to simple chat prompts?
#ArtificialIntelligence #PublicSector #AIGovernance #DigitalTransformation #TechTrends