When AI agents can complete whole tasks, the teams in these six pieces redesign the work around them. They show the result before the tool, take workflows apart first, set checks before release and make people responsible for managing agents. One engineering piece from the same day asks whether engineers still enjoy the work once their team codes with AI.
Luke Wroblewski (26 Sep · 01), a product designer, helped build Exposit, a news site whose articles are found, written and edited by AI agents. It launched looking like any news site, with headlines, articles and categories to browse. A later feature compiled a personalised report on any topic a reader asked about, and it quickly became the main way people used the site. Wroblewski draws a general lesson: with AI, a product can deliver the result first and offer the tool second.
Burke Holland (26 Sep · 02), who works on AI-powered development at GitHub, makes a similar case from the builder's side. When a person keeps asking an AI agent in chat to do the same job, each request costs tokens, the units of text an AI model is billed by. Holland instead has the agent build a small app for the job once, inside GitHub's Copilot app, after which using it costs nothing. His examples include a database browser, a blog editor and a tool that runs his whole working process and calls him in for review.
Poly Palaiogeorgou (26 Sep · 03) describes, on Microsoft's own blog, how AI adoption changed in the company's Europe South region. Teaching thousands of staff to write prompts had brought gains that faded, but agents that finish whole tasks spread from colleague to colleague. A pilot in Spain sent each idea for an agent through a triage first: a better prompt, an existing tool, or, only when needed, a new agent. Every agent at Microsoft must also pass five checks before release, including privacy, accessibility and responsible-AI reviews. The report says the hard questions have moved on: "Adoption today is no longer the primary obstacle."
A second Microsoft account, by David Hirning (26 Sep · 04), covers the company's salespeople, who spent most of their time on admin across more than a dozen tools. Rene Vejlby, a business programme manager, and his colleagues first took every sales workflow apart. They decided what the company's sales agent should do and what to cut, and Microsoft then rebuilt the agent to complete whole tasks. In Vejlby's words: "If you have a broken process, just laying AI on top of it is not going to fix it." Microsoft is now moving to judge the agent by how fast deals close, not by how many people use it.
Maggie Appleton (26 Sep · 05), a research engineer at GitHub who builds prototypes, told the podcast host Gergely Orosz how her design work has changed. She has coding agents add sliders and colour pickers to each prototype, so that she can adjust it while it runs. She no longer reads the code the agents write for prototypes, and instead writes a detailed spec that lists how the agent will verify its work. She also finds planning with agents tiring, because an agent asks multiple-choice questions "a hundred times over".
Dan Graves (26 Sep · 06), chief product officer at WitnessAI, asked seven people on his product team to manage AI agents as they would junior employees. A design agent turns designers' mockups into code, an engineering agent reviews that code, and a person approves work before it moves on. When the design agent started building after a designer only pointed it at a task, the team made it ask questions first and build nothing without an explicit go-ahead. A designer on the team, Katie Chen, sums it up: "The net effect is we all end up acting more like product owners, not just holders of one job title."
Shubham Shah (26 Sep · 07), a security researcher and co-founder of the company Assetnote, published a message from an engineer he manages, with his reply. Since the team moved to developing with AI, the engineer no longer fully understood the code they shipped and had lost touch with what colleagues were doing. Shah agreed that AI has turned engineers into something like middle managers, and said he felt the same loss in his own research. He proposed that the team present its work to restore pride in it, and said he would not push the engineer while they adjusted. He concluded that the team needs more deliberate effort on "people's pride in their work, human connection, and communication with our engineers".
Each team changed something around the agent: the order in which a user sees things, the workflow the agent runs, the checks it passes or the people who direct it. In every account a person still decides what the agent may do, and in the company accounts a person approves the result.
Four of the six pieces are companies or executives describing their own products and teams, and none reports an independent measure of results. Wroblewski's evidence comes from content products, where a generated report that misses the point costs little.
Designer Luke Wroblewski found that readers of his AI newsroom mostly asked for a report on their question instead of browsing articles, and argues for outcomes before tools.
Burke Holland of GitHub argues that once a person knows a repeated job, having an AI agent build a small app for it costs fewer tokens than asking in chat.
Microsoft reports that staff in its Europe South region now take up AI agents without being pushed, and that the hard part has moved to governing them.
Microsoft had a business programme manager take its sellers' workflows apart before rebuilding its sales AI agent, and is moving to judge the agent by deal speed rather than usage.
Maggie Appleton, a research engineer at GitHub, no longer reads the code AI agents write for her prototypes and writes a spec saying how the agent will check its work.
Dan Graves, chief product officer at WitnessAI, has seven people on his team manage AI agents like junior hires, with a person approving work before it moves on.
Shubham Shah, co-founder of the company Assetnote, published an engineer's message about losing satisfaction since the team moved to coding with AI, and his reply as their manager.