tylers-blogger-blog

At 9:47 PM, our "AI agents" were not a team. They were five talented interns talking over each other. One agent would draft a plan. Another would "helpfully" rewrite it. A third would start coding before anyone agreed on requirements. Meanwhile, the QA agent nitpicked formatting while the data agent pulled numbers from the wrong date range. Nothing was broken individually-coordination was. That night, we stopped trying to make one super-agent do everything and instead built a small system where each agent had a job, a shared memory, and rules for taking turns. The turning point: roles, boundaries, and a single source of truth The first fix was embarrassingly simple: we wrote role cards like you'd do for humans. Planner plans. Researcher gathers sources. Builder implements. QA tests. Reporter summarizes. Each role had: Inputs it is allowed to read Outputs it is allowed to write A "definition of done" checklist Then we introduced one shared artifact: a ...

Project management doesn't fail because people don't care. It fails because coordination is expensive: status updates, handoffs, "who owns this?", outdated docs, and the slow drift between what we planned and what actually happened. This manifesto is simple: stop using AI as a fancy autocomplete. Start using AI agents as teammates-purpose-built, role-based assistants that move work forward across your tools, with guardrails. 1) What "AI agents" mean in project management (and what they don't) An AI agent isn't "a bot that chats." It's a system that can: Understand a goal (e.g., "keep this sprint on track") Take actions in your workflow tools (create tickets, update fields, draft messages) Watch signals (PR activity, calendar changes, blockers in comments) Report outcomes and ask for approval when stakes are high Think of agents as specialists , not a single all-knowing manager. A practical stack might include: Intake Agent : t...

  If you've ever used a single AI assistant for a complicated project-launching a feature, writing a sales sequence, doing market research-you've probably felt the pain: it can be brilliant for 30 seconds, then drift, forget constraints, or confidently hand you something half-right. AI agent teams are the antidote. Instead of one generalist, you orchestrate a small squad of specialists with clear roles, handoffs, and checks. You stop "chatting" and start "running plays." This post is a tactical, practical guide to building AI agent teams that consistently ship usable work-without turning your process into a science project. What "AI agent teams" actually means (and what it doesn't) An AI agent team is a set of role-based agents (often multiple prompts, models, and tool permissions) working together under a shared objective. They collaborate via structured outputs and explicit handoffs. What it is: A workflow : tasks are decomposed, assigned, ve...

It started like any other "quick deploy" that absolutely shouldn't have been quick. We'd shipped a small feature behind a flag, everyone was feeling smug, and I was already planning my post-merge snack. Then the alerts hit: latency climbing, error rate spiking, and a weird pattern-only one region was melting down, and only for a subset of users who'd never touched the new feature. And that's when our AI agent-internally nicknamed "Patch"-walked into the room like it owned the place. The incident: 47 minutes of chaos (and one calm agent) Our agent wasn't a "magic button." We'd wired it into three things: read-only observability (logs/metrics/traces), our runbook repo, and a tightly scoped toolbelt (query dashboards, open PRs, suggest rollbacks, and draft incident updates). No direct production writes. No cowboy stuff. At 12:18 AM, it posted in Slack: "Spike correlates with deploy `api-7f3c` but only in `eu-west`." "Mos...

I didn't "discover" AI agents in a lightning-bolt moment. It was more embarrassing than that. It happened on a Tuesday afternoon when I was wrestling with a refactor I'd been postponing for weeks: a tangle of conditional logic, half-duplicated helper functions, and comments like "TODO: make this not terrible." I asked an AI agent to draft a cleaner version expecting something... fine. Instead, it handed me code that was not only cleaner, but also easier to test, easier to read, and annoyingly consistent. The moment it out-coded me The agent's first win wasn't clever algorithms-it was basic engineering hygiene. I fed it a simplified prompt: "Refactor this function to be easier to test; keep behavior identical; add unit tests." The before-code had a long function doing validation + parsing + persistence + logging. The agent responded with: Small pure functions (validate, transform, persist) Dependency injection for the DB and logger Tests tha...

It started as a joke: "Let's see if the bots can survive a Friday night in our office." We'd been testing AI agents in small, polite ways-summarize a doc, draft an email, pull a quick report. Useful, sure, but not exactly game-changing. Then we looked at the calendar: a Monday exec update, a customer renewal call, a backlog of support tickets, and a spreadsheet that always seemed to multiply after 5 p.m. So we did the thing you're not supposed to do: we let a handful of AI agents run while we went home. The Setup: A Few Agents, Clear Boundaries, Real Work We weren't trying to build "Skynet for spreadsheets." We created three agents with narrow jobs, strict permissions, and obvious stop signs. 1) Inbox Triage Agent (read-only + draft-only): It scanned a shared inbox, labeled threads (billing, bug, onboarding), extracted key details (customer name, urgency, due date), and prepared draft replies using our saved tone guidelines. Nothing sent automatical...

It started as a "tiny experiment" we promised wouldn't disrupt anything. Two weeks later, we realized our AI agents weren't just helping-they were quietly becoming the most reliable teammates in the building . The surprising part wasn't that the agents could write copy or summarize notes. It was how they stitched together dozens of small, annoying tasks that normally fracture a day: the follow-up emails, the "can you pull that data again?" requests, the meeting prep, the status updates. Suddenly, our best people were doing more of the work only humans can do. The moment it clicked: less busywork, more momentum Our first win was embarrassingly simple: meeting-to-action automation. Before: someone took notes, someone else "cleaned them up," tasks got lost, and we'd spend the first 10 minutes of the next meeting re-litigating decisions. After: we used an AI agent that listened to the meeting transcript, produced: A 5-bullet executive summary D...