AI in Practice — 2 March 2026

TPM at an AI company.
I use AI to do my job.

As a Technical Program Manager at Leonardo.ai, I sit at the intersection of product, engineering, and delivery. My job is to keep complex programs moving — and AI is how I do it faster, with less friction, and with better visibility across every team I work with.

Technical Program Manager Leonardo.ai · Melbourne Claude · Jira · Confluence · Slack
🗺️ Roadmap & delivery — Jira, Confluence 🤝 Cross-functional alignment — Slack, stakeholder comms ⚙️ Process automation — Claude Cowork, scripts 🔍 Codebase visibility — Cursor, Jira Rovo
"In 2025 Claude transformed how developers work, and in 2026 it will do the same for knowledge work."
— Kate Jensen, Head of Americas, Anthropic

AI as a TPM force multiplier

A TPM's job is to reduce friction between people, systems, and delivery. AI doesn't change that — it just means I can do it faster, across more workstreams, with less manual overhead.

🔗

Connect the tools, not just the people

Jira, Confluence, and Slack shouldn't require a human relay. I build automations that let these tools talk to each other so context flows without me being the bottleneck.

📡

Visibility without the overhead

Status updates, roadmap changes, stakeholder comms — these are necessary but time-consuming. AI handles the structure; I handle the judgment calls.

🛡️

Blockers are part of the workflow

Real automation hits real errors: API 404s, permission gaps, format mismatches. The skill is designing workflows that handle them gracefully and keep delivery moving.

5 TPM Workflows I Built with Claude

These aren't experiments. These are automations I built and run as part of my actual delivery work — using Claude Cowork as my AI colleague across Jira, Confluence, and Slack.

01 / 05

Kickoff

Project Kickoff in One Command

Jira Confluence Claude Cowork

Spinning up a new program used to mean: open Jira, create the Epic, switch to Confluence, create the kickoff doc from scratch, then manually link everything. Now one /project-kickoff command does all of it — Epic fields populated, kickoff doc templated, both linked automatically.

  • 1Run /project-kickoff with program name and lead
  • 2Jira Epic created with correct fields, labels, and sprint
  • 3Confluence kickoff doc generated from team template
  • 4Jira ↔ Confluence linked — ready to share with stakeholders
Takeaway What used to take 20+ minutes of tab-switching now takes one command. The time savings compound across every program kickoff.

02 / 05

Roadmap

Roadmap Updates Without the Context Switch

Confluence Claude Cowork

The TPD roadmap lives in Confluence and needs constant updates — sprint changes, scope shifts, milestone slips. Rather than navigating to the right page and editing tables manually, I push changes directly from wherever I'm working.

  • 1Call /roadmap-update with sprint, status, and change summary
  • 2Claude finds the right section of the roadmap page automatically
  • 3Table updated in place — correct format, correct columns
  • 4No tab-switching, no navigation, no reformatting
Takeaway Staying in flow matters. Every context switch costs more time than it looks like. Keeping roadmap updates frictionless means they actually happen.

03 / 05

Stakeholder Comms

Stakeholder Updates That Write Themselves

Jira Slack Claude Cowork

Stakeholder updates are repetitive by design — same structure, same tone, different details every sprint. The /stakeholder-update command pulls current Jira status, drafts a structured update in the right format, and sends it to the right Slack channel.

  • 1Run /stakeholder-update with the program epic
  • 2Claude fetches current Jira status: progress, blockers, next steps
  • 3Drafts structured update: summary → status → risks → next sprint
  • 4Sent to #program-stakeholders — reviewed, not rewritten
Takeaway The update still comes from me. Claude removes the grunt work of getting it from Jira into Slack in the right shape.

04 / 05

Cross-Tool

Jira + Confluence + Slack in One Session

Jira Confluence Slack

Starting a new workstream means creating an Epic in Jira, a spec page in Confluence, and a Slack channel — then linking all three. What normally takes 20+ minutes of tab-switching can now happen in a single Claude session.

  • 1Jira Epic created with correct fields, labels, and assignee
  • 2Confluence API returned 404 — no write access to target space
  • 3Claude created the page under my own space as a placeholder, flagged the permission gap
  • 4Slack channel created and linked — full cross-tool setup complete
Takeaway When the API blocks you, a smart fallback keeps delivery moving. Cross-tool setup across three platforms, one session.

05 / 05

Team System

AI for the Whole TPM Team

Cowork Plugin Jira Confluence Slack

The real leverage isn't automating my own work — it's packaging these workflows into a reusable plugin the whole TPM team can call. One command for kickoff, one for roadmap updates, one for stakeholder comms. Built once, used by everyone.

  • 1/project-kickoff — Epic + kickoff doc, every time, for every PM
  • 2/roadmap-update — consistent format, pushed to the right place
  • 3/stakeholder-update — structured message, right channel
  • 4Marketplace plugins layered on for additional out-of-the-box capabilities
Click to see full case study → A plugin isn't one automation — it's a reusable system. Build it once, run it every kickoff.

How AI Shows Up in TPM Work Every Day

Beyond the automations — the day-to-day ways AI has changed how I work as a TPM. Each one is a real habit, not a one-off experiment.

🌅

AI morning briefing replaces Slack scrolling

A Python script pulls Jira updates, key Slack threads, and industry news every morning at 7am and lands a structured digest in my inbox. I start with context, not catch-up.

📋

Jira Rovo makes me a better ticket writer

I use Rovo to draft ticket descriptions, acceptance criteria, and surface related work I'd otherwise miss. Tickets that used to take 20 minutes now take 5 — and they're clearer.

🖥️

Cursor gives me visibility into the codebase

As a TPM, I can now ask Cursor to explain what a function does, trace where a feature lives, or understand a PR before a review — without needing an engineer to walk me through it.

✏️

Claude reviews my Slack messages before I send

Before anything sensitive or high-stakes goes out, I run it through Claude. It catches tone issues, tightens wording, and makes sure the message lands the way I intend.

💬

Slack channel recap across 15+ channels

I stopped manually scrolling every morning. AI recap surfaces what actually matters across all engineering and product channels in seconds — nothing important slips through.

🌐

Claude for Chrome — QA without leaving the browser

I use Claude for Chrome to review docs, verify layouts, and check content accuracy without switching tools. Fast QA built into the normal review flow.