By Markus Kopko
October 15, 2026
Most project managers I meet use AI. They have a chat window open, a few prompts that work well, and notes with wording they reuse. But do you find yourself repeating the same instructions and context every time you use AI?
The solution is to move that repeated work out of the conversation into repeatable assets. Here are three practical steps to help you save time. Each steps removes a different kind of repetition, and help you get more consistent results with AI.
1. Save the process along with the wording
A saved prompt captures how you ask. It does not capture what the assistant needs to know. You still paste the template, still explain the standard, still describe the project, every time.
Every custom AI tool, regardless of platform, has three parts. Instructions define its role, required output structure, and quality standards. Knowledge provides the reference material, such as your approved template, one or two strong examples, and your organizational standard, uploaded once. Context is the specific information you provide each time you use the tool. For a charter, that means the project name, the sponsor, business case summary, constraints and target dates.
The test for whether something deserves this treatment is easy. Build a tool when the work recurs, when the output has a stable shape, when clear rules or templates govern it, and when you need to re-explain the same background. Charters and status reports pass that test in most organizations. One-off analysis and pure judgment calls do not. A good prompt is the right answer there.
2. Give the tool the history as well as the task
Some project work evolves over time and requires more than a single pass. Risk management is a clear example: you assess risks at the start, revisit them throughout the project, and track how they change. The real value comes from preserving that history, so each new assessment builds on what came before.
A tool can hold that through-line if you let it. Load the exports from your earlier assessment cycles into the knowledge, and add one instruction: “Compare against the previous cycle and label every risk as new, recurring, escalating or closing.” From that point the tool stops producing isolated snapshots.
What comes back changes character. A vendor delivery risk that was medium two cycles ago and is now high, gets called out as escalating. A theme that has appeared in every single cycle gets named as recurring, which signals that it needs a structural response instead of another one-off mitigation. A risk that is fading gets marked as closing. Nobody had to hold three months of register history in their head.
The same idea works beyond risk. Lessons learned, quality reviews and stakeholder analysis all improve when the tool can see the history.
3. Package the tool so it travels
The third step is the one most teams may not have taken. A tool that lives in your personal workspace helps exactly one person. The next move is to package it so it becomes portable.
Skills are the current shape of this. A skill bundles the instructions, the steps and the templates into one unit, and the assistant loads it only when the work calls for it. The practical gain is reach. The same tool runs in your chat application, in your development environment and through an interface other systems can call, instead of sitting inside a single workspace. It can be shared and versioned, so a team runs one tool the same way instead of six slightly different personal copies.
That last point is the real prize. A team gets consistent output when everybody uses the same configured asset. Waiting until six people are equally good at prompting takes much longer.
Where to start this week
Pick one process you run at least monthly. Write the instructions, including the role, the output structure, and the rules. Add your template and one strong example as knowledge. Run it on a real case, correct it once, then lock the instructions in and reuse it.
You will notice the difference on the second run, when you skip the briefing entirely. From then on, every run pays back the time you spent building the tool.
Markus Kopko is the founder of PMotion.ai and a project and program management consultant and trainer with more than 25 years of experience. He serves on the PMI Core Development Team for The Standard for Artificial Intelligence in Portfolio, Program, and Project Management. He will be presenting “From Prompts to PM Power Tools: Build Custom AI Tools for Recurring Project Work”, available on-Demand at IPM Day 2026.
