--- title: "Never Stop Learning: agency" client: "Atlas Copco Group" language: en date: "2026-10-12" version: "1.0.4" canonical_url: "https://20261012.jardenberg.net/agency.html" --- 01 / Agency and value # I can change how I work. Karin's first discovery was bigger than a faster attendance list. ## Karin stopped waiting. A Teams attendance report. A Word document. Copy a name, paste a name, go back for the next one. Repeat after the next meeting. In the session, I demonstrated the routine Karin, an executive assistant at Volvo Group, wanted to change. She gave Copilot the text and asked for a bullet-point list containing only the names. Then she asked for the email addresses in a format she could paste into Outlook. A frustrating sequence of manual steps became a conversation about the result she actually needed. That was the first change. The next was to combine the instructions into a prompt she could reuse. Then to work from a file rather than copying its contents. Then to give a recurring task to an agent. She learned one useful step at a time and shared what worked with her colleagues. The important word is *agency*. She could influence her own working day. You do not have to become a software engineer to begin asking whether there is a better way. - #### Try one clear request. Describe the result you need. Work with a small example you are allowed to use. - #### Improve it together. Ask for another useful format. Check whether the answer really helps the next step of the job. - #### Make the instruction reusable. Ask AI to capture what worked. Test it on a fresh example before relying on it. - #### Share the discovery. A colleague's variation can make the method useful beyond your own desk. An agent may be a later step if the task and available tools warrant it. The first request ```text From the text below, give me a bullet-point list of the participant names only. Do not invent missing names. If an entry is ambiguous, flag it separately. [Paste a small, permitted example here.] ``` Make it repeatable ```text Turn the instructions that worked in this conversation into one reusable prompt. Mark where I should insert the next input. State the output format and the checks I should perform. Keep it short enough to use again. Then help me test it with a different example. ``` Karin's story is my account of the work. Exact Copilot features and file access depend on the account and environment you use. The useful progression also works with other AI tools. The six value vectors ## What becomes possible when the friction goes away? Time saved is a useful beginning. It is not the whole account. Karin also gained confidence, more attention for meaningful work and something she could teach someone else. These six lenses help you look for the value that a stopwatch can miss. 01 #### Compliance and risk mitigation. Make checks easier to perform and exceptions easier to see. Does the new approach help you notice something important before it becomes a problem? 02 #### Organisational memory and knowledge. Find and use what is already known. Can you recover a previous decision, its source and the reasoning behind it? 03 #### Decision velocity and quality. Explore alternatives and improve the basis for a decision. Did you move forward with a clearer understanding of the trade-offs? 04 #### Cognitive capacity and wellbeing. Reduce repetitive effort, searching and unnecessary context switching. What attention or energy can you now put somewhere more valuable? 05 #### Learning and talent development. Use AI to practise, understand, get feedback and share knowledge. What can you or a colleague now do that was previously out of reach? 06 #### Access and inclusion. Lower barriers to participation, language and expression. Who gets to contribute when a difficult first step becomes easier? Which kind of value matters in the problem you want to work on? This is a working framework for asking better questions, not a measured allocation of AI's value. Percentages and broad industry estimates mentioned during the talk are not needed to use it. From thinking to trying ## Make the first step small. ![Conceptual illustration: fear rises while thinking about a change and declines after starting to act.](https://20261012.jardenberg.net/assets/fear-and-action.webp) The illustration used in the session is a metaphor, not a measured psychological curve. An unfamiliar way of working can feel unsettling. Trying one small, reversible task gives you something concrete to learn from. You do not have to settle every question about the future of AI before testing whether it can help you with this afternoon's work. Start with the [Painkiller Protocol](https://20261012.jardenberg.net/painkiller.html). It helps you choose a manageable problem before rushing to a solution. Prepared by AI + JJ. Based on the delivered keynote, final slides and reviewed source material; this is an edited learning companion.