Improve
This process works. So why does it still take so much effort?
Your organisation has processes, systems and agreements. Maybe you work with Jira, Confluence, SharePoint, Teams, ERP software or a quality system. Everything is digitised — and still people are searching for information, copying documents, retyping data and telling each other what the next step is.
With what AI can do today, you could automate such a process. But that doesn't mean you should automate the existing process.
First improve the process. Then make the improved process AI-native.
First understand what the process has to achieve
Coming from systems engineering, I don't just look at the existing procedure or software, but at what the process actually has to do.
What is the desired outcome? What information does that require? Who takes which decision? Where are the handovers? Which checks are genuinely needed? Which responsibilities must stay with people?
And above all: if we were to design this process today, knowing what people and AI can do together now, would we still organise it this way?
Among other things, we look at:
- roles and responsibilities;
- information and knowledge needed during the process;
- decisions and human checkpoints;
- unnecessary handovers and manual work;
- existing software and systems;
- quality, privacy, safety and regulation;
- what AI can support;
- what can be fully automated.
From procedure to working support
The result doesn't have to be yet another process document.
A well-designed process can become more and more executable. An employee then doesn't have to look up a procedure first and work out which steps are needed. AI can understand the context, ask the right questions, gather information, carry out work and support the employee in decisions.
People stay responsible where human accountability is needed. But they no longer have to do all the work around it themselves.
Example — corrective action
At an organisation using Jira and Confluence, starting a corrective action still took a lot of manual work: finding the right procedure and template, gathering information, filling it in, submitting and following up.
The alternative: an employee simply tells their AI assistant, “I have a corrective action.” The assistant knows the process, asks the necessary questions about risk and urgency, helps draft the corrective action and finally asks whether it may be submitted.
The process isn't just automated. The way the employee works with it changes.
Maybe the process itself doesn't need to be simpler. The work does.
The goal is not to hand as many steps as possible to AI. The goal is to design a better way of working in which people, AI and existing systems each do what they are good at.
Not ready for a call yet?
Send me one or two sentences about which process costs you unnecessary work.