From User to Designer: The Six Stages Your Team Must Take with AI

Almost everyone uses AI. That tells you almost nothing.

Treat AI capability as a continuous journey, not a destination. Users apply AI inside existing processes. Designers change the processes themselves. Between those poles lie six stages across three disciplines. Each stage moves you forward; none can be skipped. And there is no arrival point — the field evolves faster than teams can fully catch up.

Why I say this so plainly: over the last years I’ve taught AI to several hundred people. In large enterprises and in mid-sized companies, using Copilot, Claude, ChatGPT — with people who have prompted daily for two years and with people who had never opened a model. Across all those environments the challenge was the same. It wasn’t the tool, the model, or the prompt. It was the leap from output to outcome. Producing something with AI has become remarkably easy. Generating real value with it remains hard.

• 6 Stufen – Entwicklungspfad vom Anwender zum Gestalter
• 3 Disziplinen – Prompt, Context und Process Engineering strukturieren den Weg
• 2 Jahre – Teilnehmende, die seit zwei Jahren täglich prompten

Why most teams automate their past

Established processes are the norm. Almost every marketing team works in a chain like this: briefing, analysis, concept, review, presentation, handoff. Then AI enters, and each step speeds up. It feels like a restructure. In most cases it’s acceleration: the process stays exactly the same, only individual tasks run faster.

That’s not bad — in most cases it’s the right entry point. But it’s only the entry. Teams reliably move through the same phases: curiosity, individual acceleration, then a plateau. Everyone uses AI, everyone says it helps, and yet the department’s measurable outcomes don’t change. Only then comes the honest question: why does this task still exist at all?

An example from our everyday work: we consider building a complex AI process to auto-format an Excel sheet that travels by email through the team. The more honest question comes earlier: why are we emailing that sheet at all? The usual answer is: because we’ve always done it that way. A form that collects structured data replaces the whole step, and the formatting problem disappears with it.

From an Excel sheet to a form

It’s easy to be tempted to build a complex AI process just to auto-format an Excel sheet that gets emailed around the team.

A more honest question starts earlier: why are we circulating this sheet at all? Often the answer is: because we’ve always done it that way.

A form that captures structured inputs replaces the whole step. That also removes the formatting problem.

The path out of this pattern isn’t a single aha moment. It runs through six stages organized into our three engineering disciplines.

1. Prompt Engineering This is where the shift from asking to assigning begins: instead of vague requests, you hand over tasks with objectives, context, and expected outputs. At stage 2, AI doesn’t just advise — it produces concrete deliverables like presentations or analyses.
2. Context Engineering A system without context starts each task from zero; with connected knowledge it gains access to emails, calendars, documents, campaign data, and internal know-how. With memory and feedback, corrections, rules, and decisions become part of the context and generic outputs fall away.
3. Process Engineering Personal prompts become reusable workflows, skills, and playbooks with clear quality criteria. At the highest stage the process itself is re-cut: systems take on steps and humans decide where their judgment matters most.

Discipline 1: Prompt Engineering

Stage 1: Assign work instead of asking. Many people treat AI like a better search engine: question in, answer out — and if the answer is mediocre, well, AI can’t do better. The same manager would never tell a new hire “Create a marketing strategy” without goals, customer context, examples, or a definition of a good outcome. You reach the first stage when you stop asking AI questions and start assigning it work.

Stage 2: Deliverables instead of answers. Not “What might a presentation look like?” but “Create the presentation.” Not “Which metrics should I analyze?” but “Analyze the data and show the deviations.” Once AI actually produces work instead of only advising, the leverage changes noticeably.

Discipline 2: Context Engineering

Stage 3: Connect knowledge. A system without context starts each task at zero. At this stage it gains access to what your team uses: emails, calendars, documents, campaign data, internal knowledge. It knows what has already been decided, which campaigns are running, and which objectives were set.

Stage 4: Memory and feedback. Now corrections flow back into the system instead of the trash. Past decisions, feedback, and recurring rules become part of the context. This is where scattered knowledge becomes a maintained knowledge base, and where frustration with generic outputs drops significantly.

Discipline 3: Process Engineering

Stage 5: Build reusable systems. A good prompt that one person uses manually stays personal productivity. At this stage it becomes a workflow, a skill, a playbook used by the whole team, with fixed quality criteria and clear outcomes. In one case a person learned a trick. In the other, the organization built an asset.

Stage 6: Redesign work. The top stage questions the process itself. Which steps are still necessary, which can a system take over, and where must humans decide? Here a team stops accelerating its past and starts rebuilding its work. This is the designer level.

The climb doesn’t happen by observation. People who move through these stages fastest rarely master every new model. They have a lot of repetitions on real work: a difficult client briefing, a recurring report, a process that has been annoying for months. From that comes AI intuition — a sense of where AI is strong and where human judgment remains decisive.

Let the AI map your process with you

So how do you find out where your team stands and which step to take next? My answer is consistently AI-assisted. On the lower stages you can rarely judge which tasks a system can take on. You simply lack the experience of what’s possible today.

Turn it around. Give the AI your process and your real work artifacts: the latest briefings, reports, or newsletters and the steps between them. Let the system interview you about your process. Then ask for recommendations: which steps can be largely automated, where does AI support a person, where must the decision remain with you, and which steps might no longer be necessary? You evaluate the suggestions and you decide. But the map is drawn by the system, not by your gut on stage one. This exercise is also the fastest way to understand several stages at once.

Why you can’t outsource this redesign

This work is change work: questioning processes, redefining roles, guiding people through uncertainty. That’s traditionally something organizations ask outsiders to do. In this case that is the wrong approach. AI must enter the core of value-creating processes, and that core can’t be redesigned from the outside because the knowledge of why a process looks the way it does lives in the team. Whoever outsources the redesign gets a target picture and retains the problem. External support should therefore look like this: not someone who does it for you, but someone who does it with your team so the capability stays in-house.

Frequently asked questions about the six stages with AI (FAQ)

How can I tell which stage my team is at?

Observe whether AI only advises or already produces work, whether knowledge is connected systematically, and whether reusable workflows exist. The more steps the system takes independently and the clearer roles and quality criteria are, the higher the stage.

What separates acceleration from genuine redesign?

Acceleration speeds up existing steps without changing the flow. Genuine redesign questions the process itself and restructures work so some steps disappear or are taken over by systems.

What first steps help when starting Context Engineering?

Begin with the sources your team uses daily: emails, calendars, documents, campaign data, and internal knowledge. The crucial factor is that corrections and feedback flow back and become part of a growing memory.

How do I use Prompt Engineering correctly?

Provide clear goals, examples, and quality criteria. Assign concrete work instead of asking general questions. Request deliverables like drafts, analyses, or presentations and iterate using real work artifacts.

Should we outsource the redesign to external partners?

The core redesign belongs with the team, because the essential process knowledge is internal. External partners can support by working with the team and building capability, not by defining the process from the outside.

Conclusion

No matter how fast you move, you must adapt your value-creation processes yourself and stop applying AI only to your existing workflows. That will be the defining line in the coming years: not between people who use AI and those who don’t, but between users and designers. Each of the six stages brings you closer to that point, and each begins with real work rather than the next tool. The honest questions at the end: which stage is your team really at? And which one will you take on next?