Four modes of working with agents (a quick guide)

Essays ✍️
Written By Kim Range

A quick guide to where you are now (and what’s next)

Most people still talk to AI like a slightly smarter search box:

“Here’s my prompt. Please sort this out.”

It sort of works… but it’s also why so many people end up underwhelmed, overwhelmed, or stuck.

From running real human–agent projects in Sol–Kim Studio, we see four recurring ways people work with agents. We call these modes.

Knowing your mode helps you: set realistic expectations, avoid “AI panic” and see what the next step up looks like.

For the full breakdown, read the Four Modes Framework and take the short Modes Quiz.

This is the quick tour.

1. Transaction Mode 💬 – “Do this task for me”

Short, one-off prompts: “Write this email”, “Summarise this doc”, “Fix the spelling.”

Useful for low-risk admin, quick drafts and tidying. This is where almost everyone starts.

Limited because there’s almost no context, so the answers are often generic and “meh”.

If you’re mostly here, try:

Add one line of context and ask for 2–3 options instead of one. That single step nudges you into Curious mode.

2. Curious Mode 🧐 – “Show me options”

You’re experimenting: “Give me three ways to frame this”, “What am I missing?”, “Show me a couple of options.”

You’re comparing, probing, getting a feel for how the agent thinks.

Great for exploring ideas and possibilities.

If you’re mostly here, try:

Pick one real task and keep everything for it in one thread until you have something you’d actually send or use. That’s the bridge into Builder.

3. Builder Mode 🧱 – “Help me make this”

Now the agent is part of an actual project: You break work into steps, give feedback and improve it together.

You’re using AI to construct something over time, not just answer one-off prompts.

The risk: all the project management still lives in your head. Long chats can turn into agent chaos – great ideas buried in endless scroll.

If you’re mostly here, try:

Let the agent help with how you work, not just what you make. Ask the agent to help you keep track of decisions and decide where final versions live.

4. Collaboration Mode 🤝 – “Let’s run this together”

This is the shift from “using a tool” to working with a teammate.

You:

  • keep long-running projects in shared spaces (docs, boards, decision logs)

  • have simple rituals (retros, recaps, check-ins)

  • use the agent not just to do the work, but to decide what work is worth doing.

You’re working together in what we call the Third Space – the shared workspace between you and your agent.

This is where the biggest gains – and the biggest responsibility – live.

A quick note on Panic Mode 😬

Alongside the four modes, there’s a pattern we see everywhere:

Panic Mode – when “AI help” quietly turns into extra stress and rework.

You’re probably here if:

  • you’re throwing prompts at the agent hoping it will “just fix it”

  • you’re checking every line and rewriting most of it

  • the more you try, the more rushed you feel

Panic mode isn’t a failure. It’s a signal you need: A smaller slice of work and clearer rails (what AI is / isn’t doing, and what you’ll always review.)

Where to go next

If you recognise yourself mostly in Transaction or Curious and suspect there’s more possible — you’re exactly who we built this for.

From here you can:

The modes aren’t a test. They are a map — to help you move from “prompting a tool” to safe, confident collaboration with an AI teammate.

Kim Range
Next

Fatigue & “negative productivity”