Automate, collaborate, orchestrate (and how to tell which)

Three Ways to Work With AI

Once you have decided that AI should be involved in a task — whether that is Quadrant 2 (Collaborate) or Quadrant 3 (Automate) — you need to decide how you will work together. There are three distinct collaboration modes, and choosing the right one makes a significant difference to your results.

Mode 1: Automate

What it means: You define the outcome clearly, set the scope, and let AI execute. You review the output at the end.

When to use it:

• The task is well-defined with clear success criteria
• The output is relatively straightforward to verify
• Speed and consistency matter more than nuance
• You have done this type of task before and know exactly what good looks like

What it looks like in practice: You give the AI a specific instruction — "Summarise this document in 200 words, focusing on financial implications" — and it produces the output. You read it, check it against the source, make minor edits if needed, and move on. The interaction is typically one or two exchanges.

The risk: If your instructions are unclear or your quality check is superficial, you may end up with output that looks polished but misses the point. Automation mode requires clear upfront specification and honest verification.

Mode 2: Collaborate

What it means: You and AI work together iteratively. You contribute your expertise, AI contributes its processing power and breadth, and the output emerges from the back-and-forth.

When to use it:

• The task is complex or ambiguous
• You are not entirely sure what the best approach is
• The quality of the output depends on nuance and judgement
• You want to think more broadly than you would alone

What it looks like in practice: You might start by describing the problem and asking AI for initial thoughts. You react to its response — "That is interesting, but you are missing this constraint" or "Expand on that second point." You go back and forth, refining ideas, challenging assumptions, and building towards an output that neither of you would have produced alone. This might involve five, ten, or twenty exchanges.

The risk: Collaboration mode can become a time sink if you do not have a clear sense of when to stop iterating. Set yourself a time limit or a "good enough" threshold before you begin.

Mode 3: Orchestrate

What it means: You set the parameters, define the guardrails, and let AI operate with a degree of independence. You check in at key points rather than being involved in every step.

When to use it:

• The task involves multiple steps or components
• You trust the AI's ability to handle the domain
• You have clear criteria for what acceptable output looks like
• Your time is better spent on higher-value work

What it looks like in practice: You might instruct AI to analyse a dataset, identify the top five trends, generate a summary for each, and produce a one-page brief with recommendations. You give it the context and constraints upfront, then review the final output rather than guiding each step. In more advanced setups, this might involve AI agents that execute multi-step workflows with minimal supervision.

The risk: Orchestration requires the highest level of trust in both the tool and your own ability to evaluate the output. If you cannot reliably tell whether the AI has done a good job, orchestration mode is premature for that task. Start with collaboration mode and move to orchestration as your confidence grows.

Matching Mode to Task and Confidence

The right mode depends on two things: the nature of the task and your confidence level.

If the task is straightforward and you know the domain well, automate. You will save the most time with the least risk.

If the task is complex and you want to think it through, collaborate. The iterative process will produce better results than a single-shot instruction.

If the task is multi-step and you trust both the tool and your ability to evaluate its work, orchestrate. This is where you start to unlock genuinely significant productivity gains.

And if you are unsure? Start with collaboration mode. It gives you the most visibility into what the AI is doing and the most opportunity to course-correct. You can always move towards automation or orchestration as you build confidence with a particular type of task.