Designing AI collaborators
Who would you want on your team?
The dream-team question
I have worked as a solo consultant for years. Independence brings freedom, but sometimes you miss having a team around you: people who notice different things, challenge your judgement and contribute capabilities you cannot bring equally strongly to every piece of work.
That led to a simple question:
If I could design my own dream team, who would I want on it, and what would each person contribute?
I have created a small team of AI collaborators, each designed to add something specific: creative audacity, editorial discipline, commercial advocacy or a stronger instinct for professional possibility.
Mine are mainly work-related, but the same approach could be used more broadly. Someone might design collaborators to support learning, creative practice, planning, decision-making or other parts of life. The starting point is the same: understand the person, then identify the kind of support that would genuinely extend their range.
A working experiment in designing co-intelligence
This project is part of my wider work on designing co-intelligence.
Designing useful co-intelligence starts with understanding the shape of your own intelligence. What do you already do well? What do you tend to overlook, avoid or postpone? Which kinds of challenge help you think better? What does the work need that you do not always supply as readily yourself?
Those questions make it possible to design collaborators around a real need rather than a generic idea of helpfulness.
Inner Magpie, my private system for preserving self-knowledge across time, gives that inquiry continuity. It keeps selected observations, dated context and developing interpretations available without treating any one account as a finished profile.
Why create several collaborators?
A general assistant can adopt many roles, but the distinctions between them can easily blur.
Separate collaborators preserve useful differences. They have different remits, perspectives and standards. They do not need to agree because their purpose is to contribute different kinds of judgement.
I have deliberately designed productive incompatibility into the team. The value comes partly from seeing where their judgements diverge, rather than collapsing every perspective into one helpful average.
Why use fictional characters?
A fictional character acts as a compressed design specification.
It is easier to imagine, test and refine a collaborator when starting from a coherent, familiar character than from a blank list of traits.
A developed fictional character already carries a recognisable combination of motives, habits, contradictions, values, social role and ways of responding to other people. That provides a richer starting point for design. I then define the collaborator’s purpose, remit, boundaries and working relationship with me.
The character makes the collaborator imaginable. The additional design makes them useful.
The team
| Oliver from Strategy, Strategic Thinking Partner | Helps me see credible possibilities beyond the ones I can currently imagine, especially when I am underestimating my own range. |
|---|---|
| Don Draper, Creative Director | Finds the human idea inside complicated thinking and commits to a creative direction I may not have considered. |
| Elizabeth Zott, Writing and Reasoning Partner | Tests the reasoning, makes the prose clearer and preserves complexity without patronising the reader. |
| Susie Myerson, Talent Agent | Finds opportunities in work I already have, argues for the right ones and protects my time, reputation and commercial interests. |
The design process
Each collaborator is designed by asking four questions:
- What does the person already bring?
- What capability or perspective would extend their range?
- What kind of character could embody that contribution coherently?
- What purpose, boundaries and working relationship would make the collaborator useful?
The result is a deliberately assembled team whose members bring different minds to different kinds of work.