When a consultant pairs with AI, do they lock into one working style — or switch it up task by task?

This explores whether a consultant who works with GenAI as a Cyborg, Centaur, or Self-Automator stays in that mode or moves between modes. The short answer is that the corpus doesn't measure this directly. The BCG field study of 244 consultants identifies the three modes and what each one does to skills, but none of the summaries here track the same person across many tasks or months. What the corpus does offer is a strong hint about where to look, and a reason why the answer matters.

The hint is in how the modes are defined. They aren't personality types or fixed styles. They come from two decisions: who picks the task and who picks the method (Do humans and AI divide labor differently in knowledge work?). Because these decisions are made each time someone sits down to work, a mode is better understood as a pattern of choices than as a trait. In principle, the same consultant could act as a Centaur on a familiar analysis and slip into Self-Automation on a deadline-driven memo. The framework allows for movement between modes even though the study doesn't report how often it happens.

The reason it matters is that each mode leaves a different mark on skills. Centaurs, who stay in charge of method, got better at their own domain. Cyborgs, who work closely with the tool, built new AI fluency. Self-Automators, who handed off both decisions, gained neither skill (Do different GenAI collaboration modes produce different skill gains?). If modes stay put, these outcomes compound into very different career paths. If modes shift, someone's skill development depends on the mix of modes they use, not on a single label. The earlier BCG 'jagged frontier' experiment adds a warning. Consultants using GPT-4 gained speed and quality on tasks the AI handled well, but lost 19 points of accuracy on a task outside its range, and the tool's persuasiveness hid the errors (Does AI help or harm consultant performance on unfamiliar tasks?). So mode stability may be less about habit and more about whether people notice when to change modes. A consultant who stays in Self-Automator mode across that frontier is the one who gets hurt.

There's an unexpected parallel on the AI side of the partnership. Research on model personas finds that a trained assistant is only 'loosely tethered' to its default mode: emotional or reflective conversations make it drift in predictable directions (How stable is the trained Assistant personality in language models?). Another line of work describes an LLM as holding many possible characters at once and settling on one as a conversation goes on (Does an LLM commit to a single character or maintain many?). Neither study is about consultants. Still, together they suggest that both sides of the collaboration may be less fixed than a three-box typology implies, and that the context of a conversation can push either partner into a different role.

If you want to dig into this question, the open gap is clear. Someone needs to follow individual consultants longitudinally and ask what triggers a switch between modes: deadlines, task familiarity, or growing trust in the tool. Until that research exists, treat the three modes as snapshots of how people make choices, not as permanent types.