Your client says: "I don't know what to do."

And something in you lights up. You have ideas. You have experience. You can see the path they can't see. So you say: "Have you considered..."

You've just failed.

Not because your advice was bad. It was probably good. You failed because you stopped coaching and started rescuing — and neither of you noticed.

The Problem With Being Helpful

Eric Berne called them games — not because they're trivial, but because they're structured, repetitive, and invisible to both players. In Transactional Analysis, a game is a recurring pattern of social interaction with a hidden agenda and a predictable payoff. The person playing the game doesn't know they're playing. And critically: neither does the person who gets pulled in.

Coaches get pulled in through helpfulness. A client presents a problem. The coach offers a perspective. The client engages with it — then finds a reason it won't work. The coach tries another angle. Same result. By the end of the session, both feel like they've worked hard. Nothing has changed.

Berne called this one "Why Don't You — Yes, But." The client's hidden payoff: confirmation that the problem is unsolvable, which means they don't have to change. The coach's hidden payoff: feeling smart and useful, which means they don't have to sit with the discomfort of not knowing.

Three Games Coaches Don't Know They're Playing

In our work training coaches through AI-powered practice scenarios, we see the same patterns surface repeatedly. Three games account for the vast majority of coaching failures — and none of them feel like failures in the moment.

The Rescuer Game. The client asks for help directly: "What would you do?" The coach, trained to be helpful, gives advice. The client feels relieved. The coach feels effective. The client returns next session with the same pattern, because someone else did their thinking for them. The coach built no muscle — they provided a crutch.

The Collusion Game. The client describes a difficult colleague, an unfair board, an impossible boss. The coach says: "That sounds really frustrating." Empathy is an ICF competency, so this feels right. But validating the complaint without exploring the client's role in the dynamic isn't empathy — it's collusion. The client leaves feeling understood and completely unchanged. Stephen Karpman's Drama Triangle has three seats: Victim, Rescuer, Persecutor. The coach just took the Rescuer seat while the client played Victim and the absent colleague played Persecutor. Everyone has a role. Nobody grows.

The Seduction Game. Coach and client have a rich, stimulating conversation about leadership, strategy, or organizational dynamics. The coach feels intellectually engaged. The conversation is sophisticated. Both leave feeling it was a great session. But nothing personal happened. No pattern was surfaced. No discomfort was explored. The client's analytical mask — their Adult ego state performing competence — was never pierced. The game ran perfectly because it looked exactly like coaching.

What Makes This So Hard

The pull toward rescuing isn't a character flaw. It's structural. Coaches are selected for empathy, trained in active listening, and rewarded when clients feel helped. The entire feedback loop points toward helpfulness.

But coaching isn't helping. Coaching is holding space for someone to find their own answers — which often means sitting with their discomfort instead of resolving it. It means asking "What have you already considered?" when every instinct says "Have you tried..." It means noticing that you're about to give advice and choosing not to.

This is what Berne's framework reveals so clearly: the coach's ego state matters as much as the client's. When a client shifts into Adapted Child — helpless, seeking direction — the coach who responds from Nurturing Parent completes a complementary transaction. It feels natural. It feels warm. And it keeps both people exactly where they are.

The coach who responds from Adult — "You say you don't know what to do. I'm not sure I believe that" — creates a crossed transaction. It disrupts the pattern. It's uncomfortable. And it's where growth happens.

Why Reading About This Doesn't Help

Here's the uncomfortable truth: understanding Transactional Analysis doesn't make you better at resisting these patterns. You can read Berne, study Karpman, pass an exam on ego states — and still give advice the moment a client looks at you with genuine distress.

Research confirms this. Costello, Pennycook, and Rand (2024) demonstrated that practice changes behavior while passive information consumption does not. Knowing about the Drama Triangle and recognizing it in real time under emotional pressure are fundamentally different skills. The first is knowledge. The second is a trained reflex.

This is why we built The Helping Trap — a practice scenario where coaches work with AI-powered clients who are actively pulling them into games. Each client runs a different pattern. One asks directly for advice. Another builds an evidence-based prosecution and invites the coach to render a verdict. A third presents such compelling moral authority that agreeing with her feels like being on the right side of history. The pull is real, the emotional pressure is real, and the feedback shows you exactly where you got pulled in.

Try It Yourself

The Helping Trap is currently available for coaching programs, CCE workshops, and individual coach development. If you're interested in using it with your coaches or experiencing it yourself, reach out to us at hello@roleplays.ai.

References

Berne, E. (1964). Games People Play: The Psychology of Human Relationships. Grove Press.

Costello, T. H., Pennycook, G., & Rand, D. G. (2024). Durably reducing conspiracy beliefs through dialogues with AI. Science, 385(6714), eadq1814.

Karpman, S. (1968). Fairy tales and script drama analysis. Transactional Analysis Bulletin, 7(26), 39–43.

Shaikh, O., et al. (2024). Rehearsal: Simulating conflict to teach conflict resolution. Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems, 1–18. ACM.