Most leaders are not held back by effort or talent. They are held back by where they spend it. Stuck doing the work, answering, executing, fixing, they never reach the work of designing it. The research is direct: AI used poorly accelerates the wrong work; AI used well accelerates the right work1. The pause is what aims it. What follows is the practical move out.
Two Different Questions: Effective vs. Efficient
Most leaders ask the efficiency question first, “How do I do this faster?”, when the harder, more valuable question is the effectiveness question. AI helps with both, but only once you know which one you are solving. April 2026 HBR research finds executives and frontline managers diverge precisely here: executives experience AI as strategic advantage; managers confront its limits inside real workflows2.
• Effective. Doing the right things: judgment, prioritization, design. Ask, should I be doing this at all? Who else could? What is the work behind the work?
• Efficient. Doing things the right way: speed, leverage, automation. Ask, once the right work is named, how can AI help me do it faster and sharper?
Two Operating Principles
1. Interrupt first. Accelerate second. AI used poorly accelerates the wrong work. The pause finds the right work. AI then helps you do it faster and better than you could alone.
2. Do not go alone. Adaptive work has never been solo. Bring a partner: a peer, a confidant, a thinking companion3 4. Interruption is stronger when it comes from another human as well as from AI.
AI Is No Longer Just a Tool
Heimans and Timms5 argue the dominant mental model, AI as instrument, is no longer accurate. They describe today’s systems as “autosapients” with four defining traits: agentic (they plan and act), adaptive (they discover strategies beyond their programming), amiable (they simulate empathy in ways that cultivate dependency), and arcane (they are black boxes, even to their designers). The November 2025 MIT/BCG global survey found 76% of organizations already view agentic AI as more like a coworker than a tool6. The leadership practice this demands is duet and doubt: collaborate as with a capable colleague, while maintaining principled skepticism about its most confident outputs.
Five Mindset Shifts: Frame, Think, Learn, Act, Collaborate
Park, Baer, and Nickerson7 identify framing as a learnable cognitive skill and the highest-leverage place AI either helps or harms your leadership. Quantum Bridge Solutions (QBS) offers the following five “mindset shifts” to help your leadership:
• Frame. From “I bring my problem to AI” to “I interrogate my framing first; AI challenges the question.” Ask for three competing framings before solving.
• Think. From “I know the answer; let me execute” to “AI helps me see what I am missing.” Use AI as a thinking partner, not a confirmation engine.
• Learn. From “I apply best practices” to “I treat decisions as hypotheses.” Use AI to synthesize patterns and run cheap pilots in dialogue8.
• Act. From “I solve on the dance floor” to “I get on the balcony, design the intervention, and build others’ capacity.” Once the right action is chosen, AI accelerates drafting and scale9.
• Collaborate. From “I use AI privately” to “I co-create with humans and AI and build the ecosystem that makes design work safe.” AI alone cannot share identity change.
The Designer’s Ladder: A Weekly Practice
Better thinking does not come from more thinking. It comes from interruption10. The four-rung ladder maps to the QBS H-AI-H model: Human Inquiry, AI Processing, Human Empowerment.
1. Notice. Mark last week’s calendar blocks as D for doing or d for designing. Notice the ratio without rushing to fix it.
2. Name. Without AI, name the role you played. Was this technical (a known answer) or adaptive (no known answer)11?
3. Reframe. Bring it to AI not for the answer, but for the framing. Ask for three different ways a designer would see this, including a stakeholder you have not named12.
4. Redesign. Pick the framing that interrupts you most, not the most efficient. Choose one designer move this week. Then act, with AI accelerating the doing.
Five Traps to Watch For
QBS cautions leaders to avoid these five “actor traps” and focus on becoming a more effective leader:
• Efficiency-shopping the wrong problem. Ask AI to make a meeting more efficient and you may never ask whether the meeting should exist. Effectiveness comes first.
• Anchoring on your own frame. AI answers within the premise you provide13. Frame-challenging prompts must be deliberate.
• The narrow cognitive funnel. When everyone uses the same AI to prep the same meeting, inputs converge before the conversation begins14. Protect divergent sources.
• Treating AI as the only partner. AI cannot validate experience or share the burden of identity change. You still need a confidant, a sanctuary, and regular practices15.
• The lump-of-task fallacy. Squeezing more from current workflows misses the larger move: redesigning what work is, who does it, and what becomes possible for the first time.
What Stays Human
AI amplifies design capability: analysis, synthesis, drafting, scenario-building. It cannot decide for you. Mission, judgment, ethical trade-offs, accountability, and the trust you hold with those you serve remain yours16. The Lead Designer’s highest act is not what they do, but what they make possible for others.
A Final Word
QBS recognizes the “actor trap” is the most common identity trap in AI leadership today. AI is making it both more dangerous and more solvable: more dangerous because acceleration without framing scales the wrong work fast; more solvable because, used well, AI is the productive interruption that surfaces the designer’s question. Become more effective and more efficient by asking the right question first, naming where you are operating, and choosing one designer move this week. QBS partners with leaders navigating exactly this AI work.
References
- Tankelevitch, L., Kewenig, V., Simkute, A., Scott, A. E., Sarkar, A., Sellen, A., & Rintel, S. (2024). The metacognitive demands and opportunities of generative AI. Proceedings of the CHI Conference on Human Factors in Computing Systems (CHI ’24). ↩︎
- Harvard Business Review. (2026, April). Managers and executives disagree on AI, and it’s costing companies. ↩︎
- Heifetz, R. (2025). On the keys to adaptive leadership [Interview]. In Leading in the age of AI. Charter × Council Advisors. ↩︎
- Daloz Parks, S. (2005). Leadership can be taught: A bold approach for a complex world. Harvard Business School Press. ↩︎
- Heimans, J., & Timms, H. (2024). Leading in a world where AI wields power of its own. Harvard Business Review, January–February 2024. ↩︎
- Ransbotham, S., et al. (2025, November). The emerging agentic enterprise. MIT Sloan Management Review × BCG. ↩︎
- Park, C. H., Baer, M., & Nickerson, J. (2025). Looking at the trees to see the forest. Organization Science. ↩︎
- Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at work. Quarterly Journal of Economics, 140(2), 889–942. ↩︎
- Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative AI. Science, 381(6654). ↩︎
- Nemeth, C. J. (2018). In defense of troublemakers: The power of dissent in life and business. Basic Books. ↩︎
- Heifetz, R., Linsky, M., & Grashow, A. (2009). The practice of adaptive leadership: Tools and tactics for changing your organization and the world. Harvard Business Press. ↩︎
- Steyvers, M., & Kumar, A. (2024). Three challenges for AI-assisted decision-making. Perspectives on Psychological Science, 19(5), 722–734. ↩︎
- Steyvers, M., & Kumar, A. (2024). Three challenges for AI-assisted decision-making. Perspectives on Psychological Science, 19(5), 722–734. ↩︎
- Heimans, J., & Timms, H. (2024). Leading in a world where AI wields power of its own. Harvard Business Review, January–February 2024. ↩︎
- Heifetz, R. (2025). On the keys to adaptive leadership [Interview]. In Leading in the age of AI. Charter × Council Advisors. ↩︎
- Hao, X., Demir, E., & Eyers, D. (2025). Beyond human-in-the-loop: Sensemaking between AI and human intelligence collaboration. Sustainable Futures, 10. ↩︎
