Idris Bamigboye, who manages a customer success team at a mid-market software company, once ran a brainstorm he was proud of. Sixteen people, two hours, sticky notes everywhere, real energy in the room. Everyone left buzzing. Three weeks later a teammate asked him, half-joking, "Whatever happened to all those churn ideas?" Idris did not have an answer. The session had been fun and produced nothing. Worse, his quietest two analysts, the ones who actually talked to churning customers every day, had barely spoken while the loudest voices filled the air. This lesson is about how Idris turned his brainstorms from theater into a process that produces real ideas and moves work forward, using AI to handle the structure so he could focus on the people.

What This Lesson Covers

A workshop or brainstorm is a session designed to generate ideas and think through a problem together. Run well, these sessions surface options you would never reach alone and build genuine buy-in. Run badly, they reward the loudest person, lose the best ideas, and end with nothing happening.

You will learn to use AI for the mechanical parts: designing an agenda with a logical flow, generating prompts that provoke real thinking, surfacing hidden assumptions, and synthesizing a pile of raw ideas into themes and next steps. What AI cannot do is the human work at the center: creating the safety that makes people share half-formed ideas, reading the room, and deciding which ideas are worth pursuing. That stays with you.

Why Most Brainstorms Fail

Idris's failed session had three classic problems, and they are the three that sink most brainstorms. Dominant voices took over, so he heard the most confident people rather than the best ideas. Ideas were never properly captured, so good thinking evaporated. And nothing happened afterward, so the team learned that brainstorming is a performance with no consequences.

Creative sessions are worth getting right because they do real work: generating options when a problem is genuinely complex, building alignment before you commit, exploring possibilities while it is still cheap to explore, and breaking a team out of incremental thinking. AI helps on the design and synthesis side. It will draft you an agenda, generate provocative prompts, and find themes across forty raw ideas in seconds. It will not make a junior analyst feel safe enough to disagree with you. That is the part only you can do.

AI can structure the thinking and capture the output. It cannot create the safety that makes people share real ideas, or make the judgment call about which ideas to pursue. Those are yours.

Brainstorm Versus Decide

The single most common confusion in these sessions is mixing two different activities. Brainstorming generates options and explores possibilities, with judgment suspended. Deciding chooses among options, evaluates them, and commits. Running them at the same time is the fastest way to kill a session: the moment an idea gets critiqued, people stop offering new ones.

Idris now states the mode explicitly at the top of every session. "For the next twenty-five minutes we are only generating. No idea is too wild, no critique allowed. We evaluate later." When people know which mode they are in, they stop self-censoring during the generative phase and stop feeling like the session is dragging during the evaluation phase.

Psychological Safety Is the Foundation

People only do real creative thinking out loud when they feel safe to share half-baked ideas, disagree with the boss, build on what others said, and admit "I do not know." Without that safety you get a performative brainstorm where people repeat back what they think you want to hear. That is worse than no brainstorm, because it looks like agreement.

Safety is built through small, deliberate moves: separating idea from owner ("here is an idea" lands more openly than "I have an idea"), insisting on building rather than critiquing during generation, and using a parking lot so that off-topic ideas get captured somewhere instead of being shut down. These are facilitation choices a manager makes in the moment, and they are exactly the choices AI cannot make for you.

Managing the Dynamics in the Room

The core facilitation challenge is that brainstorms naturally surface whoever is most comfortable speaking, which is rarely the introverts or the junior people, and rarely the people closest to the actual problem. Idris's most effective single change was the silent brainstorm: before any discussion, everyone writes their ideas individually, one idea per card. This alone typically doubles or triples the number of distinct ideas and pulls thinking from people who would never have grabbed the floor.

His other standing moves: round-robin sharing so every person contributes at least once, an explicit invitation to the quiet ("Wen, you have been quiet, what is on your mind?"), and inviting skeptics on purpose ("what would make this idea fail?"), which turns disagreement into something productive instead of something suppressed.

A Worked Example: The Churn Workshop, Done Right

Idris got a second chance at the churn problem. He pulled together twelve people across product, support, sales, and customer success, blocked two hours, and set a clear target: leave with five to ten promising ideas, each with an owner. This time he used AI to design the session and to score the output.

He prompted the AI with the attendees, the goal, the time, and the desired outcome, and asked it to propose a structure and prompts. He adapted what it returned into this agenda:

• Problem framing (10 min). Idris presents the churn data plainly: 14 logo losses last quarter against a usual 8 to 9.
• Silent root-cause brainstorm (15 min). Everyone writes, in silence, why they think customers are leaving. One reason per card.
• Cluster (10 min). Group the cards into themes on the wall.
• Discuss top causes (15 min). Which themes are most important and most fixable?
• Solutions brainstorm (25 min). For the top three causes, generate fixes. Still no critique.
• Prioritize by vote (15 min). Each person gets three dot votes to spend on the ideas they find most promising.
• Next steps (10 min). Assign an owner and a check-in date to each surviving idea.

The silent phase produced 34 cards. Clustering collapsed them into five root-cause themes, of which the group ranked three as both important and fixable: a rough onboarding handoff, slow support response on a specific feature, and a pricing tier that trapped growing customers. The solutions brainstorm produced 19 ideas across those three. Then came the dot vote: twelve people, three votes each, 36 votes total. Here is how the top of the board scored:

• Assign a named onboarding owner per enterprise account: 9 votes.
• Same-day SLA on the high-friction feature's tickets: 7 votes.
• Mid-tier pricing plan to stop the upgrade cliff: 6 votes.
• Proactive 30-day health check call: 5 votes.
• Self-serve knowledge base for the top five issues: 4 votes.

The remaining 14 ideas split the last 5 votes and went to the parking lot rather than the bin. The dot vote did something important: it gave Idris a defensible, transparent ranking that the loudest person in the room could not override, and it surfaced that the onboarding idea, which had come from one of his quiet analysts during the silent phase, was the clear favorite. That idea would never have surfaced in his first, free-for-all session. Each of the top three got an owner and a two-week check-in before anyone left the room.

A Note on Strategy Sessions

The same mechanics scale to bigger, longer sessions, for instance a half-day to rethink a team's direction at an inflection point. A workable shape: open with context, get individual written thinking before any group talk, explicitly challenge the assumptions you have been operating on, generate possibilities, cluster them, evaluate against criteria, and close on next steps. The one move that matters most at this scale is the assumption challenge. Idris used a simple AI prompt to prepare it: "Here is what my team currently assumes about how our customers want to be supported. For each assumption, give me one reason it might no longer be true." The AI returned five challenges, three of which were obvious and two of which were genuinely unsettling, and those two became the most productive twenty minutes of the session.

For the evaluation phase he used a small scoring grid rather than a free debate. Each surviving theme got rated one to five on three criteria, strategic fit, team capability, and feasibility, by every participant, and the averages gave a starting order to discuss. The grid is not a verdict; it is a way to make the conversation specific. A theme that scores high on fit but low on feasibility tells you exactly what the conversation needs to be about.

One caution for a manager: setting the company's strategic direction is a leader's call. Your version of this session is about your team's operating choices, how you work, what you prioritize, where you focus, and feeding well-structured options upward to the people who own the broader strategy. Running a sharp session that hands leadership three evaluated options is more useful to them than running one that quietly tries to set direction above your level.

Synthesizing Ideas Into Action

The phase that separates a useful workshop from a fun one is synthesis: capturing every idea, finding the themes, grouping them, deciding what is worth exploring, and naming next steps. This is where AI genuinely shines. Idris photographs the clustered cards, types the raw list into the AI, and asks it to group them into themes and flag duplicates and overlaps. In a couple of minutes he has a clean, organized summary that would have taken him an evening to assemble by hand, which means it actually gets done instead of sitting in a pile of photos on his phone.

Two cautions keep this honest. First, the AI groups by surface similarity, so it will sometimes merge two ideas that sound alike but mean different things, or split one idea that was phrased two ways. Idris always reads the grouping against his memory of the room and re-sorts a card or two. Second, synthesis is not the same as deciding. The AI can tell him "these nineteen ideas fall into four themes." It cannot tell him which theme is worth his team's next month. He uses the dot-vote results and his own judgment for that, and he is careful never to let a tidy AI summary quietly make a decision he should be making himself.

Follow-Up and Commitment

Ideas without commitment produce nothing, and nothing is what teaches people that brainstorming is theater. Every session Idris runs now ends with explicit answers to four questions: what are we doing with these ideas, who is exploring each one, by when, and what is the decision point. Within one day he sends a recap, the root causes discussed, the top ideas, who owns what, the check-in date, and a link to the full notes. Crucially, he also reports back later, even on ideas that failed: "We tried the same-day SLA. Here is what we learned." Closing that loop is what convinces people their next idea is worth offering.

Where This Goes Wrong

The fake brainstorm. You have already decided, but you run a session to make people feel heard. They notice, every time, and they disengage from future sessions. Only brainstorm when you are genuinely open. If you have decided, say so: "Here is my thinking. I want your questions, concerns, and ways to improve it." That is an honest, different conversation.

No follow-up. Great energy, fun ideas, then silence. People become cynical and stop engaging. The fix is the four follow-up questions and the recap within a day, every time.

Dominant voices. The loud and senior crowd out the quiet and junior, and you end up with the most confident ideas rather than the best. Silent brainstorm first, round-robin sharing, explicit invitations to the quiet.

Mixing brainstorm and critique. An idea gets shot down, the room goes silent, and thinking stops. Separate the stages completely, use "yes, and" instead of "but" during generation, and defer all judgment to the evaluation phase.

Ideas that disappear. You are busy facilitating and note-taking falls away, so good thinking is lost and people feel unheard. Assign someone to capture every idea as it surfaces, not at the end, and summarize within a day.

Human Judgment Checkpoints

Across any session, these are the moments where your judgment, not the agenda or the AI, has to take over:

• Safety. Are people sharing real ideas, or performing for you?
• Representation. Are all voices being heard, or just the dominant ones?
• Mode. Is this actually the moment to brainstorm, or should you be deciding? Do people know which it is?
• Execution. Should you let an idea be critiqued now, or keep generating? Does the energy still feel safe?
• Follow-through. Can you genuinely follow up on these ideas, or are you setting up another round of theater?

Practice and Reflection

Idris improved by examining his own sessions honestly rather than by reading about facilitation. Use these exercises on the sessions you have already run and the one you will run next.

• Assess the safety in your last session. Did people share half-formed ideas, or did everything that got said sound finished and safe? Did anyone disagree with you out loud? If the answer is no, assume self-censorship rather than agreement.
• Do a representation check. Write down who spoke in your last brainstorm and who did not. Then ask whether the silent people were the ones closest to the problem, as Idris's two churn analysts were. If you cannot remember who spoke, that is itself the finding.
• Name the mode before your next session. For your next planning conversation, decide whether it is a brainstorm, where you explore options, or a decision, where you choose among them. Then check whether the people attending know which one it is. If they do not, you are the only person in the room who understands the meeting.
• Audit your follow-up. Pick a brainstorm you ran at least a month ago. What actually happened to the ideas? Did any get explored? If you asked the participants to rate your follow-up, what score would you get, and what would they say about why?
• Test the silent brainstorm. Run one session where everyone writes individually before any discussion, and count the distinct ideas. Compare that against a session you ran without it. How many more ideas surfaced, and how different were they from what the confident voices would have offered?
• Facilitate for real. Design and run a full session on something your team genuinely cares about. Use the silent phase, capture every idea as it appears, use a transparent ranking such as dot voting, and follow up on the top ideas with named owners and dates. Then compare the outcome to Idris's first session, the fun one that produced nothing.

Related Lessons

Facilitation connects to several neighbouring skills, and these lessons extend what you have just read.

• Advanced Meeting Management is the broader craft that workshops sit inside. A brainstorm is a particular kind of meeting with particular rules, and the general habits of agenda design, timeboxing, and capture apply here too.
• Structuring Complex Decisions picks up exactly where a brainstorm ends. The ideas Idris's dot vote surfaced become the input to a decision, and that lesson covers how to evaluate options rigorously once the generating phase is genuinely over.
• Presentation and Narrative Building matters for what happens after the session. The output of a workshop usually has to be presented to someone who was not in the room, and how you frame those ideas determines whether they get resourced or quietly shelved.

Key Takeaways

• A fun brainstorm that produces nothing is a failure. Idris's first churn session had energy and zero outcomes. Design for ideas with owners, not for the buzz in the room.
• Separate brainstorming from deciding. State the mode out loud. Critique during generation kills the next idea before it is spoken.
• Safety comes before ideas. Without psychological safety you get people repeating what they think you want, which only looks like agreement.
• Silent brainstorm first. Individual writing before group talk multiplies the number and diversity of ideas and surfaces the quiet people who are often closest to the problem.
• Use a transparent ranking like dot voting. It produces a defensible priority order that the loudest voice cannot override, and it lets a quiet analyst's idea win on merit.
• Let AI handle structure and synthesis. Agenda design, prompts, and theming a pile of raw ideas are exactly what AI does well, freeing you to focus on the people.
• Always close the loop. End with owners, dates, and decision points, recap within a day, and report back even on ideas that failed. That is what keeps people offering their next idea.