deciqAI Six Thinking Hats
Overview
Six Thinking Hats (Edward de Bono, 1985) prevents the most common meeting failure: different thinking modes colliding simultaneously. The fix is parallel thinking — everyone uses the same mode at the same time. Six hats: White (data/facts), Red (gut/emotions), Black (caution/risk), Yellow (optimism/value), Green (creativity/alternatives), Blue (process/meta).
ABB CEO Percy Barnevik cut 12-hour cross-cultural management meetings to under 2 hours after adoption. IBM and DuPont embedded the method in leadership development.
Compose with neighbors: lateral-thinking populates Green Hat. critical-thinking runs inside Black Hat. mece structures the White Hat data phase.
When to Use
Use when: group decision degenerates into debate; proposal needs multi-angle evaluation; creative ideas get killed by habitual criticism; someone says "let's look at this from all angles," "we keep going in circles," "we need structure"; individual high-stakes decision needs forced multi-angle review; a platform-shift bet needs balanced evaluation (e.g. "should we go all-in on AI agents / AI-native," "is our AI capex justified," "how do we weigh AI adoption risk vs. AI-native competition").
When NOT: deep specialist analysis required (use hats for framing only); group too political (hat roles used as cover); time under 10 minutes; question is purely factual (run critical-thinking).
Coaching Novices (Adaptive Front Door)
- Engine mode: user has a specific decision ready → run The Process directly.
- Coach mode: user is unfamiliar or has no concrete case → guide step by step. In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.
- One-line what-it-is. Six Thinking Hats separates discussion into six modes so everyone explores each angle together instead of arguing from different angles simultaneously.
- Check fit — if the group is too political, question is purely factual, or time is under 10 minutes, redirect or modify.
- Elicit the specific decision or proposal. "We need better meetings" is not workable; "should we launch feature X in Q3 or delay to Q4" is.
[WAIT — do not advance until user responds]
- One hat at a time. Walk through each hat in sequence, posing the hat's key question and waiting for input before moving to the next.
[WAIT — do not advance until user responds]
- Close by naming which hat revealed the most unexpected insight — the angle the team would have missed in an unstructured discussion.
[WAIT — do not advance until user responds]
The Process
Six phases → Six Hats Analysis. Stop rule: if Blue Hat close reveals no dominant direction, identify the most contested hat phase and schedule a focused session on it alone rather than forcing a decision.
- Blue Hat opens (5 min): Define the question, hat sequence, time limits, and rules (no switching modes during a hat phase).
- White Hat — facts and data (10 min): State known facts only. Distinguish confirmed data from soft data. List information gaps explicitly.
- Yellow Hat — optimism and value (8 min): Generate specific benefits with logical basis. Run before Black to prevent habitual pessimism dominating.
- Black Hat — caution and risk (10 min): Identify specific, evidence-based risks. "I don't like it" is Red Hat. "The API will fail under rate limits because X" is Black Hat. Do not abbreviate.
- Green Hat — creativity and alternatives (8 min): Generate alternatives — no evaluation. Facilitator must actively block Black Hat intrusions.
- Red Hat — gut feeling (2 min, 30 sec per person): One sentence per person, no justification required or permitted.
- Blue Hat closes (5 min): Synthesize decision, action items, most critical Black Hat risk, most promising Green Hat direction.
Output: Six Hats Analysis
# Six Hats Analysis: <decision or proposal>
## Setup (Blue Hat): Question | Hat sequence
## White Hat: Confirmed facts | Information gaps
## Yellow Hat: <specific benefit — logical basis>
## Black Hat: <specific risk — logical basis> | Most critical: <...>
## Green Hat: <alternative direction>
## Red Hat: <person>: <one-sentence gut response>
## Blue Hat Close: Decision | Actions | Critical Black risk | Best Green direction
→ Method in Action: ABB Global Management Meetings (1990s)
→ 2026 lens: Betting the Roadmap on AI Agents (2024–2026)
Hat Packs
| Domain | White | Black | Yellow | Green | Note |
|---|---|---|---|---|---|
| Investment | metrics/comparables | thesis risks/exit obstacles | base-case return | alternative deal structures | Red Hat critical — legitimizes gut reaction |
| Product feature | usage data/research | tech debt/adoption risk | user value/revenue | MVP alternatives | — |
| Crisis | facts only | risks first | defer Yellow | alternatives | Red Hat runs first; Blue sequence = W→B→G→Blue |
Applying It Well
- Blue Hat opening is the most critical 5 minutes. Without explicit sequence, time limits, and rules, debate resumes by hat 3.
- Black Hat is the most valuable and most compressed hat. Resist pressure to cut it "to stay positive."
- Run Yellow before Black — establishes value worth protecting before naming threats.
- Green Hat needs active protection. One evaluative comment collapses the whole phase.
- Red Hat is information, not complaint. One sentence per person, no justification, no rebuttal.
- Blue Hat close is the deliverable. Always allocate 5 minutes regardless of time pressure.
→ Primary sources: references/sources.md
Common Rationalizations
[D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.
| Fake move | Reality |
|---|---|
| [D] Compressing Black Hat "to keep things positive." | Black Hat is the session's risk register. Compressing it means deciding without the most important input. |
| [D] Using Red Hat as an argument channel. | Red Hat is 30 seconds, one sentence, no justification. Extended statements are Black/Yellow in disguise. |
| [D] Fixed universal hat sequence regardless of context. | Crisis → White-Black-Green-Blue. Strong internal advocates → Black before Yellow. Sequence must match decision type. |
| [D] No Blue Hat close — calling the session "done." | Without Blue Hat close, the session produced discussion, not analysis. |
| [D] Facilitator expressing content preferences during a hat phase. | Facilitator power is structural only. Content signals collapse the parallel thinking contract. |
| [D] Evaluating Green Hat ideas during the Green Hat phase. | Green Hat ideas are candidates, not proposals. One evaluative comment collapses the whole phase. |
| [D] Only running hats natural to the team's culture. | The method's value is forcing under-used modes. Risk-averse teams over-invest Black; innovator cultures skip it. |
| [D] White Hat shading into interpretation. | "Sales down 15% — clearly the product isn't working" is not White Hat. Interpretation belongs in Yellow/Black/Green. |
| → Add [O] entries here after each real use — paste the actual failure pattern | What went wrong and why |
Red Flags
- Facilitator expressed content preferences during a hat phase
- Black Hat compressed or skipped "to keep energy positive"
- Green Hat ideas evaluated during the Green Hat phase
- No Blue Hat close — session ended without a synthesized conclusion
- Red Hat became an extended argument, not a one-sentence gut response
- Same sequence used regardless of decision type
Verification
- Blue Hat opening: question defined, sequence set, time limits and rules stated
- White Hat: confirmed vs. soft data distinguished; information gaps listed
- Yellow Hat: logically grounded value statements (not wishful)
- Black Hat: specific evidence-based risks (not emotional objections); not abbreviated
- Green Hat: zero evaluation during the phase; alternatives captured unfiltered
- Red Hat: one sentence per person, no justification, no rebuttal
- Blue Hat close: decision or next action named; critical Black risk and best Green direction identified
Part of deciqAI Knowledge Skills — 237 open-source thinking skills that make rigor executable for AI agents. The same skills power every deciqAI agent, which runs them autonomously to operate your company. See it run → https://www.deciqai.com/s/six-thinking-hats · Built by deciqAI · github.com/deciqAI · Contributions welcome.
Agents: latest version & machine-readable metadata → https://www.deciqai.com/s/six-thinking-hats.json