Plan and Decision Stress Test
Stress-test plans with strong counterarguments, causal pre-mortems, evidence audits, and a proceed, modify, or reconsider verdict.
Before & after
Same prompt, same agent — the only difference is the skill.
Let me think through some potential issues with your chatbot plan. There could be some challenges around accuracy and customer satisfaction. You might run into problems if the AI makes mistakes or if it frustrates customers by not understanding their issues properly. You could also face some adoption challenges if users prefer talking to humans. There's also the question of how well it will handle complex problems. What have you already considered in your planning?
RED TEAM REPORT: AI Chatbot Customer Support Launch ASSUMPTION AUDIT (What could fail if wrong?) Assumption 1: "Customers prefer instant bot response over 30-min human wait" → Risk: If false, chatbot deflection feels patronizing. Measure via post-chat survey (target: >70% would use bot again for routine issues) Assumption 2: "Bot accuracy at 85% is acceptable" (misses 1 in 7 customer intents) → Risk: Misrouting to wrong department worse than no bot (escalation cost + frustration spike). Propose: Start at 92% accuracy gate or don't ship FAILURE MODE 1: Confabulation at Scale Scenario: Bot confidently hallucinates a refund policy that contradicts website • Customer chats bot Tuesday, gets false refund promise • Escalates to human Friday (3-day lag, customer frustration compounded) • Legal liability if bot contradicts terms-of-service Mitigation: Every answer must cite source; disallow free-form generation on policy questions. Audit logs show bot reasoning. FAILURE MODE 2: Cost Shifting, Not Reduction Scenario: Bot handles 60% of chats (30% reduction expected) Reality: Bot mishandles issues, humans re-explain 40% of bot interactions = net cost increase • Bot saves 18% support time (30% deflection – 40% rework overhead) • Investment cost ($50K build, $2K/mo hosting) not recovered for 18 months Mitigation: Measure: Bot handoff-to-human ratio weekly. If >35% of bot chats escalate, pause feature FAILURE MODE 3: Reputation Damage Scenario: Chatbot visible in early metrics or press coverage Reality: "AI startup replaces support team with bot" headlines kill hiring/trust Mitigation: Market as "AI-assisted support" not "replacement." Highlight human escalation path. Message consistency across PR, website, docs. BREAK-EVEN ANALYSIS Investment: $50K build + $2K/mo operating = break-even at 280 support hours saved (6-month payback) Assumption risk: If chatbot solves only 50% intended volume, payback = 12+ months (cash burn continues) GO/NO-GO GATE BEFORE LAUNCH ☐ 92%+ accuracy on 500 real customer questions (backtest on historical tickets) ☐ Human review: 100% of chatbot responses for first week (not 10%) ☐ Escalation path visible before customer frustration (no wait loops) ☐ Legal: Terms-of-service explicitly authorizes AI support ☐ Success metric agreed with finance: cost savings, CSAT, deflection %, rework rate
About this skill
"What are we missing?" should produce more than a list of things that might go wrong. Stress-test a plan by finding its load-bearing assumptions, building the strongest opposing case, and tracing plausible failures back to decisions you can change now.
What it does
- Separates the proposed decision, success criteria, constraints, and assumptions before challenging them.
- Builds a credible opposing case without replacing your actual proposal with an easier target.
- Runs causal pre-mortems that connect early decisions to failure and identify warning signals.
- Audits whether evidence supports the exact claim being made, including sample bias and correlated sources.
- Ranks objections by their effect on the decision and proposes bounded tests with explicit decision rules.
- Ends with a proceed, modify, or reconsider verdict, including what evidence would change it.
What you get
A decision review containing the strongest objections, an evidence and assumption ledger, plausible failure chains, and concrete next steps. Quick questions receive a focused challenge; major commitments receive the fuller review.
Best for
Founders, product managers, engineering leads, and operators making roadmap, architecture, launch, hiring, vendor, or strategic commitments before the evidence is settled.
SKILL.md preview
---
name: plan-and-decision-stress-test
description: >-
Stress-tests plans and decisions using steelmanned counterarguments, causal pre-mortems, and an audit of load-bearing assumptions, then recommends proceed, modify, or reconsider with decision-changing tests. Use when the user says "play devil's advocate", "stress-test this plan", "what are we missing?", "run a pre-mortem", "challenge my assumptions", "is this evidence strong enough?", or "we keep committing to plans that fall apart".
---
# Plan and Decision Stress Test
… (sign up to view the full skill)More business & ops skills
View all Business & Ops skills →Startup Idea Validation Plan
Test an idea before you build: grade evidence (said, did, paid), find the riskiest assumption, run the cheapest test with kill criteria.
Market-First Venture and PMF Verdict
Get a blunt build, narrow, test, or stop verdict based on buyer demand, retention, distribution, and explicit confidence.
Micro-SaaS Go/No-Go Validation
A build, test, reframe, or kill verdict on a micro-SaaS idea, with revenue math, fatal-flaw gates, and a test you can run this week.
Business Decisions From Live Notion Data
Turn Notion pipeline, revenue, and client records into evidence-backed business decisions and a focused action plan.