Every leader knows the feeling: a six-figure decision on the table, a spreadsheet graveyard in your tabs, and a team that’s been “almost ready to decide” for three weeks. That’s exactly where our client was — until we rebuilt their process around decision making with ChatGPT.
In this case study, you’ll see the exact 5-step workflow, the actual prompts we used, the results after 90 days, and — just as importantly — where ChatGPT fell short.
Note: Details below are composited from a real consulting engagement to protect client confidentiality.
Why Decision Making with ChatGPT Matters in 2026
The era of “gut feeling” leadership is over. AI-augmented decision making has become a standard operating practice for companies of every size — we broke down what this shift means in Make High-Impact Decisions with A.I.: Your Strategic Advantage in 2026.
Yet access to AI doesn’t automatically produce better choices. Most companies are data-rich but decision-poor: drowning in information, starving for a clear “yes” or “no.” The difference is process. Here’s proof.
The Case: Northtrail Gear’s $250,000 Question
The company: Northtrail Gear (name changed), an 8-person outdoor-apparel e-commerce brand, ~$2.4M revenue.
The decision: How to deploy $250K of cash over the next 12 months. Three camps had formed:
- Launch a subscription box (growth team’s favorite)
- Expand wholesale distribution (finance’s pick)
- Hold cash and optimize current operations (the founder’s instinct)
The problem: They’d spent three weeks in meetings, re-pulling the same data, and drifting toward the worst option of all — no decision at all, which as we’ve written before, is still a decision (and often the worst one). (Internal link: “Why ‘No Decision’ is Still a Decision (and Often the Worst One)”)
We stepped in with a simple rule: ChatGPT doesn’t make the decision — it accelerates every step around the decision. Here’s the workflow.
The 5-Step ChatGPT Decision Workflow
Step 1: Frame the Decision in One Prompt
Bad inputs, bad outputs. We forced the team to define the decision crisply:
“Act as a neutral decision facilitator. Restate this decision in one sentence, list the 3–5 criteria that should drive it, and identify the information actually required vs. nice-to-have. Decision: [full context, constraints, budget, timeline].”
ChatGPT’s output cut the “nice-to-have” data list by half — instantly deflating the analysis paralysis we describe in How to Make Better Decisions When You Don’t Have All the Data.
Step 2: Generate Options — Then Destroy Them
We asked ChatGPT to argue against each option, including its own earlier suggestions:
“Play devil’s advocate. For each of these three options, list the 5 strongest arguments that it will fail, the cognitive biases most likely to make us overrate it, and the early warning signs we’d ignore.”
This surfaced a classic trap: the subscription box was the team’s favorite partly because they’d already spent $18K prototyping it — textbook sunk cost fallacy. Naming the bias out loud changed the room’s energy.
Step 3: Score the Options in a Decision Matrix
Next we turned opinions into numbers. ChatGPT proposed weighted criteria (cash-flow impact 30%, strategic fit 25%, time-to-revenue 20%, risk 15%, team capacity 10%), and the team scored each option 1–5 — a structured weighing process we also cover in The Art of Cost-Benefit Analysis: Your Complete Guide to Smarter Decisions. (Internal link) We ran the scoring in our free Decision Matrix tool so the math stayed transparent.
Result: Wholesale expansion won (4.1), narrowly beating “hold and optimize” (3.7). Subscription scored lowest (2.9) once prototyping costs were excluded as sunk.
Step 4: Run a Pre-Mortem on the Winner
“Assume it is 12 months from now and we chose wholesale expansion, and it failed. Write the autopsy: the 7 most probable causes of failure, ranked, with one mitigation each.”
The pre-mortem exposed a real vulnerability — retailer payment terms straining cash flow — which became a negotiated 30-day term limit in every contract. This is exactly the kind of structured stress-test we recommend in Decision-Making Under Uncertainties: Lessons from Real Businesses.
Step 5: Decide, Document, and Schedule the Review
The founder made the call in a 45-minute meeting — with ChatGPT drafting the decision memo (rationale, assumptions, kill-criteria, review date). We logged the decision in the Daily Decision Planner and set a 90-day review. One caution from the memo: track a balanced set of indicators, not a single vanity number — see The One Metric Fallacy: Why Single KPIs Mislead Leaders.
The Results: What Changed in 90 Days
- Decision cycle time: 21+ days → 6 days (3×+ faster)
- Meeting hours on this decision: ~34 → 9
- Wholesale pipeline: +18% signed LOIs in one quarter, with payment-term protections in place
- Subscription pilot (small, low-budget): kept as a 90-day experiment instead of a $250K bet
- Team confidence: the documented rationale ended re-litigating the decision every Monday
What Didn’t Work (Honest Limitations)
- Hallucinated benchmarks: ChatGPT confidently cited two “industry stats” that didn’t exist. Every number now gets verified at source.
- Anchoring: early AI outputs anchored the team’s estimates; we now collect independent team estimates before prompting.
- Over-iteration: “one more prompt” became procrastination in disguise. Time-box every step.
- No accountability: AI advises; humans own outcomes. Keep a human-in-the-loop always — a prerequisite we detail in the AI Decision Readiness Checklist: 6 Critical Areas to Evaluate Before Implementation.
The Repeatable Prompt Template (Copy This)
“You are a neutral decision facilitator. Context: [situation, constraints, stakes]. 1) Restate the decision in one sentence. 2) Define 4–6 weighted criteria. 3) List 3 options. 4) Devil’s-advocate each option. 5) Draft a pre-mortem for the leading option. 6) List the missing information that would change the outcome. Do not choose for me; end with the 3 questions I should answer before deciding.”
FAQ
Can ChatGPT really make business decisions for me? No — and it shouldn’t. ChatGPT accelerates framing, option generation, bias-checking, and documentation; the accountable human decides.
Which ChatGPT features matter most for decision making? Structured prompting (roles, criteria, pre-mortems), long-context analysis of your documents, and draft memos. Verify all facts and figures independently.
Is decision making with ChatGPT suitable for small businesses? Especially for them. Small teams lack internal devil’s advocates — ChatGPT provides one on demand, for free.
Final Thoughts
Northtrail’s story isn’t that AI picked the winner. It’s that a disciplined decision making with ChatGPT workflow turned three weeks of circular debate into six days of structured, documented, confidence-backed choice. Try the prompts above, run your options through our free decision tools hub, and remember: the goal isn’t a perfect decision — it’s a reliably good decision process.