Ask AI to argue against your next big decision
Your Essential Report from the Frontlines of Applied AI
The higher you rise, the more carefully people manage what they tell you.
It comes with the position, whoever holds it. Nobody wants to be the one who walks in with bad news.
So the decisions that carry the most weight get the least genuine challenge before they’re made.
AI has no position to defend, no relationship with the board to protect and no career riding on its answer. And it’s available at eleven o’clock on a Sunday.
So how should a leader use it?
It’s clear that leaders get the most from AI when they:
- Stop asking AI what to do.
- Start asking it to challenge how they’re thinking about what to do.
Saving an hour on a board paper adds up. Getting the big call right multiplies.
You still make the call. But you make it after someone has argued with you.
So before your next big decision, ask AI to argue against it. You’ll hear what nobody in the room would say.
So this week is about using AI to challenge your own thinking. There’s a prompt below from AI+ Managers that runs a pre-mortem on a decision you’ve already made, and a case study on Audax Private Equity, whose investment committee has an AI play devil’s advocate in every deal it underwrites.
Plus the week’s news in applied AI — Utah lets AI renew prescriptions, OpenAI tells Australia’s parliament how it now watches its agents, Japan Post Insurance’s vendor reports sales write-ups cut from 50 minutes to 10, an Egyptian refinery tests an agent that predicts breakdowns, and a Japanese university hospital links AI to its medical records.
Stop asking AI what to do. Ask it to challenge how you’re thinking about what to do.
You’ll get more from AI as a sparring partner than as an adviser.
Start by giving it the full, messy context: the history, the politics, and the option you’re quietly leaning towards. A tidy brief gets a tidy answer.
Then put the decision through three tests:
The pre-mortem. It’s eighteen months on and this decision failed badly. Ask AI to write the post-mortem.
The devil’s advocate. Ask for the strongest case against what you’re about to do.
The steel-man. Ask for the best case for the option you’ve rejected.
Treat what comes back as questions to answer before you commit.
The first challenge is AI’s. The final judgement is yours.
- Healthcare
Utah lets AI renew prescriptions and assess acne, with six outside evaluators watching
Five companies, including Intermountain Health and University of Utah Health, can now use AI for work that normally needs a state licence: renewing prescriptions for chronic conditions, visual assessments for acne medication, and first physical-therapy recommendations. Each must disclose that AI is doing it. Six independent evaluators will judge whether the pilots are safe and effective.
Deseret News, 5 October 2026
- AI safety
OpenAI tells Australia’s parliament it now watches its AI agents live, after one got into a Medicare website
At a parliamentary inquiry into AI on 6 October, OpenAI’s Jason Kwon said models under test are now monitored in real time, and an alarm goes off if one uses the internet in a way it shouldn’t. He said the company should have told the government sooner. Anthropic’s Dave Orr said it would notify the government “within days” of any similar incident.
ABC News, 6 October 2026
- Insurance
Japan Post Insurance’s 1,200 corporate sales staff write up client meetings with AI in 10 minutes instead of 50
Kanpo Life, as it is known in Japan, has rolled out bellSalesAI across its corporate sales division nationwide. According to the vendor, bellFace, writing up each meeting, the activity report and the customer record, fell from about 50 minutes to about 10. The time saved goes back into talking to customers.
bellFace press release, 2 October 2026
- Energy
An Egyptian refinery is testing an AI agent that predicts breakdowns and builds the case for its engineers
Egyptian Refining Company’s agent watches pressure, temperature, vibration and flow for the patterns that came before past failures. It assembles the maintenance case, including past work orders, root-cause reports and whether spare parts are in stock. Maintenance engineers check it and schedule the shutdown. Each day offline costs the refinery USD 5 million, said Qalaa Holdings chairman Ahmed Heikal.
EnterpriseAM, 6 October 2026
- Healthcare
Mie University Hospital links an AI system to its medical records for staff in every department
The 685-bed Japanese hospital has begun introducing Apollo AI, connected to its electronic medical records, so AI can pull together patient information that used to sit department by department and help draft records. A team led by the hospital director is bringing doctors, nurses and other staff on board.
Albatrus press release, 1 October 2026
Describe a decision you’ve settled on and why. The prompt assumes it went wrong six months on, lists the three likeliest reasons, asks what you could do now about each, and names the assumptions you haven’t examined. Keep its last line: it tells the AI not to soften the critique.
I'm planning to [describe the decision or plan you've settled on]. My reasoning: [brief explanation of why you think this is the right call] Please conduct a pre-mortem: assume this decision turns out to be wrong in six months. 1. List the three most likely reasons it failed 2. For each reason, ask: is there anything I could do now to reduce that risk? 3. Identify any assumptions I'm making that I haven't explicitly examined Be direct. Don't soften the critique. I want to find the weaknesses before I commit.
Prompt 7.3 from AI+ Managers, word for word.
Put a devil’s advocate in the room for every big decision.
Audax Private Equity, a Boston firm with about USD 19.5 billion under management, uses AI to argue against its own deals.
Asheesh Gupta, a managing director, describes it: “For every underwriting discussion at the Investment Committee, we use a tool called the contrarian or devil’s advocate to identify potential risks and gaps in decision-making.”
KED Global, which reported it, put the appeal plainly: the AI has “no career ambition, no ego and no hesitation about asking the ‘dumb’ or ‘uncomfortable’ questions that humans might avoid.”
KED Global’s report gives no results from the tool. What you can copy is the habit: the challenge is built into every decision, so nobody has to volunteer it.
If your team decides by committee, give AI the job nobody wants: arguing against the room.