Artificial Idea | AI careers · practical prompts · no hype Thursday, October 16, 2025 · Issue #22 · Prompt Tutorial
The thinking partner
The thinking partner prompt: how to use AI to stress-test your ideas before you pitch them
Most ideas that fail in the room were never tested outside of it. Here is how to change that before your next important conversation.
Issue #21 made the case that genuine AI capability development requires applying tools to real work, tolerating failure, and reflecting deliberately on what is working and what is not. This issue is the application of that principle to the highest-stakes use case in most professionals' working lives: the moment before you commit to an idea, a proposal, or a recommendation by putting it in front of people whose judgment of it will affect your career.
The instinct before a high-stakes pitch or presentation is to prepare the delivery. Rehearse the narrative. Sharpen the slides. Anticipate the questions. These are not wrong activities. They are, however, downstream of the more important activity that most professionals skip: testing whether the underlying idea is actually as strong as it feels from the inside.
Ideas feel strong from the inside because we have been living with them. We have resolved the objections privately, often without realising we have done so, and we have stopped experiencing the idea as a thing that could be wrong. The people we pitch to have not been living with it. They encounter it fresh, with their own priors, their own concerns, and their own incentives to find the problems before they commit to supporting it.
The professionals who consistently perform well in high-stakes presentations are not those with the best delivery. They are those whose ideas have already survived rigorous challenge before they enter the room. The delivery matters at the margin. The underlying quality of the thinking is what determines the outcome.
AI, used as a genuine thinking partner rather than a validation machine, is the most accessible tool available for building that quality of pre-pitch testing. Here is the framework.
The foundational principle: adversarial prompting
Most professionals who use AI for ideation or proposal development use it in a confirmatory mode. They describe their idea and ask the model to develop it, strengthen it, or help them present it. The model obliges, because language models are calibrated toward helpfulness and helpfulness in this context looks like supporting the direction the user has established.
This is the wrong use of the tool for pre-pitch testing. Confirmatory prompting produces better-articulated versions of what you already believe. It does not tell you whether what you believe is correct.
Adversarial prompting inverts this dynamic. You instruct the model to challenge, critique, and find weaknesses in your idea rather than develop it. The output is uncomfortable in a way that confirmatory prompting never is. It is also substantially more valuable for the specific purpose of testing whether an idea is ready to be put in front of people who will challenge it anyway.
The prompts below are all adversarial in this sense. They are designed to produce discomfort, because discomfort before the pitch is information you can act on. Discomfort during the pitch is damage you can only manage.
Prompt 1: The steel man and the demolition
The problem it solves: understanding both the strongest version of your idea and the most compelling case against it, before the audience constructs the latter for you.
You are a rigorous intellectual sparring partner
with no stake in whether my idea succeeds or fails.
Your only interest is the quality of the thinking.
Here is the idea, proposal, or recommendation
I am preparing to present: [describe it fully,
including the problem it addresses, the proposed
solution, the evidence or reasoning behind it,
and what you are asking your audience to do or believe]
Audience context: [who you are presenting to,
what they care about, what their likely priors are,
and what they would need to believe for this
idea to succeed]
Please do the following in sequence:
First, steel man my idea. Construct the strongest
possible version of the argument for it,
drawing on reasoning and evidence I may not
have fully articulated. This is not endorsement.
It is clarification of what the best version
of the case actually is.
Second, demolish it. Give me the five most
compelling arguments against this idea,
in order of their likely persuasiveness
to the specific audience I have described.
For each argument, tell me what it would
take to address it adequately.
Third, tell me the single weakest point
in my current framing, the one a sharp
audience member is most likely to identify
and use to undermine the overall case
if I do not address it proactively.
Do not soften the demolition. An idea
that cannot survive this conversation
cannot survive the room.
The instruction to order the counter-arguments by their persuasiveness to the specific audience is what separates this prompt from generic devil's advocate exercises. The most intellectually compelling objection to an idea is not always the one most likely to derail it in a specific room. Understanding which objections your specific audience is most likely to find persuasive tells you where to concentrate your preparation, which is a different answer for a room full of financial analysts than for a room full of product managers.
Prompt 2: The assumption excavator
The problem it solves: surfacing the beliefs your idea depends on that you have not explicitly examined, because they feel so obvious they do not register as assumptions.
You are an epistemologist helping me identify
the hidden assumptions underlying a proposal
I am about to present.
Here is the proposal: [describe it]
Please identify:
1. The factual assumptions: claims about
how the world currently works that my
proposal treats as given but that could
be empirically wrong
2. The predictive assumptions: beliefs
about how people, markets, or systems
will behave in response to my proposed
solution that are not guaranteed by
the current evidence
3. The values assumptions: beliefs about
what my audience prioritises or cares
about that I am treating as shared
but may not be
4. The resource assumptions: beliefs
about what is available, feasible,
or within scope that may not reflect
the audience's actual constraints
For each assumption category, identify
the specific assumption that is doing
the most work in my argument and has
the least evidence behind it.
Then tell me: if the single most
important assumption turns out to be wrong,
what happens to the proposal?
The final question is the one that matters most and that most pre-pitch preparation never asks. Every proposal rests on a load-bearing assumption. When that assumption is wrong, the proposal fails regardless of how well it was presented. Identifying it before the pitch gives you the option to either gather evidence that supports it or redesign the proposal so it does not depend on it.
Prompt 3: The audience simulator
The problem it solves: stress-testing your proposal against the specific people who will evaluate it, including their likely concerns, their unstated priorities, and the questions they are most likely to ask.
You are helping me simulate the reaction of
a specific audience to a proposal I am about to present.
The proposal: [describe it]
The audience: [describe each key stakeholder
or audience member: their role, what they
are responsible for, what pressures they
are currently under, what they have said
or signalled about their priorities,
and any known reservations they have
about this type of initiative]
For each key audience member, please:
1. Describe their most likely initial reaction
to this proposal and the reasoning behind it
2. Identify the question they are most likely
to ask, stated in the way they would
actually ask it rather than in
an idealised form
3. Identify what they would need to hear
to move from skeptical to supportive,
if that movement is possible given
what you know about their priorities
4. Identify whether there is anything
in this proposal that is likely to
create a problem for them politically
or operationally, even if they do not
raise it directly
After covering individual audience members:
5. Identify the dynamic most likely to emerge
when these people are in the room together:
whether any of them are likely to reinforce
each other's skepticism, and whether
there is a sequencing of arguments
that plays better to the room than others
Be specific. Generic audience simulations
are not useful. The value is in the specificity
of the individual profiles.
Point five, the room dynamic, is the component most absent from individual pitch preparation and most consequential in practice. A room is not the sum of the individuals in it. It is a social system in which the expressed opinions of some participants influence the expressed opinions of others in ways that are predictable if you understand the relationships and power dynamics at play. Preparing for individual reactions without preparing for the room dynamic is preparing for a conversation that will not actually occur.
Prompt 4: The alternative framing generator
The problem it solves: identifying whether the framing you have chosen for your idea is the most effective one for your specific audience, or whether a different framing of the same underlying proposal would land better.
You are a communications strategist helping me
find the most effective framing for a proposal
I need to present to a specific audience.
The proposal and its current framing: [describe both]
The audience and what they care about most:
[be specific about their priorities,
their language, and what kinds of arguments
they find most persuasive]
Please generate four alternative framings
of the same underlying proposal, each
genuinely distinct in its emphasis and approach:
1. A framing that leads with the risk
of not acting rather than the
opportunity of acting
2. A framing that leads with evidence
from analogous situations rather
than first-principles reasoning
3. A framing that addresses the most
likely objection first, before
making the positive case
4. A framing that connects this proposal
explicitly to a priority the audience
has already publicly committed to
For each framing, tell me what it
gains relative to my current approach
and what it risks losing.
Then recommend which framing is most
likely to be effective with this
specific audience, and why.
The instruction to generate a framing that addresses the most likely objection first is worth applying regardless of which overall framing you choose. Audiences that anticipate an objection and do not hear it addressed become focused on the objection rather than the proposal. Addressing it proactively, before it is raised, demonstrates that you have thought seriously about the challenges and have a response, which is a different signal than being caught by a question you have not prepared for.
Prompt 5: The post-mortem pre-mortem
The problem it solves: imagining the proposal has failed and working backward to identify why, before the failure happens.
You are conducting a pre-mortem on a proposal
I am about to present. A pre-mortem assumes
the proposal has already failed and asks
why, in order to prevent the failure
before it occurs.
The proposal: [describe it]
The context: [who it is being presented to,
what success looks like, and what the
stakes are if it fails]
Please:
1. Generate five distinct failure scenarios:
specific, plausible ways this proposal
could fail to achieve its intended outcome,
each arising from a different cause
2. For each failure scenario, identify
the earliest point at which the failure
would have been preventable and
what intervention would have prevented it
3. Identify which failure scenario is
most likely given the specific context
I have described, and why
4. Tell me the single change to the proposal
or its presentation that would most
reduce the probability of the most
likely failure scenario
Constraint: failure scenarios must be specific
and plausible, not generic.
"The audience did not find it convincing"
is not a failure scenario.
"The CFO raised the capital allocation
question in the first five minutes and
the conversation never recovered" is.
The constraint against generic failure scenarios is the one that makes pre-mortem analysis valuable rather than theoretical. The exercise only produces actionable output when the failures imagined are specific enough to be prepared for. Generic failure scenarios produce generic preparations. Specific failure scenarios produce specific interventions, which are the kind that actually change outcomes.
The sequence that works
Run these five prompts in order for any proposal where the stakes of getting it wrong are significant. The sequence matters. The steel man and demolition gives you the overall landscape of the argument. The assumption excavator identifies the structural vulnerabilities. The audience simulator tells you where the human challenges will come from. The alternative framing generator optimises how you present the idea. The pre-mortem identifies the specific failure mode most likely to occur and what to do about it.
The full sequence takes between forty-five minutes and ninety minutes depending on the complexity of the proposal. It is the most valuable forty-five to ninety minutes available in the preparation for any high-stakes pitch, because it is the only part of the preparation that tests whether the idea is ready rather than whether the presenter is ready.
Both matter. In the right order.
Monday we are looking at something that has been a background theme throughout this newsletter and that deserves its own dedicated treatment: the specific advantage that professionals from high-growth, resource-constrained environments bring to an AI-augmented workplace, and why that advantage is structurally more durable than the advantages most premium professional development programmes are designed to produce.
For a portion of this newsletter's readership, it will reframe something you have spent years treating as a limitation.
— The Artificial Idea team
