When AI Sounds Certain, Can You Still Name Your Own View?

Write your first view before absorbing the proposal

Take a moment to describe the decision using what you already know. What does the company need? Which constraint matters? What would a satisfactory choice accomplish? You do not need a complete independent analysis. You need enough of an initial account to notice when the proposal changes the question.

Suppose the organization needs a dependable way to serve a demanding client. A proposal framed entirely around reducing preparation time may be answering a narrower question. Time matters, but reliability, responsibility, and the client's expectations also belong in the decision.

Your first view can be wrong. Keeping it visible allows you to explain why it changed rather than assuming the last confident paragraph you read must be your own conclusion. It also gives the director something useful to challenge. Do not present your initial view as the answer everyone must now defend.

Find the claim on which the conclusion depends

A recommendation may contain many accurate statements and one unsupported assumption that determines the outcome. Look for the sentence that makes the preferred option appear clearly better. Perhaps it assumes a customer will tolerate a delay, that a team can absorb review work, or that a pilot result applies to an entire division.

Ask the author to point to that claim. Then ask how they know it. Separate records from estimates, and separate estimates from hopes. The fact that an AI tool can produce a reasonable argument does not establish the fact on which that argument rests.

Spend attention in proportion to consequence. If a small reversible trial can answer the uncertainty, you may not need a lengthy debate. If the decision creates a difficult-to-reverse commitment, the missing evidence deserves a more deliberate review. There is no virtue in challenging every sentence equally.

Follow the evidence to its actual scope

Open the relevant source rather than relying on the proposal's summary. A result from one team, vendor, or period can be useful while leaving your intended use untested. Check whether the source describes the same work, people, and outcome. If it does not, ask what justifies the extension.

The NIST AI Risk Management Framework is voluntary organizational guidance for managing AI risks. It can inform a company's review process. It does not certify this recommendation or establish that a particular management decision is safe.

Keep that distinction throughout the brief. A framework, case study, benchmark, and your own operating records serve different purposes. A recommendation can draw on all of them, provided it says what each supports. Where the source is inaccessible or ambiguous, record the limitation instead of smoothing it into certainty.

Ask for a real alternative, not the weakest objection

A recommendation can appear balanced while comparing the preferred option with an obviously unattractive alternative. Ask what a competent person would choose if they rejected the recommendation. They might prefer a smaller test, a different workflow, or keeping the current arrangement until a particular constraint changes.

Invite the author to explain the strongest version of that choice. What does it protect? What does it cost? What evidence would make it preferable? This is especially useful when the proposal was prepared to satisfy a senior person's stated enthusiasm.

Do not turn the exercise into a performance where everyone must invent objections after the decision is already settled. If leadership has made a nonnegotiable commitment, say so and review the questions that remain open. Genuine disagreement requires knowing that relevant evidence can affect the result.

Use a challenge table without pretending it decides for you

A table can keep claims, sources, and uncertainties from dissolving into a general feeling that the proposal is persuasive. It also prevents a discussion from circling the same objection without identifying the information needed to resolve it.

Use a few consequential claims. If every sentence becomes a row, the exercise can bury the decision. Ask the responsible author to distinguish an answer already available from evidence that must be gathered. Put a person and a next action beside the unresolved item.

You still have to judge the tradeoff. A table that contains several green cells is not an approval rule. The important issue might be the single assumption you cannot yet verify. Or the remaining uncertainty may be acceptable for a limited test. Explain the reasoning in the decision note.

A hypothetical managing partner changes the question

A managing partner receives an AI-assisted proposal for preparing client recommendations in the following hypothetical example. The presentation emphasizes faster drafts. Its preferred option reduces the time assigned to senior review because the initial output appears more complete.

The partner writes an initial question: can the firm give clients dependable advice with a clear accountable reviewer? That exposes a missing assumption. The pilot counted drafting time but did not include the review of unusual client circumstances. The team has evidence about speed, but only partial evidence about the full service.

Instead of rejecting the entire proposal, the partner asks for a bounded trial with the review responsibility retained and the omitted work recorded. The author revises the recommendation. This scenario does not prove that retaining review is always right. It illustrates how a specific challenge can produce a better next step without turning skepticism into permanent refusal. It is a hypothetical example, not a claim about a past participant.

Hear dissent without asking people to predict your reaction

Your colleagues may know that you favor the proposal. Asking whether anyone has concerns is not enough if your next response makes disagreement costly. Explain which questions are still open and request evidence before announcing how certain you feel.

You can ask someone with relevant experience to examine a particular assumption. Give them access to the material and time to think. Do not demand a spontaneous argument in a crowded meeting and treat silence as agreement. Some concerns need a separate conversation or an established confidential route.

When a colleague raises a useful objection, show what changed because of it. If you disagree, explain why. The aim is a review process people can understand and use again, including when their account is inconvenient. Your responsibility does not disappear because you invited dissent, and theirs does not depend on pleasing you.

Decide when more review stops helping

Executives can overtrust a polished recommendation. They can also use the possibility of an error to postpone a call indefinitely. Distinguish missing evidence that matters from a wish to feel certain before taking responsibility.

Ask what the next piece of information could change. If it would alter the option, scope, timing, or authority, identify how to obtain it. If it would simply repeat what you already know, make the decision you are responsible for making. Record important uncertainty instead of pretending the review eliminated it.

For a reversible test, decide what would lead you to stop or revise it. For a consequential commitment, use the company's appropriate specialist and governance processes. This article offers a practical way to organize your questions. It cannot replace the expertise or authorization the decision requires.

Keep the decision note useful after the meeting

Write what you chose, the reason, the material assumption, and the condition that would make you reconsider. Name who is responsible for the next action. Keep the underlying evidence accessible according to company requirements. A person arriving later should be able to tell how you reached the decision without reconstructing an entire conversation.

If you keep endorsing a proposal before examining it, or rejecting one that challenges your preferred view, bring an actual example into reflection. What did you notice? What did you choose not to ask?

AI and the Andes Retreat, part of The SoulSearching Experience, gives you time to work through your question with guidance and a small group in Peru’s Sacred Valley. Participants share perspectives and examine possible responses. You remain responsible for deciding what fits your circumstances.

The retreat lasts five days. Bring a de-identified account of the situation rather than confidential company records, and ask for current practical details. The next step you choose may be as concrete as a conversation you have been avoiding.

Bring the practical question back to your own response

What makes stating a provisional view feel risky?

Choose the most recent moment this happened. Write what was actually said or requested, what you concluded it meant and the next action you took. Keep those parts distinct. You may discover a real practical gap, a response that repeats in several settings, or both. The observation is a starting point, not a diagnosis.

Write your initial reasoning privately, then check it against actual evidence and appropriate advice.

Ask what your usual action provides immediately and what it displaces afterward. Being available may give you contact; intervening may restore familiar influence; another explanation may postpone a difficult conversation. These are possibilities to examine against your own experience. They should not be imposed as the reason you act.

Choose a response you can try in an appropriate, ordinary situation. Respect actual company duties, your household’s agreements and other people’s boundaries. If specialist advice or a formal arrangement is needed, obtain it through the right route. Personal insight cannot supply missing authority or decide another person’s obligations.

At the AI and the Andes Retreat, this personal question is the work to bring. Florian facilitates and offers useful guidance while you investigate what fits your circumstances. You participate in finding your answer and choosing an action for home. The group is not a corporate advisory team, and the experience does not guarantee a business, career or relationship outcome.

Write what you will observe after returning: the moment that usually begins the pattern, the action you choose and the consequence. Review whether you tried the response and what it revealed. Support and accountability continue afterward; ask about current details. Progress need not take the form of an impressive announcement.

How do you say what you think before you know whether others agree?

Naming your view first is a separate skill from being right about it. You might notice that you wait until the room settles before speaking. Ask whether you have changed your view because of something you learned, or whether you have not yet examined what you think.

Prepare one sentence before the discussion: given what I currently know, I would lean toward this option, for this reason, and this is what could change my mind. Say it early, while it is still visibly provisional. Explain that you are offering a working position and want to hear what you have missed. You do not need to guess how everyone will receive it before making your reasoning clear.

Suppose a discussion concerns an AI-assisted recommendation you have been asked to endorse. You could say: I understand the proposed conclusion, but I cannot yet explain why this assumption holds. I want to check it before agreeing. That gives the group a specific question to examine while leaving your position open to change.

If the room goes quiet, do not rush to fill the silence with concessions. Ask what others see that you have not yet seen. You may experience disagreement as disapproval. Check what the person is actually questioning before responding to that interpretation. If someone argues well, say plainly that your view moved and why. Making your reasoning visible gives other people something specific to question.

If you hesitate to question a recommendation because you feel you should already understand it, read how to let yourself learn AI after years of being the expert for the difficulty of asking a basic question.

To protect the time you need to form your own view before the proposal arrives, see what happens when you make time to think.