Can AI Help You Make a Decision When Information Is Missing?
· By Michel Valenzuela, Founder of Daryl
Yes. AI can help identify what is missing, distinguish essential information from optional detail, and work out what to check next.
It cannot turn an unknown into a fact simply by producing a confident answer.
When information is incomplete, useful assistance should make the uncertainty easier to handle. Sometimes that leads to a decision. Sometimes it leads to one question that needs answering first.
Do you need all the information before deciding?
Often, you cannot obtain everything.
A project may need an answer before every dependency is confirmed. A supplier may provide an estimate rather than a guarantee. A new idea may have no track record to examine.
The task is to identify which uncertainties matter enough to affect your choice.
An assistant can help you ask:
“Which missing fact could make an option unsuitable?”
That is more useful than treating every unanswered question as equally urgent.
What kinds of information might be missing?
Separate the gaps.
| Type of uncertainty | Example | Useful next step |
|---|---|---|
| Missing fact | A supplier’s availability is unconfirmed | Ask the supplier |
| Unclear requirement | Nobody has defined who approves the work | Clarify ownership |
| Estimate | The work may take three to five days | Compare both cases |
| Future outcome | You do not know how people will respond | Consider a limited trial |
| Conflicting information | Two briefs list different deadlines | Confirm the authoritative version |
Different gaps need different responses.
More analysis will not establish a supplier’s availability. Another summary will not resolve which deadline was agreed. Someone needs to confirm the fact.
How should you describe the situation to AI?
State what you know and what you do not.
For example:
“I need to choose a delivery approach today. The deadline is Friday. Our internal team can complete a simpler version. A supplier could provide the full version, but their availability is unconfirmed.”
Then ask:
“Compare the options without assuming the supplier is available. Identify what I should confirm and what I can decide now.”
This keeps the unknown visible.
If the assistant recommends the supplier as though capacity were confirmed, correct that assumption before continuing.
How can you tell which missing fact matters most?
Ask how each possible answer would affect the decision.
Suppose you are waiting for three details: a supplier’s availability, their preferred file format, and whether they can attend a kickoff call.
Availability may determine whether the option is viable. File format may be easy to accommodate. The kickoff call may be replaceable with a written brief.
You can ask:
“For each unknown, explain whether the answer could eliminate an option, change its ranking, or simply affect implementation.”
That helps focus your attention.
Can AI compare different scenarios?
Yes, provided the scenarios are clearly labeled as possibilities.
For example:
“If the supplier is available, compare the full version with internal delivery. If they are unavailable, identify the best remaining approach.”
This creates conditional guidance.
The assistant should preserve the difference between a scenario and a confirmed situation.
A useful comparison may reveal that your next step is the same in both cases. Perhaps the internal team should begin a task that either approach will require.
That lets you move forward while another detail remains unresolved.
What if the deadline arrives before you get an answer?
Describe the deadline explicitly.
“Supplier availability must be confirmed by noon. If we have no answer then, we need to proceed internally to meet Friday’s delivery.”
That gives the decision a practical rule.
The assistant can help identify whether the rule fits your requirements and what needs to happen before noon.
It should not invent a probability that the supplier will respond or assume silence means agreement.
You are deciding how to proceed under uncertainty, using the commitments and information available.
Can a small test reduce the uncertainty?
Sometimes.
A trial can provide evidence about a question that discussion alone cannot settle.
If you do not know whether a new workflow is manageable, test it on one representative task. If an explanation is unclear, ask someone unfamiliar with the project to read it.
Before testing, define what you want to learn.
“Can we complete this task with fewer corrections?”
“Does the reader understand who the service is for?”
A test without a clear question can generate more activity without resolving the decision.
When should the assistant avoid recommending a choice?
When an essential gap prevents a meaningful comparison.
If two options depend on delivery dates you do not know, a ranking may be premature.
A useful response should explain the limitation:
“We cannot establish which option meets the deadline until availability is confirmed.”
It can still help prepare the questions or identify a fallback.
That is useful progress. The conversation has clarified what blocks the decision and how to address it.
What does a practical request look like?
You can use this structure:
“Help me examine this decision. Separate confirmed facts, assumptions, and unknowns. Identify the missing information that could change the choice. Compare realistic scenarios, then suggest the next step I can take now.”
Review the categories before accepting the recommendation.
An assumption mistakenly placed under confirmed facts can distort everything that follows.
How does this relate to Daryl?
Decision assistance is central to what we are building with Daryl.
Daryl is being developed as a personal AI for Mac, with local core conversations, voice and typing, and help examining possible next steps.
Incomplete information is part of ordinary decision-making. The experience needs to help people identify what can be decided, what needs checking, and what remains conditional.
Daryl is still in development. We are not promising flawless handling of uncertainty.
Its behavior needs to be tested against situations where guessing would produce an unsuitable recommendation.
Frequently asked questions
Can AI fill in missing details?
It can suggest possibilities or assumptions. Those should remain clearly labeled until verified.
Should I ask for a confidence percentage?
Only if you understand what the number represents. An unsupported percentage does not resolve missing evidence.
Can I proceed while some information is unknown?
Sometimes. Look for a useful step that remains appropriate across the realistic scenarios.
What is the best question to ask first?
“Which missing fact could change what I should do?”
Turn uncertainty into a next step
Name the gap, identify its effect, and decide how to address it.
You may need a question, a test, a fallback, or a deadline for proceeding.
That is the kind of practical clarity we want Daryl to help people reach.
Meet Daryl and get the launch announcement