Prompting Guidelines

Decompose

Break the request into a task list that Kris should explore and perform before providing an answer.

Artifacts

Add the source material for Kris to mirror results and structure, such as sample emails or generated reports.

Context

Explain the project, date range, audience, constraints, and decision you need to make.

The DAC Template

Copy this structure when you need a reliable answer:

Open AP/AR Aging
RFI Summary
Schedule Risk

Goal:
Show open accounts payable and accounts receivable by aging bucket.

DAC:
1. Decompose the analysis into AP invoices, AR invoices, open balances, aging buckets, project/vendor/customer breakdowns, and exclusions.
2. Use these artifacts: AP invoice table, AR invoice table, payment records, vendor list, customer list, and project list.
3. Context: This is for a construction finance review. Bucket open invoices into 0-30, 31-60, 61-90, and 90+ days. Exclude void invoices from every total.

Output:
Return a table with columns: Aging Bucket, Open AP, Open AR, Net Position, Invoice Count, and Notes. Then list any assumptions, including whether aging is based on due date or invoice date.

Goal:
Summarize RFI 042 so the PM can prepare a response to the architect.

DAC:
1. Decompose the issue into design intent, affected drawings/specs, impacted trades, schedule risk, and open questions.
2. Use these artifacts: RFI text, drawing A-301, spec section 07 21 00, and the May 12 coordination meeting notes.
3. Context: This is an occupied hospital renovation. Work is in Area B. The response should help the PM ask the design team for a clear decision.

Output:
Return 5 bullets: issue, likely interpretation, impacted trades, risks, and questions to ask. Cite the artifact used for each point.

Goal:
Identify the top schedule risks for next week's OAC meeting.

DAC:
1. Decompose the analysis into critical activities, delayed predecessors, missing decisions, procurement risks, and trade coordination issues.
2. Use these artifacts: current P6 lookahead, RFI log, submittal log, procurement log, and last week's meeting minutes.
3. Context: Focus on work planned in the next 3 weeks. Audience is owner, architect, and GC leadership.

Output:
Return the top 5 risks with owner, evidence, likely impact, and recommended next action.

If the answer matters for cost, schedule, safety, contract terms, or client communication, ask the AI to list assumptions and missing information before giving the final recommendation.

Best Practices

1. Ask for the thinking path, not hidden reasoning

Use a clear process:

Before answering, break this into the checks you need to perform. Then give the final answer in the requested format.

This helps the AI avoid jumping straight to a confident but incomplete answer.

2. Name the artifacts clearly

Be specific about what the AI should use:

Instead of	Use
”Look at the project docs"	"Use RFI log, submittal log, drawing A-401, and spec section 09 91 00"
"Analyze costs"	"Use the change event export, commitment log, and approved budget view"
"Check schedule"	"Use the current P6 file, 3-week lookahead, and procurement log”

3. Define the output before the AI starts

Construction teams usually need decision-ready formats:

A table for reviews, comparisons, and logs
A short narrative for owner updates or change events
Bullets for meeting prep
A checklist for field follow-up
A risk register for leadership review

4. Set boundaries

Tell the AI what not to do:

Do not invent contract requirements.
Do not include void invoices in aging totals.
Do not make legal conclusions.
Do not assume schedule impact unless it is supported by the artifacts.

5. Require missing information

Good construction answers should show gaps:

If information is missing, list what is missing, why it matters, and who likely owns the answer.

6. Use examples when tone matters

If you want the AI to match your company’s communication style, provide 1-2 examples:

Match this tone: factual, concise, and owner-facing.

Example:
"The team is reviewing the added scope against the current drawings and will confirm cost and schedule exposure after subcontractor pricing is received."

Prompt Quality Checklist

Before sending a prompt, check for these five items:

Decision

Is it clear what decision, draft, summary, or analysis you need?

Artifacts

Did you include or name the source records the AI should rely on?

Context

Did you include project, trade, dates, audience, and constraints?

Format

Did you define the output format?

Gaps

Did you ask for assumptions, missing information, and next steps?

Common Fixes

If the answer is…	Improve the prompt by…
Too generic	Add project context, role of the audience, and source artifacts
Too long	Specify the exact number of bullets, rows, or paragraphs
Too confident	Ask for assumptions, evidence, confidence level, and missing information
Missing key records	Name the exact RFIs, specs, drawings, logs, or tables to use
Hard to act on	Request owner, due date, risk, and next action columns

A Strong Starting Prompt

I need help preparing for a construction project review.

Goal:
Identify the most important open issues that could affect cost, schedule, or owner decisions.

DAC:
1. Decompose the review into RFIs, submittals, procurement, schedule, change events, and field coordination.
2. Use these artifacts: [paste or attach current logs, meeting notes, schedule export, and cost report].
3. Context: [project type], [phase], [audience], [date range], [known constraints].

Output:
Return a table with columns: Issue, Source Artifact, Why It Matters, Owner, Due Date, Recommended Next Action. Then list assumptions and missing information.

Only share project data in tools approved by your company. For sensitive contract, personnel, financial, or safety information, follow your internal data handling policies.

Using the App

Analytics

Permissions

Integrations

Prompting Guidelines

Decompose

Artifacts

Context

The DAC Template

Best Practices

1. Ask for the thinking path, not hidden reasoning

2. Name the artifacts clearly

3. Define the output before the AI starts

4. Set boundaries

5. Require missing information

6. Use examples when tone matters

Prompt Quality Checklist

Common Fixes

A Strong Starting Prompt

Decompose

Artifacts

Context

​The DAC Template

​Best Practices

​1. Ask for the thinking path, not hidden reasoning

​2. Name the artifacts clearly

​3. Define the output before the AI starts

​4. Set boundaries

​5. Require missing information

​6. Use examples when tone matters

​Prompt Quality Checklist

​Common Fixes

​A Strong Starting Prompt

The DAC Template

Best Practices

1. Ask for the thinking path, not hidden reasoning

2. Name the artifacts clearly

3. Define the output before the AI starts

4. Set boundaries

5. Require missing information

6. Use examples when tone matters

Prompt Quality Checklist

Common Fixes

A Strong Starting Prompt