Cup of Wit
Writing

How to “Brief” AI Like a Human Teammate (So It Actually Understands the Assignment)

How to "Brief" AI Like a Human Teammate (So It Actually Understands the Assignment)

You wouldn't walk up to a colleague and say: "Write something about business capabilities."

Yet that's exactly how most people prompt AI.

They type vague instructions, hit enter, and then wonder why the output is generic, off-brand, or completely misses the mark. They blame the AI—"it's not smart enough" or "it doesn't understand my industry."

But here's the truth: AI isn't failing. Your brief is.

When creative directors work with copywriters, they don't say "write an ad." They provide a creative brief that defines the audience, objective, tone, constraints, and success criteria. The brief eliminates ambiguity so the writer can focus on execution, not interpretation.

AI is no different. Treat it like a senior contributor who needs a proper brief—and watch your outputs transform from guesswork to precision.

Why Vague Prompts Produce Garbage

When you give AI a vague prompt, you're not being "efficient"—you're asking it to make hundreds of invisible decisions on your behalf:

You say: "Analyze our operational pain points."

AI has to guess:

  • Which operational areas? (All functions or specific capabilities?)
  • What counts as a "pain point"? (Process inefficiency? Technology gaps? Skills shortages?)
  • What framework should I use? (Generic categories or industry-specific taxonomy?)
  • What depth of analysis? (High-level summary or detailed root cause analysis?)
  • What format? (Bullet list, narrative, table, heat map?)
  • What's the purpose? (Executive briefing, detailed diagnosis, or implementation roadmap?)

Every guess AI makes is an opportunity for misalignment. And when the output doesn't match your expectations, you've wasted time—yours on giving feedback, AI's on regenerating.

The solution isn't more powerful AI. It's better briefs.

The Creative Brief Framework for AI

In creative agencies, briefs follow a structured format that leaves zero room for ambiguity. Here's how to adapt that discipline for AI:

Component 1: The Brief (Context & Objective)

What to include:

  • Who is this for? (Audience: C-suite executives, operational managers, technical teams?)
  • What are we creating? (Deliverable type: capability assessment, strategic recommendation, process analysis?)
  • Why does it matter? (Business objective: support budget decision, identify transformation priorities, defend headcount request?)

Why this works: AI understands the purpose behind the task, not just the task itself. This prevents technically correct but strategically useless outputs.

Example:

❌ Vague: "Analyze the customer service function"

✅ Briefed:

textAudience: VP of Operations preparing Q2 budget justification Deliverable: Customer service capability assessment Objective: Identify top 3 capability gaps that justify a $2M technology investment Context: Company is a mid-market insurance provider with 200-person service team handling 50K claims annually

Component 2: The Technique (Tone, Style, Mandatories)

What to include:

  • Tone: Professional/conversational, technical/accessible, formal/direct
  • Style reference: "Match the analytical depth of Example_Assessment.pdf" or "Use the structure from our standard TOM template"
  • Mandatories: Required frameworks (APQC, TOGAF), compliance considerations, terminology to use/avoid

Why this works: You're giving AI unambiguous reference points instead of subjective descriptions.

Example:

❌ Vague: "Write this professionally"

✅ Briefed:

`textTone: Senior consultant addressing C-suite—authoritative but accessible, data-driven but not jargon-heavy Style: Match the structure and analytical depth demonstrated in uploaded file "Example_Maturity_Assessment_Q3_2025.pdf" Mandatories:

  • Use APQC capability framework (uploaded in /sources/)
  • Reference only data from the uploaded client CSV—no external assumptions
  • Avoid technology vendor names (client has procurement sensitivities)
  • Every capability gap must cite specific evidence from process documentation`

Component 3: The Output (Format & Acceptance Criteria)

What to include:

  • Exact format: Markdown table, executive summary, slide deck outline, narrative report
  • Length specifications: Word count, page limits, section breakdowns
  • Acceptance criteria: What must be true for this output to be "done"?

Why this works: You're defining success upfront. AI knows when it's finished, and you have objective criteria to evaluate quality.

Example:

❌ Vague: "Give me a summary"

✅ Briefed:

`textFormat: Markdown table with columns: Capability, Current Maturity (1-5), Evidence, Gap Description, Impact (H/M/L)

Length: 8-12 capabilities maximum, evidence limited to 1-2 sentences per capability

Acceptance Criteria:

  • Every maturity rating must reference specific evidence from the uploaded process documentation (cite page numbers)
  • "Impact" ratings must align with the prioritization rubric in uploaded file "Impact_Scoring_Rules.pdf"
  • Any capability rated below Level 3 must include a gap description that explains why it's below target
  • Final output must be client-ready (no placeholders, no generic consulting-speak)`

The Complete Brief Template

Here's the full framework you can adapt for any task:

`text=== BRIEF === Audience: [Who will consume this? Their role, knowledge level, decision authority] Deliverable: [Specific output type] Objective: [Business outcome this supports] Context: [Background info AI needs: industry, company size, current situation, constraints]

=== TECHNIQUE === Tone: [Authoritative/accessible, formal/conversational, technical/simplified] Style Reference: [Link to example file or describe structural approach] Mandatories:

  • [Required frameworks, terminologies, data sources]
  • [Compliance considerations, sensitivities to avoid]
  • [Evidence requirements—every claim must be cited, quantified, etc.]

=== OUTPUT === Format: [Exact structure: table/narrative/bullets, with column headers if applicable] Length: [Word count, page limit, section breakdown]

Acceptance Criteria:

  • [Criterion 1: e.g., "Every recommendation must include quantified business impact"]
  • [Criterion 2: e.g., "All data must trace back to uploaded source files—no external assumptions"]
  • [Criterion 3: e.g., "Tone must match Example_File.pdf"]
  • [Criterion 4: e.g., "Output must be client-ready with no placeholders"]

=== YOUR SPECIFIC INSTRUCTION === [Now give the specific task: "Assess the maturity of the uploaded process documentation using this brief"]`

Real-World Example: Before and After

Let's see this framework in action for a capability maturity assessment:

Before (Vague Prompt)

text"Analyze the attached process documentation and create a capability maturity assessment."

AI's output:

  • Generic capability list that doesn't match client's organizational structure
  • Maturity ratings with no supporting evidence
  • Recommendations that ignore budget constraints
  • Tone that's either too technical or too simplistic
  • Format that doesn't align with your firm's standards

Result: 90 minutes of back-and-forth edits to get something usable.

After (Briefed Properly)

`text=== BRIEF === Audience: CFO and VP Operations at mid-market insurance company preparing transformation business case Deliverable: Customer Claims capability maturity assessment Objective: Justify $3M investment in claims modernization by demonstrating current-state gaps Context: 200-person claims team, 50K annual claims, legacy systems from 2010, regulatory pressure to improve cycle time from 45 days to 20 days

=== TECHNIQUE === Tone: Senior business architect addressing finance and operations executives—data-driven, pragmatic, focused on ROI Style Reference: Match analytical structure from uploaded "Example_Claims_Assessment_2024.pdf" Mandatories:

  • Use uploaded Custom_Claims_Capability_Framework.pdf (9 capabilities, Levels 1-4 scale)
  • Cite specific evidence from uploaded Process_Documentation_Nov2025.pdf with page numbers
  • Every gap must quantify current vs. target performance using metrics from uploaded KPI_Dashboard.csv
  • Avoid vendor names or solution recommendations (client has separate RFP process)

=== OUTPUT === Format: Markdown table with columns: Capability, Current Level (1-4), Target Level, Evidence (with page citations), Performance Gap (quantified), Business Impact ($)

Length: 9 capabilities (all from framework), evidence limited to 2 sentences max per capability

Acceptance Criteria:

  • Every maturity level must cite page-specific evidence from Process_Documentation_Nov2025.pdf
  • Every performance gap must reference actual metrics from KPI_Dashboard.csv (e.g., "Current: 45-day cycle time, Target: 20 days")
  • Business Impact must show annual cost of gap using labor rates from uploaded Finance_Data.csv
  • Final table must be copy-paste ready for executive deck—no editing required

=== INSTRUCTION === Using the uploaded process documentation, KPI dashboard, and capability framework, create the capability maturity assessment following this brief.`

AI's output:

  • Capability list perfectly aligned with client's framework
  • Every maturity rating supported by page-specific evidence
  • Quantified gaps tied directly to client's actual performance data
  • Business impact calculated using client's real cost structure
  • Format ready for immediate insertion into executive presentation

Result: Client-ready output in one attempt. 10 minutes of review vs. 90 minutes of editing.

Why This Approach Eliminates Revision Cycles

When you brief AI properly, three things happen:

1. AI makes decisions you would make

Because you've explicitly defined audience, objective, and constraints, AI's judgment aligns with yours. It's not guessing—it's following your brief.

2. You evaluate against objective criteria

Instead of "this doesn't feel right," you can check: "Did it meet the acceptance criteria?" If yes, it's done. If no, you know exactly what to fix.

3. AI becomes a reliable contributor

Just like a well-briefed human teammate, AI delivers predictable quality. You stop micromanaging and start leveraging.

Common Briefing Mistakes (And How to Fix Them)

Even when people adopt the framework, they make predictable errors:

Mistake 1: Vague acceptance criteria

❌ "Output should be high quality"

✅ "Every claim must cite a source document. Every recommendation must include quantified ROI. Tone must match Example_File.pdf."

Mistake 2: Missing context

❌ "Analyze customer experience pain points"

✅ "Analyze customer experience pain points for a B2B SaaS company with 500 enterprise clients, average contract value $150K, NPS score 42, targeting improvement to 60+ to reduce churn from 12% to 8%"

Mistake 3: No style reference

❌ "Write in a professional tone"

✅ "Match the structure, sentence length, and data-to-narrative ratio demonstrated in uploaded Example_Report.pdf"

Mistake 4: Format ambiguity

❌ "Summarize this"

✅ "Create a 3-column markdown table: Pain Point | Evidence (cite page #) | Estimated Annual Cost. Maximum 8 rows. Each evidence cell limited to 15 words."

The Time Investment That Pays Dividends

"This seems like a lot of work just to ask AI a question."

You're right—it takes 5-10 minutes to write a proper brief vs. 30 seconds to type a vague prompt.

But here's the math:

Vague prompt approach:

  • 30 seconds to write prompt
  • 60-90 minutes editing AI's output across 4-5 revision cycles
  • Still uncertain if final output meets your quality standard
  • Total time: 90+ minutes

Proper brief approach:

  • 5-10 minutes to write structured brief
  • 5-10 minutes reviewing AI's output against acceptance criteria
  • Client-ready deliverable in first attempt
  • Total time: 15-20 minutes

And here's the compounding benefit: Once you build a brief template for a recurring task (capability assessments, executive summaries, pain point analyses), you reuse it forever. Your 10-minute investment becomes 30 seconds: "Use the Standard_Maturity_Assessment_Brief.txt with this client's data."

When to Brief (And When a Simple Prompt Is Fine)

Not every AI interaction needs a full brief. Use judgment:

Use a simple prompt when:

  • Quick factual lookup ("What's the APQC definition of Supply Chain Planning?")
  • Brainstorming or exploration ("Generate 10 potential names for this capability")
  • Tasks with low stakes (internal notes, personal research)

Use a full brief when:

  • Client-facing deliverables
  • Strategic recommendations that drive decisions
  • Analysis that requires specific frameworks or methodologies
  • Anything where quality inconsistency wastes time

The rule: If you'd brief a human teammate before assigning the task, brief the AI the same way.

The Bottom Line

AI isn't a search engine. It's not a magic genie. It's a capable contributor that performs as well as the brief you give it.

When creative directors treat AI like a junior copywriter—throwing vague instructions and expecting brilliance—they get intern-level output.

When they treat AI like a senior partner—providing structured briefs with clear audience, objective, tone, constraints, and acceptance criteria—they get strategic-level output.

The difference isn't the AI. It's the brief.

Start with one deliverable type you create regularly. Write a full brief using the framework:

  • Brief (audience, objective, context)
  • Technique (tone, style reference, mandatories)
  • Output (format, length, acceptance criteria)

Run the same task twice—once with a vague prompt, once with your structured brief. Compare the outputs.

You'll never go back to vague prompts again.

What's the deliverable type you'd benefit from briefing properly? Have you experimented with structured briefs vs. vague prompts? Drop your experience in the comments.

LINKEDIN POST:

You wouldn't tell a colleague: "Write something about capabilities."

Yet that's exactly how most people prompt AI.

Then they wonder why the output is generic, off-brand, or completely wrong.

The problem isn't the AI. It's the brief.

Senior consultants don't tell analysts "do some research." They provide structured briefs with:

→ Audience (who is this for?)

→ Objective (what decision does it support?)

→ Scope (which capabilities, what frameworks?)

→ Constraints (budget limits, data sources, mandatories)

→ Acceptance criteria (what must be true for this to be "done"?)

The same discipline transforms AI outputs.

The framework (adapted from creative agencies):

BRIEF → Define audience, deliverable, objective, context

TECHNIQUE → Specify tone, style reference, mandatories

OUTPUT → Detail format, length, acceptance criteria

Why this eliminates revision cycles:

  1. AI makes decisions you would make (because you defined the parameters)
  2. You evaluate against objective criteria (not subjective "feel")
  3. AI becomes a reliable contributor (like a well-briefed teammate)

When to use a full brief:

✓ Client-facing deliverables

✓ Strategic recommendations

✓ Analysis requiring specific frameworks

✓ Anything where quality inconsistency wastes time

Simple prompts are fine for quick lookups or brainstorming. But if you'd brief a human teammate before assigning the task, brief the AI the same way.

I wrote the complete briefing framework—including the exact template and real before/after examples:

📖 https://open.substack.com/pub/cupofwit/p/how-to-brief-ai-like-a-human-teammate?r=59sawq&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Question: Are you still using vague prompts or structured briefs? What deliverable would you benefit from briefing properly?

SUBSTACK NOTE:

You wouldn't tell a colleague: "Write something about capabilities."

Yet that's how most people prompt AI. Then wonder why outputs are generic or wrong.

The problem isn't the AI. It's the brief.

Senior Consultants use structured briefs:

→ Audience (who is this for?)

→ Objective (what decision does it support?)

→ Tone + constraints + acceptance criteria

The same discipline transforms AI outputs.

Vague prompt:

"Analyze process documentation and create a maturity assessment"

Result: 90 minutes editing generic output

Structured brief:

Audience: CFO preparing $3M business case

Framework: Custom_Claims_Framework.pdf

Format: Table with evidence citations

Acceptance: Every gap must quantify using actual KPIs

Result: Client-ready in one attempt. 15 minutes total.

The complete briefing framework (with template and examples): https://open.substack.com/pub/cupofwit/p/how-to-brief-ai-like-a-human-teammate?r=59sawq&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true