The ONHT Framework: Beginners Guide

Summary: Stop getting generic AI responses. Learn the four-letter framework that transforms vague requests into precise results. The ONHT framework: Objective (what problem you're solving), Needs (key information that matters), How (the thinking approach), and Trajectory (clear steps to the answer), teaches you to think WITH AI, not through it, turning "analyse customer feedback" into board-ready insights. Real examples show how adding context and structure gets you from Level 1 basics to Level 3 mastery, where AI delivers exactly what you need. The difference? Knowing how to ask.

Your First Step Towards AI Mastery

Here's the truth. Right now, someone with the same AI tools as you is getting 10x better results.
Not because they're smarter. Because they know how to ask.
This guide teaches you that skill. In 20 minutes, you'll go from getting generic fluff to getting exactly what you need.

Why So Many AI Prompts Fail (And How to Fix Them)

Picture this. Sarah from marketing needs customer insights. Tomorrow's board meeting. She types: "analyse customer feedback."
The AI responds with 10 pages. Generic. Useless.
Why? Nobody told the AI what Sarah actually needed.
Three bullet points. For non-technical board members. About why sales dropped. That's what she needed.
The ONHT framework fixes this. Let's learn how.

What is ONHT?

Think of ONHT as a conversation blueprint. Four simple parts:

O - Objective: What problem are we solving?
N - Needs: What information matters?
H - How: How should the AI think?
T - Trajectory: What path gets us there?

That's it. Four letters. Massive difference.

Three Rules Before We Start

1. Real People Have Real Problems

Sarah doesn't need "customer sentiment analysis."
She needs three points explaining why sales dropped. In plain English. For tomorrow.
Always start with the human need. Not the AI capability.

2. Context Beats Commands

Bad: "Analyse data"
Good: "I have 10 minutes to explain to my CEO why our sales dropped 15%"
See the difference? Context gives the AI everything it needs to help.

3. First Drafts Are Starting Points

Nobody gets it perfect first time. That's normal.
Plan to refine. Each round gets better.

Breaking Down ONHT

ONHT Framework - Objective - Needs - How - Trajectory

O - Objective - Start with Why

Define what you're really trying to achieve. Start with why it matters. That tells you what to create. Always specify who needs it and what they'll do with it.

Bad objective:
"Generate a report on customer feedback"

Good objective:
"Help me identify the top 3 reasons customers cancelled last month, so I can present solutions to the board in simple terms"

See the difference?
The good example has a real person with a real deadline and a specific need. The AI now knows exactly how to help.

Levels 1 to 3 of Examples
Now to help you understand the layers that can be used with GenAI, we will share here 3 examples over additional layers to explain the objective from Level 1 to Level 3. As you progress you will see how giving more details will help clarify your objective:

Level 1: Basic Objective

Example:
"Help me explain why sales dropped to my boss"

This tells AI the task. But misses crucial details.

Level 2: Add Key Detail to Objectives

Example:
"Help me explain why Q3 sales dropped 15% to my CEO during tomorrow's board meeting. She needs three clear points in simple language. I have 5 minutes to present."

Now AI knows:

Specific problem (15% drop)
Audience (CEO, board)
Format (three points)
Constraints (5 minutes, simple language)

Level 3: Complete Objective Details

Example:
"I'm the Sales Director preparing for tomorrow's 9am board meeting. Sales dropped 15% in Q3 (£2.3M to £1.95M). The CEO expects clear explanations and solutions. The board includes non-technical members who make funding decisions. I need:

3 bullet points for a PowerPoint slide
Each point max 15 words
Focus on fixable problems
Include one quick win we can implement this month
Tone: confident but not defensive"

See how Level 3 gives AI everything about your Objectives? It knows who you are, who's listening, what format you need, even the emotional tone required.

N - Needs - Define what Information Matters

Give the AI the right facts. Not all the facts. Think: what would someone need to know to solve this specific problem? Not everything. Just what's relevant.

Bad needs statement:
"List all the functions in group A with their descriptions."

Good needs statement:
"Draft the base structure of a PRD using the functions in group A. Group the functions by user journey, with enough context for designers and engineers to expand from it."

See the difference?
These two good narratives frame the output around real human goals aiming for collaboration and usability, not just raw data extraction.

Additional guidance: A CFO thinks about quarters and budgets. A developer thinks in sprints and features. Same data, different stories.

Level 1: Basic Facts

Example:
"Sales are down. Customers are complaining about price."
Minimal context. AI will give generic responses.

Level 2: Key Data Points

Example:
"Q3 sales: £1.95M (down from £2.3M in Q2) Main complaints:

Pricing too high (45% of feedback)
Competitor has better features (30%)
Support response slow (20%)"

Better. AI can now identify patterns and priorities.

Level 3: Rich Context

Example:
"Q3 Performance:

Revenue: £1.95M (down 15% from Q2's £2.3M)
Lost 47 enterprise accounts (worth £420K annually)
New customer acquisition down 30%

Exit Interview Data (sample: 127 customers):

Price concerns: 45% (average comment: "35% more than competitor")
Feature gaps: 30% (main ask: mobile app, real-time sync)
Support issues: 20% (average response time: 48hrs vs promised 4hrs)
Other: 5%

Market Context:

Main competitor launched freemium tier in Month 2 of Q3
Economic downturn hit our retail sector hard
Our last price increase: 18 months ago (+12%)

Internal factors:

Support team reduced by 3 people in Q3 (budget cuts)
Mobile app delayed again (now 6 months late)
Sales team morale low after commission structure change"

Level 3 provides complete context for the Needs. AI can now give you nuanced, actionable insights.

H - How: Guide the Thinking

Tell the AI how to approach the problem. Here's where you set the mindset. Think of it as hiring a consultant. Do you want the mind of a brutal cost-cutter or the creative strategist? Tell the AI which hat to wear.

Bad approach:
"Tell me what to do"

Good approach:
"Think like a board member: What's the financial impact? Is this fixable? What's the cost of fixing vs not fixing? No technical jargon."

See the difference?
The good example gives the AI a clear thinking framework. The output will match what board members actually need.

Level 1: Basic Instruction

Example:"Think like a business person"
Too vague. Could mean anything.

Level 2: Specific Perspective

Example:"Think like an experienced Sales Director who needs to maintain credibility while admitting problems. Consider what the board cares about: protecting revenue and maintaining growth."
AI now understands the professional lens and key concerns.

Level 3: Detailed Thinking Framework

Example:"Approach this like a McKinsey consultant would:

Start with the financial impact to get attention
Identify root causes, not symptoms
Focus on what we control (not market conditions)
Present solutions with clear ROI
Acknowledge mistakes without appearing weak
Use board-friendly language (ROI, market share, competitive advantage)
Balance honesty with confidence
Remember: the board wants to know we can fix this, not hear excuses"

Level 3 provides a complete Thinking framework. AI will structure its response like a top consultant would.

T - Trajectory - Map the Journey

Break complex requests into clear steps. Start simple, then build.
Here's the magic: you control the journey. Not just what information appears, but how it unfolds. Need to soften bad news? Start with context. Selling an idea? Build to the big reveal. Trajectory turns information into communication.

Bad trajectory:
"Give me everything about our customers"

Good trajectory:
"First, summarise the main issue in one sentence. Then explain the biggest cause. Follow with 2-3 solutions we could implement. End with your recommended next step."

See the difference?
The good example creates a logical flow. The AI delivers information in the order you need it.

Level 1: Simple Steps

Example:"First explain the problem, then give solutions"
Basic structure, but lacks detail.

Level 2: Clear Sequence

Example:"Structure the response as:

One-sentence problem summary
Three main causes (most to least important)
Three solutions (quickest to longest-term)
Recommended first action

AI now has a clear path to follow.

Level 3: Detailed Journey

Example:
"Build the response in this exact structure:

Executive Summary (one sentence, 20 words max)
o Start with financial impact
o End with 'but it's fixable'
Three Root Causes (as bullet points for PowerPoint)
o Each point: 10-15 words
o Format: 'Problem: Impact'
o Order by ease of fixing (easiest first)
o Use specific numbers where possible
Three Solutions (matching the three problems)
o Each solution: 10-15 words
o Format: 'Action: Result'
o Include timeline (30/60/90 days)
o Focus on what we control
Quick Win Recommendation (one specific action)
o Something visible we can do this week
o Low cost, high impact
o Name who owns it
Confidence Closer (one sentence)
o Acknowledge challenge but emphasise capability
o End on positive trajectory"

Level 3 creates a precise roadmap, a Trajectory to achieve our end goal. Every element has purpose and parameters.

Some ONHT Level 1 to 3 Examples

Let's look at some examples linked to unhappy customers and how we can address the issues from Level 1 to Level 3 examples.

Example 1: Email to Angry Customer

Beginner Level 1 Prompt:

[OBJECTIVE]
Reply to angry customer email

[NEEDS]
Customer paid £500, product broke after 2 days

[HOW]
Be professional

[TRAJECTORY]
Apologise and offer solution

Result: Generic apology that might make things worse.

Beginner Level 2 Prompt:

[OBJECTIVE]
Reply to angry premium customer threatening social media complaint

[NEEDS]
- Customer: Tech startup CEO, 50K Twitter followers
- Product: £500 enterprise software licence
- Issue: Software crashed during important demo
- Previous contact: None (straight to CEO email)

[HOW]
Be empathetic but professional. Acknowledge frustration without admitting fault yet.

[TRAJECTORY]
1. Acknowledge their frustration
2. Take it seriously
3. Offer immediate help
4. Provide direct contact

Result: Better response that shows understanding.

Beginner Level 3 Prompt:

[OBJECTIVE]
Draft urgent email response to prevent social media crisis. Customer is influential tech CEO threatening to tweet about our "terrible" product. Need to send within 30 minutes. Goal: Turn them into advocate, not detractor.

[NEEDS]
- Customer: James Chen, CEO of StartupTech (50K Twitter followers)
- Product: £500/month Enterprise Plan (high-value customer)
- Issue: Software crashed during pitch to investors yesterday
- Complaint: Emailed our CEO directly at 8am today (it's now 10:30am)
- Tone: Furious - "This is unacceptable", "Will tell everyone"
- Background: Been a customer for 3 months, normally happy
- Our context: We had server issues yesterday 2-4pm (his demo was at 3pm)

[HOW]
Channel the mindset of a crisis management expert who:
- Takes full ownership without legal admissions
- Transforms anger into partnership
- Knows every minute counts before they tweet
- Balances corporate needs with human empathy
- Understands executive ego and public image concerns

[TRAJECTORY]
1. **Subject Line** - Something that makes them want to read, not rage-tweet

2. **Opening** (1-2 sentences)
   - Personal, immediate acknowledgment
   - Show we understand the gravity

3. **Take Ownership** (2-3 sentences)
   - No excuses, but context if helpful
   - Focus on their experience, not our systems

4. **Immediate Action** (bullet points)
   - What we're doing RIGHT NOW
   - Specific compensations
   - Direct line to senior tech support

5. **Partnership Pivot** (1-2 sentences)
   - Position them as valued advisor
   - Invite them to help us improve

6. **Close** (1 sentence + signature)
   - Personal mobile number
   - Commitment to follow up today

Format: Email, max 200 words, readable on phone

Result: Crisis-preventing email that could turn enemy into advocate.

Example 2: Budget Presentation

Beginner Level 1 Prompt:

[OBJECTIVE]
Create budget presentation

[NEEDS]
Need 20% more budget for marketing

[HOW]
Make it convincing

[TRAJECTORY]
Show why we need it
Result: Generic slides that won't move anyone.

Beginner Level 2 Prompt:

[OBJECTIVE]
Convince CFO to approve 20% marketing budget increase for next quarter

[NEEDS]
- Current budget: £100K/quarter
- Requesting: £120K/quarter
- Reason: Competitor taking market share
- CFO: Very numbers-focused, sceptical of marketing

[HOW]
Think like a CFO - focus on ROI and financial metrics

[TRAJECTORY]
1. Current situation
2. Financial impact of not acting
3. Expected ROI from investment
4. Risk mitigation

Result: Solid presentation with financial focus.

Beginner Level 3 Prompt:

[OBJECTIVE]
Design PowerPoint presentation to secure 20% marketing budget increase from sceptical CFO. Meeting: Friday 2pm (preparing Wednesday). Need to send slides by Thursday noon for pre-read. Success = walk out with approval.

[NEEDS]
Current Situation:
- Q3 marketing budget: £100K
- Q3 results: 200 leads, 20 customers, £400K revenue
- Cost per acquisition: £5K
- Customer lifetime value: £50K

Competition:
- Main competitor increased marketing spend 40% in Q3
- They took 5 of our target accounts (worth £300K)
- Their new campaign launching next month
- We're losing "share of voice" online (down from 35% to 22%)

Our Request:
- Q4 budget: £120K (+£20K)
- Plan: Digital campaign + 2 events
- Expected: 280 leads, 28 customers, £560K revenue
- Break-even: Need just 2 extra sales

CFO Context:
- Name: Sandra Williams, been CFO 2 years
- Style: Ex-Deloitte, loves spreadsheets
- Pet peeve: Fuzzy marketing metrics
- Current focus: Improving EBITDA by 10%
- Recently rejected IT's budget increase

[HOW]
Think like a CFO turned CMO:
- Lead with numbers, not narratives
- Show risk of NOT investing
- Every slide must have ROI calculation
- Use their language: payback period, IRR, EBITDA impact
- Anticipate objections with data
- Make it easy to say yes

[TRAJECTORY]
Slide Structure (5 slides max):

1. **Executive Summary** (1 slide)
   - The ask: £20K for 40% revenue increase
   - Payback: 6 weeks
   - Risk of inaction: £300K+ revenue loss

2. **The Threat** (1 slide with graph)
   - Competitor spending vs market share
   - Project forward 6 months
   - Revenue at risk calculation

3. **The Opportunity** (1 slide with table)
   - Investment vs return breakdown
   - Conservative/Expected/Optimistic scenarios
   - All show positive ROI

4. **The Plan** (1 slide with timeline)
   - Specific campaigns with metrics
   - Weekly tracking commitment
   - Kill switches if not performing

5. **The Ask** (1 slide)
   - Approval for £20K
   - Success metrics agreed
   - Review date set

Design: Clean, professional, lots of white space, charts not text

Result: CFO ready presentation that takes into account the objective and expectations, target audience, specific level of detail and with an output that should be convincing, in a polished presentation.

One additional example

An extra example for Sarah: the board presentation she's worried about:

[OBJECTIVE]
"I need to explain to our board why customer churn increased 23% last quarter. They're not technical. I have 10 minutes. They'll want solutions, not just problems."

[NEEDS]
"Last quarter: 2,300 cancellations (up from 1,870). Exit surveys show pricing concerns (45%), competitor features (30%), support issues (20%), other (5%). Our main competitor launched a freemium tier in Month 2."

[HOW]
"Think like a board member: What's the financial impact? Is this fixable? What's the cost of fixing vs not fixing? Keep explanations at executive level - no technical jargon."

[TRAJECTORY]
1. State the problem with financial impact
2. Explain the main driver  
3. Give 2-3 actionable solutions
4. Recommend next step with timeline

The result? Sarah gets exactly what she needs. Not a research paper.

Making It Stick: The Human-in-the-Loop

Here's where most frameworks fail. They assume once you prompt, you're done.

Wrong. Here's the secret. You're not done after one prompt.

Real work happens in loops:

Check: Does this solve my actual problem?
Adjust: "Make it shorter" or "Add more data"
Validate: "Would my target audience actually read this?"
Refine: "Perfect, just change the opening"

This isn't failure. This is the process.

Refinement Examples

After Level 1 Output: "Too generic. Add specific examples from our industry."

After Level 2 Output: "Good structure, but tone is too formal for our startup culture. Make it more conversational."

After Level 3 Output: "Excellent, but can you make the financial projections more conservative? Our CFO always cuts numbers by 20%. Explain this has been done as it not therefore needed"

Format Considerations: Where Will This Live?

Always specify your output format:

For Email:

"Max 150 words, readable on mobile"
"Short paragraphs, no bullet points"
"Subject line that prevents deletion"

For PowerPoint:

"Bullet points, max 5 per slide"
"Each point under 15 words"
"Speaker notes with details"

For Word Document:

"Executive summary at top"
"Clear headers for scanning"
"Page numbers and professional formatting"

For Slack/Teams:

"Casual tone, use emojis where appropriate"
"Break into multiple short messages"
"Include TL;DR at start"

Common Beginner Mistakes

Mistake 1: Forgetting The Audience

Wrong: "Write a report about project delays"
Right: "Write a 1-page report for the steering committee (CEOs who want solutions, not technical details) explaining why the project is 2 weeks late, key issues and risks, how we'll recover, what the next steps are and who is responsible for each step"

Mistake 2: Missing Emotional Context

Wrong: "Email client about contract renewal"
Right: "Email our nervous client Alison Dentfort who's been a client of ours for 4 years but is considering switching to competitor about contract renewal. She is upset about last month's outage that put her in a difficult position at her company. Need to rebuild trust while aiming at securing a 2-year commitment."

Mistake 3: No Success Criteria

Wrong: "Help me with my presentation"
Right: "Help me create 10-minute presentation that convinces board to approve £2M investment. Success = they approve it same day."

Note: in the Trajectory of aiming to get the perfect outcome, you need to provide details about the format, the emotional context but also as part of the prompt, the same way you may say to a client "What does a succesfult outcome for you look like?" you can also explain what you consider to your new AI sidekick, what success looks like for you.

Your Practice Exercise

Try this yourself. Think of a real task you have.

Now write:

Objective: What you need and why
Needs: Key information (5-6 facts maximum) layer them by priority
How: How to think about it
Trajectory: 3-4 clear steps

Remember: specific beats clever every time.

Take a real task you have tomorrow. Write three versions:

Level 1: Just the basics

What you need
Who it's for
Basic approach

Level 2: Add crucial details for each area of the prompt: Objectives / Needs / How / Trajectory

Specific context
Key constraints
Clear structure

Level 3: Complete blueprint: bring it home and marvel at the results you get through iterations

Full background
Detailed requirements
Exact success criteria
Format specifications
Emotional considerations

Watch how your outputs improve with each level.

What Happens Next?

Master these basics first. Get comfortable with ONHT. Start with Level 1, then push yourself to Level 2, then Level 3.

Within a week, Level 3 thinking becomes natural.

Then you'll discover something powerful. You're not just getting better at AI.

You're getting better at:

Clarifying your thinking
Understanding your audience
Communicating clearly and precisely
Solving problems systematically

These skills transfer everywhere. Every tool. Every role. Every company.

That's your real competitive advantage.

Take one real task. Right now.
Apply ONHT. See what happens.

Then tomorrow, try another.Within a week, you'll see the difference. Within a month, others will too.

If you want to know more, move on to the Intermediate level and start using some of the next level steps.

Remember: The tool might change. The company might change. But these skills? They're yours forever.

Key Takeaways for Beginners

Start with the human need, yours or the end-user, not the AI capability
Be specific about your Objectives context, constraints, and audience
Use simple, clear language - fancy prompts aren't better prompts
Expect to refine. First drafts are for direction, not perfection. It's part of the process. Part of the Trajectory
You're learning to think WITH AI, not through it

Level 1 gets you started, but rarely gets great results
Level 2 adds the context AI needs to be helpful
Level 3 provides complete blueprints for exceptional outputs

Always specify format, audience, and success criteria
Emotional context often matters as much as facts
Plan to refine - even Level 3 prompts improve with iteration

Your Next Step

Take one real task. Right now.
Write a Level 1 version. See what you get.
Upgrade to Level 2. Notice the improvement.
Push to Level 3. Experience the transformation.
Then tomorrow, start at Level 3.

Within a week, you'll be getting outputs that make colleagues ask: "How did you do that?"

Remember:
AI is a thinking partner, not a magic oracle. The clearer you are about what you need, the more helpful it becomes.
AI amplifies what you bring. Bring clarity, get clarity. Bring confusion, get confusion.
And the tool might change. The company might change. But these skills? They're yours forever.

Coming up next:
The ONHT Framework for Intermediate users
The ONHT Framework for Advanced users
Some ONHT Tips and Supporting Prompts
Key Advanced Features of the ONHT Framework

Article written by John Garner