Beyond “Should We Use AI?”: An AI Decision Framework (AIDF) for Modern Professionals

Introduction

Over the last few years, Artificial Intelligence has become part of our daily work. Whether you’re a software tester, developer, product manager, marketer, HR professional or business leader, you’ve probably heard questions like:

• “Can AI do this?”
• “Should we automate this?”
• “Let’s ask ChatGPT.”

But I’ve noticed something interesting.

Most organisations approach AI from one of two extremes.

One group wants AI to do everything.

The other doesn’t trust AI at all.

In my opinion, both approaches miss the point.

The real question isn’t:

“Can AI do this?”

The better question is:

“What role should AI play in this task?”

That single shift in thinking changes how we collaborate with AI.

While revisiting Ethan Mollick’s excellent book Co-Intelligence, particularly Chapter 3, I realised it provides outstanding principles for working with AI. However, I wanted a practical framework that any professional could use to decide when AI should lead, assist, or stay out of the task altogether.

That led me to develop the AI Decision Framework (AIDF).

The Inspiration: Four Rules for Co-Intelligence

Before introducing the framework, it’s worth acknowledging the ideas that inspired it.

In hi book Co-Intelligence, Ethan Mollick presents four simple but powerful rules.

1. Always Invite AI to the Table

Before starting any task, ask whether AI could contribute.

Not because AI will always be the best option—but because it often offers ideas, speed or perspectives you might not have considered.

2. Be the Human in the Loop

AI generates.

Humans decide.

No matter how capable AI becomes, responsibility still belongs to us.

3. Treat AI Like a Person (But Tell It What Kind of Person)

Better context produces better results.

Instead of vague prompts, provide:

• Context
• Objectives
• Constraints
• Expected output
• Audience

The quality of AI’s output often reflects the quality of the instructions it receives.

4. Assume This Is the Worst AI You Will Ever Use

AI is evolving rapidly.

Tasks that seem impossible today may become routine tomorrow.

Rather than treating AI as a fixed capability, we should continually reassess where it can add value.

See content credentials

The Missing Piece

While reading these principles, one question kept coming to mind.

How do we actually decide whether AI should be involved in a task?

The principles explain how to collaborate with AI.

They don’t provide a repeatable decision-making process.

So I created one.

Introducing the AI Decision Framework (AIDF)

The purpose of this framework isn’t to determine whether AI can perform a task.

Instead, it helps answer a more practical question:

What is the appropriate role for AI in this task?

The framework guides professionals through a series of questions that evaluate the nature of the work before assigning one of four AI Role Levels.

The Framework

Rather than making assumptions, ask these questions in order.

1. Can AI reasonably help?

Would AI make the task faster, easier or more effective?

If the answer is no, there’s little reason to involve AI.

2. Is the task repetitive or execution-focused?

Tasks involving:

• Formatting
• Summarisation
• Documentation
• Data processing
• Standardised outputs

are often excellent candidates for AI.

These activities consume time but usually require limited judgement.

3. Does the task require creativity?

Many people assume creativity belongs exclusively to humans.

In reality, AI excels at brainstorming.

Generating ideas is different from choosing the best one.

AI can propose dozens of possibilities.

Humans determine which are valuable.

4. Does the task require deep business judgement or domain expertise?

Some decisions depend on organisational priorities, customer relationships, regulations or years of experience.

Examples include:

• Release approval
• Pricing strategy
• Hiring decisions
• Legal interpretation
• Risk acceptance

AI can provide insight.

Humans remain accountable.

5. Are the consequences of mistakes significant?

Not every mistake carries the same impact.

An incorrect meeting summary is inconvenient.

An incorrect medical recommendation could be catastrophic.

The greater the consequences, the greater the need for human oversight.

6. Can the output be easily verified?

This may be the most important question in the framework.

If AI generates an answer that cannot be validated with reasonable confidence, it should never become the final authority.

Trust should always be proportional to our ability to verify.

See content credentials

The Four AI Role Levels

One of the biggest misconceptions surrounding AI is that tasks are either “AI” or “Human.”

Reality is much more nuanced.

The framework identifies four distinct levels of AI involvement.

Level 1 — Human Only

AI has little or no role.

These tasks require:

• Ethics
• Accountability
• Empathy
• Final approval
• Strategic judgement

Examples:

• Release approval
• Legal agreements
• Performance reviews
• Security sign-off
• Executive decisions

Level 2 — AI Assistant

This is where most knowledge work belongs today.

AI contributes ideas, drafts or recommendations.

Humans create, review and own the final outcome.

Examples:

• Writing documentation
• Bug reports
• Requirement analysis
• Test case generation
• Research summaries
• Presentation drafting

Level 3 — AI First

AI performs the first iteration.

Humans review, refine and approve.

This dramatically reduces effort without removing accountability.

Examples:

• Meeting summaries
• Data analysis
• SQL generation
• Email drafting
• Translation
• Standard reports

Level 4 — AI Automation

AI performs the task autonomously while humans monitor exceptions.

Suitable for highly repetitive, rule-based activities.

Examples:

• Automated testing
• CI/CD notifications
• Monitoring systems
• Duplicate detection
• Build validation
• Scheduled reporting

See content credentials

How to Use the Framework

The framework is intentionally simple.

For every new task:

Step 1: Define the task clearly.

Step 2: Walk through each decision question.

Step 3: Stop when the answers naturally indicate the appropriate AI Role Level.

Step 4: Assign AI accordingly.

Step 5: Periodically revisit the decision as AI capabilities continue to evolve.

Remember:

The goal isn’t maximum automation.

The goal is appropriate collaboration.

See content credentials

See content credentials

The Golden Rules of Co-Intelligence

Regardless of the role AI plays, I always come back to the principles that inspired this framework.

✔ Always invite AI to the table.

✔ Be the human in the loop.

✔ Treat AI like a knowledgeable colleague by providing clear context.

✔ Assume today’s AI is the least capable AI you’ll ever use.

These four rules ensure that AI remains a collaborator—not a replacement.

The Golden Principle

If I had to summarise this entire framework in a single sentence, it would be this:

AI amplifies capability. Human judgement creates value.

AI can generate.

AI can analyse.

AI can automate.

But humans provide context.

Humans understand consequences.

Humans make decisions.

Technology should strengthen human capability—not replace human responsibility.

Final Thoughts

The conversation around AI often focuses on one question:

“Can AI do this?”

I believe we should start asking a different one.

“What role should AI play?”

That subtle shift changes everything.

It moves us away from fear.

Away from hype.

Away from blind automation.

Instead, it encourages thoughtful collaboration—placing humans and AI where each delivers the greatest value.

The future isn’t Human vs AI.

It’s Human + AI, working together with the right balance of automation, judgement and accountability.

And perhaps that’s the most important decision framework we’ll need in the age of AI.

Recommended Reading

This framework was inspired by the ideas presented in Ethan Mollick’s Co-Intelligence: Living and Working with AI, particularly the chapter “Four Rules for Co-Intelligence.” If you’re interested in developing a practical and balanced approach to working with AI, I highly recommend reading the book. It explores not only what AI can do today, but also how humans can collaborate with it effectively as its capabilities continue to evolve.