Skip to main content

Explainer · 4D framework

AI fluency, explained.

AI fluency is the ability to work with AI well: knowing what to hand over, communicating clearly, judging what comes back and using it responsibly. This is the reference behind GAFI™ and its provisional assessment result.

9 minute read · Updated August 2026 · Grounded in the 4D AI Fluency Framework

The idea

Using AI is common. Using it well is rarer.

Most people can open an AI tool and get an answer. Fluency is the harder workplace capability: choosing the right task, asking in a way that gets a useful result, and knowing when the answer is wrong. It can be learned, demonstrated and improved.

AI fluency is

  • Deciding what to hand to AI and what to keep yourself
  • Getting a useful result through clear context and iteration
  • Spotting when an answer is confident but wrong
  • Being open about where AI was used and checking it first

AI fluency is not

  • Knowing the most tools or the newest model
  • Accepting the first answer because it looks polished
  • Letting AI make the call on something that matters
  • Hiding that AI was involved

The 4D framework

Four capabilities that add up to fluency.

The 4D AI Fluency Framework was developed by Professors Rick Dakan and Joseph Feller in collaboration with Anthropic. Each capability feeds the next, creating a practical loop for human and AI collaboration.

01

Delegation

Knowing what to hand to AI and what to keep yourself.

Strong looks like: You match the task to the tool and keep the decisions that need human judgement.

02

Description

Communicating clearly with AI through prompts and iteration.

Strong looks like: You provide useful context, examples and constraints, then refine the work deliberately.

03

Discernment

Critically evaluating AI output for errors, bias and trust.

Strong looks like: You question polished answers, check important claims and notice missing context.

04

Diligence

Verifying, documenting, disclosing and iterating.

Strong looks like: You protect sensitive information, disclose AI use and keep a human accountable.

Source: Anthropic, AI Fluency Framework and Foundations.

What it looks like

One task, handled two ways.

Your manager sends 40 open-text customer survey responses on Tuesday and asks for the themes by Thursday morning. The tool and deadline are identical. The working habits are not.

Around L1

Using AI

Delegation

Pastes everything in, including identifying details, and asks for a summary.

Description

Uses one short prompt with no audience, format or decision context.

Discernment

Moves polished themes into the deck without checking the responses.

Diligence

Leaves no record of how the work was made or checked.

Around L4

Working with AI

Delegation

Removes identifying details, delegates clustering and keeps the business judgement.

Description

Provides the audience, decision and examples, then refines the format.

Discernment

Checks every theme against source responses and removes weak claims.

Diligence

Discloses AI clustering, records human review and keeps traceable working files.

The five levels

Fluency grows in stages, from aware to architect.

The curve is directional, not a ranking of people. Each level describes a practical shift in working habits and builds on the one before it.

L1

L2

L3

L4

L5

L1

Aware

You are using AI, and building more deliberate habits.

You have used AI, but structured delegation, prompting, checking and disclosure are not yet consistent.

Your next move: Choose one weekly task and run it with AI twice, changing the approach the second time.

L2

Practitioner

You use AI regularly and are starting to refine how you work.

You add context, follow up to improve answers and are beginning to notice when an output needs checking.

Your next move: Name the audience and purpose in every prompt for a week, then note what changes.

L3

Integrator

AI is part of how you work.

You choose tasks deliberately, ask in structured ways and check the work that carries risk.

Your next move: Make verification a written step so it still happens when the output looks finished.

L4

Strategist

You work with AI well and help others do the same.

Your practice is strong across the four capabilities, and colleagues can learn from your approach.

Your next move: Write your approach down so a teammate can follow it without you in the room.

L5

Architect

You design how teams work with AI.

You design repeatable workflows, catch subtle failures and shape responsible norms for others.

Your next move: Keep the practice current and re-demonstrate it as tools and expectations move.

How GAFI™ measures it

How a provisional level is reached.

GAFI™ turns the framework into 30 short workplace scenarios. The assessment result is provisional and based on self-reported choices. It is not an independently verified credential.

Scenarios, not trivia

Choose what is closest to how you work today, not the answer that sounds best.

Four capability bands

The pattern across all four capabilities determines your level. No numeric score is shown.

Positional, never pass or fail

Every result names a level, a focus capability and a useful next step.

Current through re-demonstration

Take the assessment again after 8 weeks to show how your working habits have changed.

Where it fits

One assessment result inside a larger record.

A GAFI™ Assessment Result is not a standalone certificate. It can contribute to a future Digital Human Profile™ while remaining distinct from a Verified GAFI™ Credential.

Step one

GAFI™

Receive a private, provisional result describing current AI working habits.

Step two

DHP™

A future Digital Human Profile™ can bring capability evidence into one living record.

Step three

VCI™ and VEI™

Graduates see VEI™. Employers access VCI™ when verified evidence is available and shared.

Common questions

What people ask first.

Can I fail?

No. Every result is a level with a next step. L1 is a starting point, not a verdict.

Do I need to use AI at work already?

No. The scenarios use everyday work situations. Answer with what you would actually do.

Is this tied to one tool?

No. The four capabilities transfer across products and models.

What if my four capabilities differ?

That is normal. The difference helps identify your strongest habit and clearest growth opportunity.

Does finishing a resource raise my level?

Not on its own. Your level changes when you re-demonstrate the working habit.

Key terms

The words we use, in plain language.

AI fluency
The ability to work with AI well through delegation, clear description, discernment and responsible diligence.
GAFI™ Assessment Result
A dated, provisional output from the self-reported GAFI™ scenario assessment.
GAFI™ Level Badge
A shareable visual representation of a provisional assessment result, including its date and version.
Verified GAFI™ Credential
A future credential requiring defined identity and evidence verification beyond the current assessment.
Learning Library
A public catalogue of reviewed resources organised around the four GAFI™ capabilities.
Re-demonstration
Showing the capability again after time has passed so the result reflects current practice.

Get started

Find out where you sit.

Answer real workplace scenarios and receive a provisional fluency level with one practical next step.