The 4Ds: the core skills for working with AI
Summary: The skills that separate teams doing remarkable work with AI from teams typing questions into a chatbot aren't technical – and there are four of them. Or there were, until Luke Taylor of United Us added a fifth on stage with us at Brandland this month. Here's the full framework, with a question to test your team against each one.
Author: Emma Wharton
Read length: 3 minutes
Date: 10th July 2026
Spark AI is a strategy-led consultancy helping agencies and creative teams build organisation-wide AI capability. Together, we'll build the engine that changes not just how your team works, but what your organisation is capable of. Our blog provides the techniques, insights and industry discussions needed to navigate AI with confidence.
It’s easy to assume the AI skills gap is a technical one – that somewhere out there is a prompt engineer or an automation specialist who holds the answer.
Ten years ago, that might have been true. Specialist skills and capital were the price of entry for new technology, which is why the big networks always got there first.
AI has changed the terms. The tools cost less than your project management software, and the skills that separate teams doing remarkable work from teams typing questions into a chatbot are not technical at all. After working with 80+ agencies, we can tell you there are four of them.
1. Description
Brief AI the way you'd brief a person – with full context.
AI produces generic work for the same reason a freelancer would: it wasn't told enough. The teams getting the best output treat every prompt like a creative brief – background, audience, tone, constraints, examples of good. If your agency is good at writing briefs, you're closer than you think.
Would a new starter succeed with the brief you just gave the AI?
2. Discernment
Know what good looks like – and what to reject.
AI will hand you something plausible every single time. Plausible isn't the same as good. Discernment is the editor's skill: spotting the generic phrase, the wrong strategic take, the almost-right design. It's why experienced creatives often get more out of AI than digital natives – they have deeper references to judge against.
Could everyone explain why an AI output isn't good enough?
3. Delegation
Know what to hand over – and what to keep.
AI is good at tasks. It is not good at jobs. Delegation is knowing the difference: hand over the synthesis, the research, the first drafts and the formatting; keep the relationships, the taste and the judgement. Teams that get this wrong either automate what they shouldn't or drown in work AI should be doing.
What did you do this week that AI should have done instead?
4. Dialogue
Direct the work through conversation, not one-shot prompts.
The first output is the start of the conversation, not the end. Push back, ask for alternatives, make the AI argue against its own recommendation. The best AI users work the way good creative directors do – through rounds of direction, not a single instruction.
Do you refine AI output through conversation and actively direct and challenge it, or accept the first answer?
Why this is important
With everyone holding roughly the same tools, technology access is no longer an advantage. The gap between you and your competitor now comes down to capability – and capability is built, not bought.
That's what the 4Ds give you: a shared language for working with AI across your whole team. A way to brief it, judge it, direct it, and know when not to use it at all. No specialist hires required.
Frequently asked questions:
Want to build the 4Ds across your team?
Our AI Foundations workshop takes your whole team through the framework using your real client work – everyone on the same page in a day.
Where does the 4Ds come from?
The 4Ds come from Shift - AI for Agencies, the #1 bestselling book by our co-founder Jules Love. Get the book.
Where can I find more AI insights like this?
We publish a fortnightly newsletter called Spark Intelligence, where we share the latest AI insights and updates for agency and brand teams.
