Is your agency ready for AI? Assess four things: context, people, process and platform. Then ignore your best score. According to one Singapore agency founder who spent 18 months trying to go AI-native, your weakest dimension sets your readiness level. That is why sending everyone on an AI course, the default answer to "how do we upskill?", so often changes very little.
The people question is getting sharper. Marketing Dive reported Gartner research last month showing that, of 1,303 senior marketing leaders, 18% have already eliminated some positions because of AI and 16% have created new roles. Gartner predicts most high-performing marketing teams will eliminate traditional entry-level jobs by 2030, and urges CMOs to give juniors higher-value work earlier rather than cut the talent pipeline. That is hard to do if the rest of the agency is not ready to support them.
What does an AI readiness assessment actually measure?
The founder, speaking at an AI adoption session in Singapore this week, boiled readiness down to four dimensions. As the diagram shows:
- Context: what is written down that AI can draw on. Briefs, brand guidelines, past work, client history, the way you do things.
- People: who gets real work out of AI, and who is falling behind.
- Process: how work actually runs, beyond the SOP.
- Platform: what you pay for, and how deeply it is used.
Most agencies only assess the last one, and badly. They count licences. A licence nobody uses properly is a cost, not a capability.
Why does your weakest dimension decide your readiness?
Because AI work flows through all four and stops at the narrowest point. As his slide put it, your weakest dimension sets your readiness level; your strongest one does not show readiness at all.
Picture three agencies. A creative shop with brilliant, curious people but almost nothing written down: every AI output sounds generic because the model has never seen the brand book or the last three years of work. A media agency with excellent platforms and tidy dashboards, where the real process lives in the heads of two senior planners and a chat group, so automation keeps breaking on undocumented exceptions. A CRM agency with immaculate processes and a team that has quietly decided AI is not for them.
Each one has a strength to point to. None of them is ready. And the fix is different in each case, which is why a generic AI programme rarely works.
Why aren't AI courses enough to make a team ready?
Courses address one dimension, people, and only part of it. The founder raised a harder question: if you succeed in turning staff into hybrid experts, part strategist, part engineer, part client lead, what happens when they leave? And when you win business and need to hire, where do you find more of them?
Agencies need to be able to deploy new hires quickly. A readiness strategy that depends on a few rare hybrids is fragile. One that puts context into shared systems, documents how the work really runs and uses tools a new starter can pick up in days scales with the agency.
This matters doubly for juniors. If Gartner is right that traditional entry-level roles are disappearing, the juniors you do hire need to reach higher-value work fast. That only happens when context, process and platform do some of the teaching.
His conclusion was blunt: adoption likely requires restructuring the team and reorganising how work is done. Training is part of that, not a substitute for it.
Why do some teams pull ahead while others stall?
Digital Divide Theory was developed to explain why access to technology does not guarantee equal benefit. The gap opens in layers: first who has access, then who has the skills, then who turns it into real outcomes. Every agency has a version of this inside it. Everyone has the same licence; a handful of people get real work out of it; the rest fall further behind each month.
The same divide runs between holding networks and independents. The networks have spent heavily on platforms. Independents cannot match that spend, but they can close the outcome gap through connection: shared context, people learning from each other on live work, peers swapping what actually works. That is what anyone typing "how can I learn from others in my field?" is really after. Innovation thrives in connection, not competition.
In practice, the fastest way to close the internal divide is to pair the people getting value from AI with those who are not, on real client work rather than in a classroom.
How do you run a quick AI readiness assessment?
Score each dimension from 1 to 5 using the questions below, then put your effort into the lowest.
| Dimension | Question to ask | Signs you are weak | First fix |
|---|---|---|---|
| Context | What is written down that AI can draw on? | Briefs buried in inboxes, brand rules in one person's head, past work unsearchable | Build a shared library of briefs, guidelines and best work for each client |
| People | Who gets real work out of AI, and who is falling behind? | A few power users, many sceptics, no shared standard for "good" | Pair power users with sceptics on live work; name approvers |
| Process | How does work actually run, beyond the SOP? | Exceptions handled in chat threads; nobody can map a job end to end | Shadow one workflow and document what really happens |
| Platform | What do we pay for, and how deeply is it used? | Overlapping subscriptions, low usage, tools nobody owns | Audit licences; cut, consolidate and give every tool an owner |
- Pick one team or one client. Assessing the whole agency at once produces a vague average.
- Interview, don't survey. Ask five people to walk you through a recent job. An hour of that teaches you more about process and context than any questionnaire.
- Score each dimension honestly. Use the questions above. Disagreement between scorers is useful data, not noise.
- Circle the lowest score. That is your readiness level, whatever the other three say.
- Spend the next quarter on that one dimension. Resist buying another tool if platform is not the problem.
- Re-score and repeat. As one dimension rises, another becomes the constraint.
What changes once you know your weakest link?
You stop buying solutions to problems you do not have. If context is the constraint, the work is unglamorous: write things down and put them where people and AI can find them. If process is the constraint, map it before you automate it. Restructuring may follow, and according to the Singapore founder that is usually the point, not a failure.
If people is the constraint, the answer is often a tool that asks less of them. The Defy platform includes Defy MCP, which lets your team interrogate client and category data inside Claude and prompt it like a co-worker, with no dashboard to learn. That lowers the skills bar, so a new hire is useful sooner and the gap between power users and everyone else narrows. Which, in the end, is what readiness means.
