Defining and rolling out a five-level AI readiness framework for the CarTrade Tech design team, from Awareness to AI Strategist, with workshops, tool evaluations, and workflow automations that meet designers where they are.
Within months of GPT-4 going mainstream, a quiet split was opening up inside the design organisation. A small group of designers had started actively experimenting with AI in their daily work, integrating Claude into research synthesis, Cursor into design-to-code workflows, and various AI tools into the gaps in between. This group was shipping faster and producing visibly better artifacts. A much larger group watched cautiously from the sidelines, oscillating between curiosity and anxiety, and worrying privately about their relevance in a profession that suddenly felt mid-disruption.
Leadership wanted everyone to adopt AI but had no operating plan for how. The usual mechanisms were not working. Generic 'try this tool' Slack messages were producing scepticism, not adoption. Lunch-and-learns produced one week of interest and then evaporated. The few designers who were already adopting AI had figured it out themselves and were too busy applying it to teach the rest.
The underlying problem was that the team had no shared definition of what 'AI capable' meant at each stage of a designer's career, no learning path, and no way to recognise progress. Without those, individual motivation was the only force driving adoption, which is the slowest possible force at organisational scale. The brief I gave myself was to fix that vacuum: build a framework that gave every designer a clear next step, regardless of where they were starting from.
I started with a structured baseline. Every designer in the organisation went through a short self-assessment, paired with a one-on-one conversation about their current tool use, biggest blockers, and where they wanted to be in six months. The conversations were more useful than the assessment. They surfaced a pattern that the data alone would not have shown: most designers were not opposed to AI, they were opposed to feeling stupid.
The second strand of research was a competitive scan of how peer design organisations were approaching AI capability. I read public posts from design leaders at Linear, Vercel, Stripe and several Indian product companies, and where possible spoke to design leaders one on one. The consistent thread was that no one had cracked it. Most organisations were either at the 'random Slack messages' stage or at the 'mandate from above' stage, neither of which produced lasting behaviour change.
The research output was a gap map that placed every designer on a rough axis from AI-curious to AI-fluent, and identified the specific friction points blocking each one from moving up. The map made it possible to design a programme that addressed the real friction rather than a generic syllabus.
I designed a five-tier framework with explicit competencies, expected outputs, and tools mastered at each level. The progression deliberately mirrors how a designer naturally grows in any specialism.
Level 1: Awareness. The designer knows what AI tools exist, has a basic mental model of how they work, and can articulate a perspective when AI comes up in conversations. No daily tool use is expected.
Level 2: Applied User. The designer uses AI tools fluently in daily work, with established workflows for research synthesis, ideation, and design support. They can show three concrete examples of how AI has changed their workflow in the past month.
Level 3: Builder. The designer creates reusable AI workflows for the team, builds custom prompts and templates that others adopt, and contributes to the team's AI playbook. They are net producers of AI capability rather than just consumers.
Level 4: Architect. The designer designs how AI integrates into product features, not just into internal workflows. They make architectural calls about where AI belongs in the user experience and where it does not. They partner with engineering on AI-powered product surfaces.
Level 5: AI Strategist. The designer owns AI direction at the organisation level, evaluates emerging tools and platforms, and shapes how the design function evolves as the underlying technology shifts. This is a long-horizon role that I expect very few designers to occupy in any given quarter.
Framing the levels as a growth path rather than a mandate changed the conversation. Designers stopped asking 'do I have to' and started asking 'how do I get to the next level'. That shift is everything.
Tool evaluation as a habit, not a one-off. Rather than telling designers which tools to use, I set up a quarterly tool evaluation cycle. The team scores a curated shortlist of AI tools against three criteria: time saved in real workflows, output quality after review, and integration with our existing stack. The shortlist itself is deliberately small, around five tools per quarter, so designers can go deep rather than shallow. This produced two outcomes that mattered more than any individual tool decision: designers built the muscle to evaluate new tools on their own, and the team developed an internal point of view on AI in design that was not just an echo of whatever social media was loud about that week.
Workshops anchored on real work, not toy examples. Most AI training material uses generic examples that disconnect from how designers actually spend their time. I built workshops around real CarTrade design problems, with the AI tool applied to the actual artifacts a designer would produce that week. The signal that this worked was that designers started bringing their own real problems to the next workshop unprompted.
Recognition without ranking. I deliberately avoided publishing individual designer levels or creating internal leaderboards. The framework was a self-assessment tool and a manager-conversation tool, not a public hierarchy. Senior designers initially saw the framework as a threat to seniority earned the old way, and the privacy of the assessment was what made it possible for them to engage honestly with their current level rather than defending a level they were not actually at.
Self-assessment, growth conversations, quarterly reviews
Without quoting specific numbers, the most visible shift was in how designers talked about their own work. AI moved from being a defensive topic to a generative one. Designers started bringing AI experiments to design critiques unprompted, and the Builder level became aspirational, with several Applied Users self-organising into pairs to skill up together.
The second visible outcome was on hiring and role definition. Job postings now reference specific AI capability expectations using the framework's vocabulary, which has saved interviewers and candidates from the chronic miscommunication that 'AI fluent' produces in 2025. Both sides know what is being asked.
The most durable outcome is the framework itself as a reusable artifact. Other functions inside CarTrade Tech, particularly engineering, have started adapting the structure for their own AI capability mapping. A framework designed for designers has turned into a cross-functional vocabulary for talking about AI capability at the organisational level, which was not the goal but is the most useful thing the work has produced.
Designing the framework was the easy half of the project. The hard half was the human side. Several senior designers initially read the framework as an indirect criticism of how they had built their careers up to that point, and I spent more time on one-to-one conversations than I had planned, listening to those concerns and adjusting how I positioned the framework in team forums.
The lesson, and one I will carry into every team transformation I run from here, is that any change at organisational scale is one part strategy and three parts care. The framework worked because people felt seen at whatever level they started from, not because the levels themselves were clever. I would scope the next iteration with explicit one-on-one time built into the rollout calendar rather than treating it as informal overhead.