Embedding Enterprise AI tools for Designers and Engineers

Five AI tools for Designers and Engineers. One design philosophy: use AI to raise quality, not just cut time.

  • AI and tooling
  • Design systems
  • Enterprise delivery
Elevate AI suite logo on a black card

My role - Head of Design, Elevate Design System

As Head of Design, I created the vision and the direction for a suite of AI tools that focused on quality, not just efficiency. My leadership ensured the team had the clarity and capability needed to deliver these tools.

Collage of the Elevate AI suite brand mark, a code snippet, and the Elevate design system home screen

The impact - AI tools that deliver quality and efficiency

  • Designers

    75%reduction in time spent creating flows for existing platforms

    Output reuses Elevate Design System components rather than hallucinating them, so it is higher quality and system aligned by default.

  • Engineers

    86%reduction in component and build time

    Code is generated to enterprise standards rather than corrected downstream.

  • Enterprise

    255guidance rules embedded in the AI landscape

    Five tools available for reuse across the organisation, encoding NAB's standards into the tooling itself rather than relying on individual compliance.

  • Commercial

    Self-fundingROI metrics generated by the tools themselves

    Those metrics have been used to fund subsequent initiatives, creating a self-sustaining investment case for quality-led AI work.

5 key AI tools - driving quality for design and code

  • Lighthouse

    Design quality

    Scores a Figma file against design system standards, shows the gaps, and returns a plan with the time and cost of closing them.

    • Figma
    • Cursor

  • Scribe

    Content quality

    Checks any piece of content against NAB brand and UX writing guidelines, then aligns it. Same answer every time.

    • Copilot
    • Cursor

  • Generator

    Code quality

    Builds application code on NAB standards and Elevate components, so what ships is tested and consistent by default.

    • Cursor

  • Navigator

    Project effectiveness

    Audits the design and code in a project, plans the fixes, and puts an effort and resource figure against them.

    • Cursor

  • Cosmos

    Strategic planning

    Maps every React application across the organisation and its dependencies, with a traffic light on what needs investment.

    • Cursor
    • Web app

My approach - lead with quality, prove it, then scale

Dark-mode screenshot of a design token and component states audit panel
  1. 01

    I made the Design System AI ready.

    I worked with our team tirelessly to ensure every component, token, decision and piece of documentation was machine readable, by creating a markdown version of the entire design system.

  2. 02

    I mandated a quality first focus.

    Speed is a given for AI because it's expected. I pushed the team to lead with making any AI output better quality. That meant all output from AI needed to meet our high human standards.

  3. 03

    I started with a pilot.

    A contained project helped us quickly measure the value of the tools. We then validated them and scaled using the crawl, walk, run method.

  4. 04

    I drove adoption through storytelling.

    I screamed about the value and replayed the metrics at every opportunity I got. I used my existing relationships with senior stakeholders to quickly gained traction across more than 10 product teams, with multiple squads in each.

How I did it - with no budget

* Diagram for illustration purposes only, no real data used.

There was no budget. Using my relationships in the organisation, I ensured that the Elevate team were sought after for their skills. As such I was able to create a chargeback model for parallel work that then funded the creation of the suite of AI tools. I then used the efficiency and quality metrics to prove ROI for the tools and was able to influence senior stakeholders to unlock further investment.

What I would do differently

Enlist the service of our vendors, earlier. While we were co-designing and co-facilitating sessions with our vendor partners (e.g. Cursor, Microsoft, Amazon) I would have engaged their help much earlier to help us build out the tools.