AI-Powered Creative Transformation

Creative demand was outpacing headcount, and AI adoption was happening ad hoc.

Citi's in-house creative organization was under growing pressure to produce more — more journeys, more markets, more variations — without a proportional increase in headcount. At the same time, generative AI tools were already showing up in individual workflows across the 50+ person creative org, informally and without any shared standards.

In a regulated bank, that's not just a productivity story, it's a risk story: inconsistent outputs, no audit trail, and no clear line between experimentation and anything that could touch a customer, legal, or brand review.

A founder-led task force, purpose-built tools, and governance from day one.

I founded Citi's internal AI Task Force and have led it since, personally training the 50+ person creative organization on responsible generative-AI usage, hosting recurring lunch-and-learns, and building decks that brief leadership on use cases and measurable impact. Alongside that, my team and I design, build, and run a growing set of custom AI tools aimed at specific, recurring bottlenecks in the creative workflow.

What follows is a small sample of those tools — a few can't be shown here because the work is IP and proprietary to Citi, but these give a clear sense of the range: from creative ideation, to compliance review, to the automation layer running quietly underneath both.

• Copy Lab — a headline-generation tool tuned to each Citi brand's voice (Wealth, Citigold, Citigold Private Client). Creatives enter a brief, audience, and campaign goal, plus any legal/compliance notes, and get back several structurally distinct headline directions — contrast pairs, tension pairings, audience-led, brief-driven — so the team starts ideation from six strong options instead of a blank page.
• ProseCheck — an AI compliance and brand-voice reviewer that scores marketing copy against grammar, brand voice, legal/compliance, and UX criteria before it ever reaches a human legal reviewer, flagging specific issues (weak or hedging language, missing required disclosures, passive voice) with a clear pass/fail signal and a director's-note style explanation of what to fix and why.
• An internal automation workspace that runs scheduled AI jobs the team relies on weekly — automatic detection of banner-spec changes across channels, a recurring competitor-marketing sweep, and scheduled share-outs of what's new in tools like Claude and Claude Code — plus a shared library of reusable prompt presets (creative brief generation, copy-variation automation, hallucination-checking) so good prompting isn't something each person has to reinvent.

Beyond these tools, my team and I also work daily in Adobe's AI suite — Firefly and GenStudio — along with Microsoft Copilot and other licensed tools, and we code our own purpose-built utilities using LLMs including Gemini, Claude, and Grok. My design lead, Sandon Spalding, is a key partner in a lot of this work and is exceptionally strong with AI — his own tools aren't pictured here until he's ready to share them, but the broader push to drive AI fluency, experimentation, and responsible adoption runs across my team and the entire 50+ person agency, not just through me.

A validated path to scale creative output without scaling risk.

The task force and its tools gave Citi's creative organization a shared, sanctioned way to use AI — turning a scattered set of individual habits into an auditable, governed capability that legal and risk could sign off on, and that leadership now treats as a model for the rest of the business.