Mandala.
An AI orchestration layer for Google Marketing. Keeping the strategic thread intact from planning to launch.
The connected marketing workspace · Mandala
The problem
Marketing teams lost strategic context as work moved between hundreds of tools, briefs, and approvals.
The solution
An AI workspace that connects planning, audience insights, and creative work and explains the reasoning behind its suggestions.
The impact
60% fasterCampaign workflows in the pilot. Satisfaction rose from 34% to 89%.
- My role
- I led experience design across the north-star vision, AI interaction model, enterprise workflows, and cross-platform specifications, aligning product, engineering, and leadership through research and working prototypes.
- The central decision
- Carry context between tools. Show the inputs and rationale behind recommendations so marketers can evaluate and explain them.
- Project scope
- Marketing workflow prototypes and pilot results, followed by a dedicated team and multi-year roadmap. Includes longer-term concepts.
The decisions behind the experience
Watch the interaction. Follow the decision.
The brief carries forward into audience research. Each step builds on the previous one, so the team can develop the work without rebuilding its context.
Distinct concepts begin with the same brief. Marketers can compare directions and refine the work while keeping the strategy intact.
The friction lived between the tools.
Google Marketing worked across hundreds of tools, dashboards, and spreadsheets. Planners were spending time reconnecting information that already existed: the brief, audience research, creative decisions, approvals, and performance data.
Improving an individual tool would leave the coordination problem intact. I mapped the ecosystem and its handoffs to understand where strategic context disappeared and where an AI layer could help carry it forward.
Understanding the existing ecosystem
01 / 02A working vision gave the organization something concrete to decide on.
I led heuristic evaluations, archetype activities, and cross-functional design jams before converging on the workspace. Sketches made it possible to compare interaction patterns quickly; prototypes let marketing leads respond to the behavior rather than the promise.
The design argument was that AI should live inside briefs, plans, and approvals. If it demanded a separate place to work, it would become another tool to manage. The prototype showed what continuity could feel like.
Exploration and alignment
01 / 07
“Your ability to transform our strategy document into a compelling story is superb. Your future-looking mocks will be used by a VP of marketing to pitch our value.”
Plan. Understand. Create. Ship. Keep the context.
Mandala connected a family of surfaces through one orchestration model. A decision made in planning could inform the brief, shape an audience, guide creative directions, and carry into the asset workflow.
The workspace adapts to the task. Audience intelligence brings signals into a usable picture, while concept generation offers distinct directions grounded in the brief. The marketer can shape the work without starting from scratch at each handoff.
Mandala in motion
01 / 05
“Why this?” made the intelligence defensible.
An early confidence score did little to help a marketer defend a recommendation. A number could not explain the choice. I changed the pattern so each proposal surfaced its inputs, a plain-language rationale, and the trade-offs it made.
Inputs link back to their sources. Rejecting an input lets the proposal recompute in place. This makes the AI something a marketer can question, steer, and bring into a real decision, not simply accept or dismiss.
Evidence at the point of decision
01 / 04
Designing for how enterprise AI fails
Numbers must come from a system of record. Brand checks happen before an output appears. AI can draft, compare, and recommend, while humans commit. Progressive responses expose the decision first, then its rationale and sources.
The shared model mattered as much as the workspace.
The broader body of work carried the same model into planning, audience building, creative tools, and measurement. Adopting one capability should make the next feel familiar: intelligence grounded in the team’s material, inside the work.
I developed Human–AI creativity principles around agency, contextual relevance, curated divergence, iterative collaboration, and explainable confidence. They gave teams a common frame for evaluating new capabilities, beyond any individual design.
Across the marketing organization
01 / 05
North-star concepts
01 / 02
Faster work. A case for continued investment.
Pilot workflows increased campaign speed by 60%, and satisfaction rose from a 34% baseline to 89%. The prototypes earned SVP support for a dedicated team and multi-year roadmap. Eight partner organizations gained a shared language for moving from isolated tools to connected work.
Results describe the campaign-workflow pilot. The wider ecosystem explorations show the longer-term product direction.
I design the rationale alongside the answer. People need to explain a recommendation to someone else; the interface should help them do that, with the inputs and trade-offs close at hand.
