WorkflowÂ
Omnicom
Client
Omnicom
Role
Product DesignerÂ
Duration
In Progress
Challenge
Creative and marketing production teams at Kinesso relied on manual, multi-tool workflows to move a campaign from brief to approved asset — with no self-serve way to configure AI agents, orchestrate multi-step production tasks, or keep human approval in the loop for brand and legal compliance. The goal was to design a system that let non-technical production users build, test, and run their own AI-powered workflows end-to-end, without engineering support, while preserving quality control at every critical step.
Approach
Started with ideation and system mapping — sketched out how users, agents, tools, and workflows (ACTIFlow) would interact, using existing node-based orchestration tools like n8n as a reference for a more visual, non-technical builder.
Defined two core user journeys:
-
A workflow-centered path (Production Console), where users build and run custom agents directly.
-
A campaign-centered path (Research Console → Production Console handoff), where research and creative ideation flow into a structured production brief before reaching execution.
-
Validated logic before visuals — worked through low-fidelity flows and decision points first, to make sure the orchestration logic (approvals, handoffs, agent triggers) made sense before investing in UI.
Designed for real-world edge cases early — mapped out states like paused executions, disconnected nodes, and error handling, not just the happy path, since production reliability was a key concern from the start.
Moved into high-fidelity UI once the flow logic was validated, focusing on the Workflow Builder canvas, node configuration panels, and the workflow library.
Solution
I designed a full Workflow Builder experience: a drag-and-drop canvas with modular nodes (Agents, Tools, Logic, Human-in-Loop, Output), each with its own configuration panel (prompts, model selection, retries, error handling, output variables).
