| Director of Product Design
Turning around a stalled platform consolidation
Delivering $1M annual savings and enabled a new AI model evaluation capability.
FOCUS: Platform Consolidation | Workflow Architecture | Operating Model | Stakeholder Alignment

Context
Appen operated a large-scale AI data platform connecting enterprise customers with a global contributor workforce.
After several acquisitions, annotation workflows were fragmented across multiple systems. A consolidation program was launched to unify them into a single platform and reduce operational duplication.
Nearly two years later, the effort had stalled. Architectural decisions were unresolved, workflows conflicted, and delivery confidence was eroding.
As Director of Product Design, I led the reset of the consolidation model, reframe the workflow architecture, and restore structured execution across product, design, and engineering.
​
Stakeholder alignment artifact used to reset the consolidation from overlapping tools to a single scalable system:


Reframing the core model
1. The consolidation stalled because teams were debating features without agreement on the underlying workflow model.
Documentation existed, but it was fragmented and interpretations diverged. Each platform defined annotation jobs differently. Quality control, contributor management, and payments were embedded inconsistently. Without agreement on the core unit of work, architecture and prioritisation could not progress.
During a research interview, a pivotal comment helped reframed the problem:
"A job is a job."

2. Redefining the unit of work established a clear operational boundary across systems.
Annotation work remained contained within a job. Quality review became a separate quality job, each with its own configuration rules, contributors, and payment structure. This structural separation clarified ownership, configuration, and incentives across production and quality workflows.
With this structural model agreed, engineering, product, and design aligned on a unified information architecture and resumed execution across both frontend and backend systems:

Restoring execution discipline
Structural clarity alone was not enough. The program also required execution reset.
I introduced:
-
A structured delivery cadence across product, design, and engineering
-
Clear architectural decision boundaries
-
Weekly working sessions with engineering leadership
-
Definition of “done” aligned to the new operational model
This shifted the program from reactive iteration to governed delivery.
Organisational impact
The stalled consolidation was reset and brought into active delivery within twelve months. Three legacy annotation platforms were unified into a single enterprise system, eliminating structural duplication and generating over $1M in annual savings.
More importantly, alignment across acquired teams was restored. A shared workflow model replaced fragmented interpretations of how annotation work should operate, allowing product and engineering to prioritise against a coherent roadmap.
The consolidated architecture did more than reduce cost. It established a unified platform that enabled new AI model evaluation capabilities to be delivered at scale and accelerated the evolution of Appen’s enterprise offering.

Key takeaways
The breakthrough was not adding features, it was redefining the operational model.
When teams operate with different assumptions about how work is defined, friction is inevitable. Resetting the workflow model did more than simplify the architecture, it reduced ambiguity and rebuilt confidence across functions.
Durable momentum does not come from pushing harder on delivery. It comes from aligning structure, incentives, and execution around a shared model.

