Structure before scale: rebuilding a workflow so AI could help
Leadership wanted AI to improve efficiency — but the core workflow, which took up to twelve screens to complete, lacked the structural clarity intelligence requires. We rebuilt the workflow first, creating the foundation that let AI be introduced without adding complexity or compromising control.
This was a commercial engagement delivered by the team behind This Waay. It is presented as relevant experience for public-systems work — not as a government contract reference.
Context
Following consolidation, Portico operated across three overlapping student platforms. Core workflows were duplicated, interaction models diverged, and terminology varied between products. One of the most central experiences — logging a skill — required up to twelve screens to complete, with repeated data entry, inconsistent field structures, and unclear progression.
Leadership was interested in introducing AI to improve efficiency. But the workflow itself lacked the structural consistency intelligent assistance requires: fragmentation and ambiguity created more friction than AI could realistically solve.
Before intelligence could scale, the workflow needed to be rebuilt. This engagement is a companion to our broader work unifying the student journey.
We rebuilt and unified the skill-logging workflow across platforms, establishing the structural clarity required for responsible, scalable AI assistance.
Methodologies
- Workflow decomposition and friction mapping
- Cross-platform information architecture alignment
- Heuristic and standards evaluation
- Data model and attribute normalization
- Iterative prototyping and validation
- Cross-functional stakeholder alignment
Deliverables
- Unified skill-logging workflow framework
- Normalized data structure and required-field model
- AI voice-capture and structured-suggestion concept
- Review-and-approval interaction pattern
- Engineering-ready design specifications
The workflow wasn't ready for AI. It lacked the structure to support it.
The existing workflow required students to move through up to twelve discrete steps with limited guidance, re-enter information already provided, navigate non-sequential screens introduced through system consolidation, and make decisions without sufficient context or confirmation.
As platforms merged, interaction models diverged and terminology became inconsistent. Students relied on workarounds and institutional knowledge to complete tasks accurately.
Introducing AI into this environment risked amplifying confusion rather than reducing it. The workflow lacked the structure that intelligence requires.
Before introducing AI, we rebuilt the workflow into a coherent, canonical system.
Because this workflow would serve as the backbone for AI-assisted acceleration, it required structural clarity, consistent data modeling, and a reliable manual path people could trust. The rebuild focused on:
- Unifying required attributes across products
- Normalizing terminology and field structures
- Sequencing interactions into a coherent, standards-aligned flow
- Eliminating redundant steps and duplicated effort
The objective was a single canonical workflow that could support both manual entry and AI-assisted acceleration — with the same validation for both.
A structural cleanup that became the backbone of the AI validation sequence.
The redesigned experience became a guided, sequential workflow aligned with how students think about the task. Redundant steps were removed, information was consolidated, and progress indicators clarified where students were and what came next. Structural decisions made the system one that AI could operate within:
- A canonical workflow across platforms — one consistent structure, eliminating fragmentation
- A normalized data model — manual and AI-generated inputs map to the same structured system
- A shared path for manual and AI input — voice capture and manual entry operate within the same validation framework
- AI as an accelerator inside the workflow — speeding input without bypassing structure
- Explicit user validation before submission — every AI-generated input is reviewable, editable, and confirmed by the person responsible
A scalable, governable foundation for intelligence.
- Faster task completion with improved accuracy
- Reduced reliance on institutional knowledge and training
- Increased confidence in completing complex workflows
- A validated backbone supporting both manual and AI-assisted paths
- A foundation for future AI capability that is governable by design
Because the workflow itself served as the validation layer for AI suggestions, later expansion of intelligence was more reliable, more governable, and lower risk.
AI doesn't fix fragmented systems. It exposes them.
California has directed agencies to adopt AI transparently, with safeguards, and in service of better delivery for residents. The uncomfortable truth underneath that directive: when workflows are fragmented across systems, teams, or data models, AI amplifies the inconsistency rather than resolving it.
Structure before scale is what responsible adoption looks like in practice — rebuild the workflow, normalize the data, keep a trustworthy manual path, and require human confirmation. Agencies that sequence the work this way get AI that is governable by design.
More engagements that show how we work.
Let's find the signal.
Wondering whether your workflows are ready for AI? An honest readiness assessment is a right-sized first step.
Start a conversation →