Embedding intelligence in a system people already depend on
In a dense, mission-critical automation platform, interpreting failures depended on experience, not clarity. We embedded AI as an interpretation layer within existing workflows — improving decision speed and consistency without displacing operator 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
JAMS is a mature automation platform trusted to orchestrate complex, mission-critical workflows — high volumes of jobs, dependencies, logs, and configurations forming the operational backbone for organizations that rely on precision and uptime.
Over time, the platform's flexibility had created a dense interface environment. Experienced staff valued the control, but the cognitive load required to interpret logs, trace dependencies, and diagnose failures introduced friction — particularly under time pressure.
Introducing intelligence here required more than feature enhancement. It required architectural sensitivity: preserving system integrity, respecting operator workflows, and embedding assistance without disrupting established patterns of control. This was not a greenfield AI opportunity — it was the harder, more common problem of integrating intelligence into a system people already depend on.
We partnered with product and engineering leadership to embed intelligence within existing workflows without destabilizing system integrity — leading opportunity modeling and AI-assisted experience design, and establishing patterns to support future expansion.
Methodologies
- AI opportunity and risk modeling
- Workflow decomposition and operational analysis
- Information architecture restructuring
- Intent modeling and operator mental-model mapping
- Rapid concept prototyping and validation
- Cross-functional product and engineering alignment
Deliverables
- AI-assisted troubleshooting concepts embedded within existing workflows
- Workflow restructuring artifacts and system interaction maps
- Suggestion and guidance interaction frameworks
- Foundational AI integration architecture supporting future expansion
- Engineering-ready design specifications and implementation guidance
Resolution depended on experience, not clarity.
When jobs failed, staff were presented with detailed logs, system metadata, and configuration outputs. The information was technically available — but interpretation required deep system familiarity and accumulated institutional knowledge. Staff needed to answer critical operational questions quickly: What actually happened? Where did the failure originate? What is the safest corrective action?
Support teams were frequently involved, resolution time varied, and new staff faced steep cognitive barriers. This was a structural gap between data availability and decision clarity.
How do you introduce intelligence into a mission-critical system without undermining reliability, eroding trust, or displacing operator control? Replacing the interface with an automated decision-maker was not an option.
Making the data staff already had easier to interpret.
AI was treated as structured reasoning support, not an automated decision-maker. Three principles guided the strategy:
- Amplify clarity without introducing new complexity
- Strengthen established workflows, not bypass them
- Earn trust through transparency and restraint
Structural design decisions followed: introduce an interpretation layer inside the workflow where troubleshooting already happens; preserve the existing system structure; anchor every AI output to visible data sources so staff can validate it; prioritize a few high-quality signals over volume; and keep interpretation and action fully under human control.
Assistance appeared alongside the data — never in place of it.
An AI-supported troubleshooting panel was introduced within the existing job detail view. When a job failed, staff could immediately access:
- A plain-language explanation of what occurred
- Likely root causes grounded in execution context
- Structured corrective suggestions aligned with system constraints
Intelligence appeared directly alongside logs and metadata, translating technical outputs into interpretable reasoning without obscuring the underlying data. The panel distinguished clearly between explanation and recommendation. No actions were executed automatically — operators retained full visibility and control.
Rather than redefining the workflow, the solution accelerated it: failures interpreted faster, resolution paths clearer, and new staff guided without disrupting the precision experienced operators relied on.
Easier to reason about, without becoming less precise.
- Faster, more consistent interpretation of failure conditions
- Clearer resolution pathways grounded in system context
- Reduced dependency on support escalation and institutional knowledge
- Lower cognitive load for new and infrequent staff
- A repeatable pattern for embedding intelligence across other areas of the platform without architectural disruption
Because intelligence was layered into existing workflows rather than replacing them, adoption remained high among experienced operators — the hardest audience to win.
This is what responsible AI looks like inside government systems.
Public-sector platforms share this exact profile: dense, mission-critical, trusted, and dependent on institutional knowledge. California's direction to agencies is to adopt AI transparently, with privacy and security safeguards, in service of better delivery — not to hand decisions to automation.
An interpretation layer — explanations anchored to visible data, recommendations distinct from actions, humans retaining full authority — is the pattern that satisfies both the governance requirement and the staff who have to trust the system every day.
More engagements that show how we work.
Let's find the signal.
Curious where AI genuinely belongs in your systems — and where it doesn't? An AI opportunity assessment is a right-sized first step.
Start a conversation →