Skip to content
TH.Tai HabershamTECHNOLOGY. BUSINESS. IMPACT.Contact
Selected work
Applied AI & knowledge engineering

Turning Watch expertise into a reusable diagnostic system.

A retrieval-augmented prototype for testing how AI interprets activation and lifecycle scenarios.

MY ROLEKnowledge-system designer · Mobile & wearables SME
CONTEXTCox Mobile / Spectrum career work
PERIOD2025–2026
PROBLEM

What needed to work.

Wearable activation spans order entry, provisioning, customer records, billing, and device lifecycle events. Scattered documentation and informal SME help make it easy to apply the wrong flow or interpret an expected delay as a failure.

MY ROLE

The ownership I brought.

  • Designed the multi-document knowledge structure for an Apple Watch activation diagnostic prototype.
  • Structured scenario-specific guidance for wearable activation and lifecycle troubleshooting.
  • Organized canonical flow definitions and signal timing so diagnosis could be anchored in the correct lifecycle context.
  • Connected the knowledge work to hands-on readiness and validation across CRM and billing systems.
THE KNOWLEDGE ARCHITECTURE

Four focused domains.
One connected diagnostic model.

The structure follows the questions people need to answer. The master RAG connects the domains while preserving scenario-specific context.

THE JUDGMENT BEHIND THE DELIVERY

Decisions that shaped the work.

01

Separate knowledge by the decision it supports.

Terminology resolves what a term means. Canonical flows establish expected behavior. Signal timing explains what can legitimately arrive later. Troubleshooting guides the next diagnostic step. A master layer keeps those domains connected.

02

Preserve scenario differences.

Customer-owned devices, retail orders, exchanges, and relinking do not automatically share the same expected sequence. The knowledge structure preserves the relevant path instead of flattening everything into a generic activation answer.

03

Challenge confident answers.

I challenged a sourced answer that appeared to mix two workflows. I also set a test-credentials-only requirement. That was a design requirement, not a verified production control.

04

Extend integration discipline to AI.

My next step for agent-enabled operations is to apply explicit permissions, evaluation, observability, and governance to workflow orchestration. Those are design priorities; this case does not claim an autonomous agent deployment or measured benchmark.

THE RESULT

What changed.

The prototype made scenario-specific troubleshooting testable and exposed where confident AI answers still needed domain review. Production rollout, adoption, and measured accuracy were not established.

WHAT IT DEMONSTRATESI apply AI to enterprise operations through domain-grounded knowledge, scenario-aware evaluation, and clear human judgment.

Source & outcome context

Supporting records show prototype design, an example of human answer review, and a test-credentials-only requirement. This case does not claim production rollout or measured adoption. The diagram is illustrative and contains no internal systems or customer data.

LET’S GET SPECIFIC

Questions this work can answer.

How are you applying AI to real business problems?
NEXT CASE / 06

Prince Tai Fitness VIP Client Tracker

THE NEXT CONVERSATION

Let’s talk about the work
you need to move forward.

Technical programs. Enterprise integration. Practical AI.