Advanced practical AI · CX Operations · Human-in-the-loop

Janackeh
Blackwell.

Senior Customer Support Specialist at Zoro. Most support and CX candidates use AI casually. I get curious before I solve — what’s it for, what size, how do you use it — then design, verify, and ship human-in-the-loop workflows on company-approved tools, and run my own AI agents as personal projects. Hiring teams get discovery depth and AI fluency most peers don’t bring.

Explore my AI toolkit Connect on LinkedIn Download resume
Practical AI workflows · Investigation · Prompt systems Los Angeles · Remote only

The principle

AI in support should do the repetitive work so humans stay the warm, accountable element in the relationship.

Anyone can paste a ticket into a chatbot. The scarce skill is designing the prompt, the verification step, and the handoff so the human stays accountable for what goes into the CRM and out to the customer.

Most customers don’t call with a part number. They call with what it’s for, how they use it, or the item they bought years ago. I start with discovery questions before I offer a fix — those answers are what get us to the right item, not the first guess. I’m at my best when the case is hard, and my goal is the work that prevents repeats: customer experience operations.

That through-line runs through my whole career. I started in 911 dispatch, triaging emergencies under pressure, then carried that composure into insurance claims and mortgage processing — investigating files, enforcing standards, and coordinating across underwriters, closers, realtors, and vendors. I proved I could run operations when I lifted property occupancy from 78% to 96%, then retooled for tech: a software engineering fellowship, an open-source fellowship at Sauce Labs, and hands-on AI workflow design for support.

At Zoro, working complex cases has clarified what I want next: the systems side of customer experience, where practical AI — on company-approved tools only — and investigation habits prevent repeat issues instead of only resolving them one at a time. Personal AI agent projects stay outside of work.

How I use AI

Tools I run.
Checks I insist on.

Day-to-day AI practice — what I actually use, how I verify outputs, and demos you can try yourself.

Daily stack

Gemini · ChatGPT · Five9 Agent Assist → NetSuite

At work I use only company-approved tools. Gemini (including custom Gems) for deeper product and part research and mispick verification concepts; Five9 Agent Assist summaries shaped into concise NetSuite notes (Subject + short Call Summary). I review and edit every draft before it goes into the case — nothing writes to NetSuite automatically, and customer data stays inside approved systems. ChatGPT supports prompt libraries and research in contexts that don’t put customer data at risk.

Proof: call-note demo · case study · interactive mispick workflow

Workflow design

Prompt libraries & verification habits

Reusable prompts for ownership-driven customer emails and structured documentation. I emphasize checking sources, verifying item numbers, and clearly labeling what cannot be confirmed before anything reaches a customer or the CRM.

Human-reviewed by design

Method

Search Gap — from vague request to the right item

When the customer’s language doesn’t match the catalog, I start with their exact words, use company-approved Gemini to clarify search terms, match on function, size, and use (not a guessed SKU name), find a suitable alternative when there’s no exact match, and verify item number, specs, and fit before anything reaches the customer.

Human-reviewed method · company-approved tools only

Personal projects

AI agents outside of work

Outside work, I run my own AI agents for research, drafting, and career projects — with my approval required before anything goes out. At work, I stick to company-approved tools like Gemini, and customer data never leaves them.

Personal practice · human approval gates · separate from work systems

Portfolio prototype

Signal — AI Operations Copilot

A working prototype for ApartmentIQ’s AI Operations Lead role: triage-and-drafting for tickets and call notes, with human review, confidence scoring, and an evaluation dashboard. Same HITL pattern as my support workflows, applied to a new domain.

Try the demo · GitHub

Portfolio build

Resolve

A working end to end demo I built outside of work. It reads a messy complaint, proposes 2 or 3 resolution options, and runs a PayPal sandbox refund only after a person approves.

View on Devpost · GitHub

Literacy

MCP & multi-agent awareness

Actively learning the agent ecosystem — Model Context Protocol, multi-agent patterns, and how local tools connect — so I can partner with eng and AI teams with shared vocabulary and clear handoffs.

Community sessions · ongoing practice

How I keep it safe: At work, every AI output is human-reviewed and runs only on company-approved tools — customer data never leaves them. Personal AI agents stay outside of work.

Selected work

Workflows you
can try.

Interactive demos and playbooks — human-in-the-loop tools for search, summarization, and verification so people stay focused on judgment and complex resolution.

Resolve graphic: a vendor invoice email next to a $237.50 PayPal payment confirmation
Portfolio build

Resolve

A working end to end demo I built outside of work for the PayPal AI Hackathon. It reads a messy customer complaint, pulls out the key facts, and proposes 2 or 3 resolution options (full refund, partial, store credit, or no refund). A short approval packet comes next. A PayPal sandbox refund runs only after a person approves, and a fail-closed gate re-checks the packet right before any PayPal call. Extraction and proposal logic is mocked for reliable demos, behind an LLM-ready interface.

Next.jsTypeScriptPayPal sandboxHuman-in-the-loop
View on Devpost GitHub
Human-in-the-loop AI · Featured

AI-Assisted Call-Note Drafting

A Gemini Gem that turns AI call-transcript summaries into clean, consistently formatted CRM notes — a Subject line plus a 1–3 sentence Call Summary. The rep reviews, edits, and posts — human judgment stays in the loop for consistency and accountability.

GeminiFive9 Agent AssistNetSuite
Try the demo → Read the case study →
Investigation · Human-in-the-loop AI

Mispick Verification & Part Search

An evidence-first investigation workflow that turns “you shipped the wrong item” into a root cause — true mispick, packaging change, vendor substitution, customer error, listing error, or variation confusion. Compares what was ordered against what was received, spec by spec, and classifies the case. Started as a Gemini Gem for verified part search; grew into a full mispick playbook.

Gemini GemOrder managementRoot Cause Analysis
Try the demo → Read the case study → GitHub →
Prompt design

Customer Email Prompt Library

A prompt library that drafts brief, empathetic, ownership-driven customer emails — aligned to core CX values across delayed shipments, marketplace confusion, RMAs, refunds, and upset customers. Always human-reviewed before send.

ChatGPTPrompt Design
Portfolio prototype · ApartmentIQ

Signal — AI Operations Copilot

A portfolio prototype designed for ApartmentIQ’s AI Operations Lead application: triage-and-drafting for support tickets, CS call notes, sales summaries, and data-quality complaints, with human review, confidence scoring, and an evaluation dashboard.

LLM PromptingHuman-in-the-Loop QAEvaluation Dashboards
Try the demo → GitHub →
Quality & CSAT

QA Call-Language & CSAT Playbook

Reusable call language for greetings, holds, ownership statements, transfers, and closings to improve consistency — plus tested closing language aimed at clearer survey participation asks.

QA StandardizationCSAT
Concept · Future state

AI Call-Coaching Dashboard

A proposed concept: an AI call-analysis surface that turns call transcripts into coaching signals — trends, recurring friction, and improvement opportunities for the rep and the team.

ConceptCoaching
Concept · Future state

“Apple-style” Agent Workspace

A proposed concept: redesigning the support agent workspace around clarity, focus, and intelligent triage — fewer tools, better signal, calmer work.

ConceptWorkspace Design

For hiring teams

How I can help
your team.

Concrete value — what I bring that a typical support hire usually doesn’t.

  1. Curiosity-first discovery

    Ask what’s it for, what size, and how they use it before offering a fix — so product search starts from the customer’s real need, not the first catalog guess.

  2. Stand up practical AI without losing control

    Prompt patterns and human review gates for notes, research, and customer email on company-approved tools — so AI drafts never silently become the record of truth.

  3. Turn “wrong item” chaos into root-cause clarity

    An evidence-first mispick playbook with six classifications, spec comparison, and customer-ready outputs — so replacements and refunds follow the real cause, not the first assumption.

  4. Turn one fix into a reusable playbook

    Call language, note templates, investigation workflows, and peer training on product search — so a solution that worked once can help the next ten cases.

  5. Right item first time, fewer repeats

    When discovery and verification land the correct product (or a real alternative), customers don’t need to call back about the same gap — better for them and for retention-minded teams.

  6. Investigate and own hard escalations

    Daily NetSuite order investigation, multi-party delivery coordination, and complex cases spanning business customers and consumers.

  7. Speak builder and ops

    JavaScript, React, Python, and Git from web-development training and a Sauce Labs open-source fellowship — a technical foundation that helps me partner with eng and AI teams as an ops collaborator.

The journey

From triage under
pressure to systems
thinking.

Investigation habits first — then tech fluency and practical AI for support operations.

  1. Mar 2026 — Present

    Senior Customer Support Specialist · Zoro

    Senior-level support in a blended phone-and-email queue, resolving complex cases end-to-end while developing practical AI-assisted workflows and standardized processes for my own support work — with an emphasis on consistency, accuracy, and human review. Daily across company-approved tools: NetSuite, Five9, Five9 Agent Assist, and Gemini. Customer data stays inside those systems.

  2. Nov 2025 — Mar 2026

    Escalation Specialist · Zoro

    Stretch specialization handling escalated, high-impact customer situations and sensitive issues requiring higher-level problem-solving, ownership, and an investigative approach to case resolution.

  3. Aug 2023 — Oct 2025

    Customer Support Representative · Zoro

    Frontline support on a high-volume queue across order questions, returns, shipping, and product issues — building the credibility that earned rapid early promotions.

  4. 2022 — 2023

    Open Source Fellow · Sauce Labs

    Selected fellowship building open-source tooling around Selenium, mentored across content, software QA, and development. Stack: Vue.js, Express, Node.js, MongoDB, Strapi.

  5. 2020 — 2022

    Senior Mortgage Loan Processor · Keller Mortgage

    Processed purchase and refinance loans, coordinating across loan officers, underwriters, closers, title, insurance, and appraisal providers — driving loans to on-time closing.

  6. 2014 — 2016

    Property Manager · Greenleaf Management

    Managed two communities (174 and 66 units) including staff, budgets, vendors, and turns. Increased occupancy from 78% → 96% at one community and 80% → 99% at another.

  7. Earlier

    911 Dispatch · Insurance Claims · Mortgage Files · Operations lineage

    Started in 911 emergency dispatch — triaging emergencies under pressure — then carried that composure into insurance claims and mortgage file review. A genuine lineage for investigations, RCA, and Trust & Safety work.

Throughout: Founder, Black Tech Denver (Sep 2022 — Present) — building community for underrepresented people in tech through career and social networking.

Roadmap

What I'm learning.

Credentials that back CX/Ops credibility and deepen AI-ops fluency — reinforcing practice I already use every day.

In progress

ChatGPT Prompt Engineering for Developers

DeepLearning.AI · Andrew Ng & Isa Fulford · free · ~1h40m

Reinforces the prompt-design thinking behind my call-note workflow — summarizing, transforming, and iterating on LLM output with a human review step.

Next

ITIL 4 Foundation

PeopleCert · exam in ~6 weeks

The baseline service & support-operations credential — the keyword recruiters filter CX/Ops roles on. Prepping via LinkedIn Learning + Udemy Business.

Exploring

Lean Six Sigma Green Belt

Accredited body · IASSC / ASQ

Process-improvement and root-cause analysis credibility for CX/Ops and RCA work.

Optional

KCS v9 · AWS AI Practitioner · Azure AI Fundamentals

Knowledge management & broader AI-ops

Credentials to pursue depending on which roles I target next — knowledge management, or deeper AI-ops fluency.

Capabilities

What I bring to a CX/Ops team.

Support & CX Operations

  • Customer Experience Operations
  • Complex Case Resolution
  • Root Cause Analysis
  • QA Standardization & Scripting
  • CSAT Optimization
  • De-escalation
  • Customer Empathy
  • Order Management & Logistics
  • SLA Ownership

AI & Tooling

  • Curiosity-first discovery
  • Product search / Search Gap
  • Playbook design
  • Human-in-the-Loop AI for Support
  • Prompt Design & Iteration
  • Gemini Gems / Custom Assistants
  • ChatGPT
  • Five9 Agent Assist
  • HITL QA & Source Checking
  • Personal AI Agents (outside work)
  • MCP Literacy
  • Evaluation-minded Prototypes
  • NetSuite
  • Five9

Systems & Methods

  • Process Improvement
  • Continuous Improvement
  • Problem Management
  • Postmortems
  • Knowledge Sharing
  • Data-Driven Decision Making
  • Dashboards & Reporting

Leadership & Community

  • Community Building & Leadership
  • Mentorship
  • Peer Training
  • Cross-Functional Collaboration
  • Stakeholder Management
  • JavaScript / React / Python / Git

Let's connect

Seeking remote Customer Operations, Customer Success, Product/Technical Support, and AI-enabled support roles at technology and AI companies.

Los Angeles · Remote only. Recruiters, hiring managers, and collaborators welcome.

Email janackeh@gmail.com LinkedIn /in/janackeh GitHub github.com/janackeh