Find the useful problem.
Start with an existing pain point instead of adding AI because it is available.
Practical AI · implementation · enablement
I turn AI into workflows people can actually use.
I build practical AI tools, test where they help and where they fail, and help people use them with confidence.
My work sits between technology and the people using it: finding useful problems, designing the workflow, testing the output, keeping human review in the loop, and teaching people what to trust and what to check. I come from enterprise IT support and workflow systems, so I’m interested in AI that fits into real work — not demos that only look good on launch day.
At a glance
- Tier 3 End User Support Specialist — ServiceNow, daily since 2020
- 30–50 cases a day across incidents, problems, service requests, and tasks
- B.S., Information Technology (AI concentration) — Strayer University, 2025, summa cum laude
- ServiceNow Certified System Administrator (CSA) — in progress, exam November 2026
- Configured practical Copilot agents and coached coworkers on using them, with human review in the loop
- Based in the greater Nashville area — remote preferred
- Resume available on request
AI in real workflows
Build the tool. Test it. Help people use it.
I’ve used AI inside support workflows, not just in personal experiments. I configured Copilot agents that helped coworkers choose response templates and ServiceNow categories from existing SharePoint knowledge sources.
Employees stayed in control: they reviewed the result, chose what to use, and checked procedures when something was unclear.
I tested difficult examples, revised the agents’ instructions when they missed the point or included unsupported details, and ran the same cases again to see whether the changes helped.
I also walked coworkers through the tools, answered questions, created instructional material, and kept the agents’ source material current as templates changed.
The goal wasn’t to automate people out of the workflow. It was to make a familiar task easier while keeping judgment where it belonged.
How I approach implementation
Start with the work, not the model.
A useful AI workflow usually starts with a boring question: what is someone already trying to accomplish?
Give the system the right context.
Use the knowledge, procedures, or structured information the task already depends on.
Test the ugly cases.
Easy examples rarely tell you much. I look for ambiguity, missing information, conflicting instructions, stale memory, and unsupported answers.
Keep people in control.
Suggestions should be inspectable, editable, and easy to reject before they become part of the work.
Teach the check.
Adoption is more than showing someone where the button is. People need to understand what the tool does well, where it can fail, and when to verify its answer.
Measure what changed.
Where possible, compare quality, usability, latency, cost, or the amount of correction required instead of relying on whether the demo felt impressive.
Experiments that became evidence
Selected builds
Three projects that pushed on workflow design, reliability, and repeatable AI behavior. See all builds →
Applied AI · Workflow
TinyDesk
A fictional help desk where AI summarizes ticket history, suggests categories, identifies missing information, and drafts editable replies. A twelve-ticket evaluation compares usefulness, errors, latency, and cost while keeping the technician in control.
Read the case study →
AI · Memory reliabilityTinyTalk
A local assistant that separates current facts, historical facts, recent conversation, and retrieved memories so stale information does not quietly become truth.
Read the case study →
Agent systemsCricket Mode
A portable behavior layer for AI coding agents built around scope, challenge, evidence, and cleanup across Codex, Cursor, and Antigravity.
Read the case study →
Writing
On the shelf
The tools I'm using, the things I'm building, and questions that come to me as I learn and experiment.
Field notes
The AI meter you can't fully see
What a token actually is, why AI bills by it, and the playbook I use to spend fewer…
Field notes“Could AI Do This?” Is the Wrong First Question
Before building another agent, workflow, or Copilot integration, there's a more important question to answer: is this actually a good use of AI? Sometimes the answer is “no AI needed.”
Field notesBeyond Coding: How People Use Cursor for Research, Notes, and Writing
Notes that talk to each other, transcripts into memos, and reviewing an AI edit like a code diff: how people use Cursor for work that isn't code.
Creative technology
The weird stuff stays on the workbench.
Not every build needs to look like enterprise software. Commit City, Mutation Microscope, and the rest of the lab are where I explore interfaces, visualization, local AI, and ideas that are useful precisely because they started a little strange.
Open to practical AI work
I'm interested in AI Implementation, Enablement, Adoption, and related roles where the work is less about showing off a model and more about finding useful workflows, testing them carefully, helping people use them, and improving what happens after the first demo.
My background is enterprise IT support and workflow systems, including ServiceNow. That experience shapes how I think about users, procedures, troubleshooting, knowledge, and human review.
ServiceNow track
Tier 3 support work since 2020, 30–50 cases a day, CSA in progress.
Reach me at cyb3rcricket@proton.me or message @Cyb3rCricket on X.
Resume available on request.
I can start within two weeks' notice.