AI Workflow Development Case Study

AI Employment Navigator

A working concept and web prototype I created to explore how AI could support job discovery, application preparation and AI literacy while keeping human judgment, privacy and program integrity at the center.

Current status

Status: working concept and prototype. A participant pilot and staff training are proposed next steps.

Working prototype

AJC Navigator — workflow preview

ajcnavigator.netlify.app
Start Here
Job Watch
Resume
Interview
AI Literacy
AI Employment Navigator

What do you need to accomplish?

A structured starting point for job search, application preparation and responsible AI use.

Build candidate profileskills, salary, location, interests and exclusions
Run Job Watchsearch broadly, screen carefully, verify before action
Tailor applicationalign truthful experience to employer needs
Prepare interviewresearch, STAR examples and likely questions
Human review remains part of the workflow before important decisions.
Interface representation of the live concept. The Launch Project button opens the working Netlify site.

The problem I set out to solve

Job search is fragmented across platforms and can produce a large amount of low-value noise. I wanted to explore whether a structured AI-assisted workflow could better reflect an individual job seeker's real qualifications, salary needs, geography, interests and exclusions — while still requiring human review.

Design principle

Search broadly. Filter intelligently. Keep humans in control.

How I built it

I used generative AI extensively as the primary development tool. I defined what the system needed to accomplish, developed the workflows and requirements, reviewed and corrected outputs, rejected approaches that did not work, and repeatedly refined the prototype and supporting materials.

What I created and directed

  • Original concept, project vision and service architecture.
  • Candidate-profile intake, privacy choices and minimum-data considerations.
  • Job Watch screening rules, qualification logic and realistic-match methodology.
  • Application and interview-support workflow.
  • Participant-facing workflow design and working web prototypes.
  • Netlify deployment and repeated site revisions.
  • Responsible-AI safeguards, human-review rules and truthfulness controls.
  • Executive briefs, implementation guidance, career-planner and client materials, FAQs and stakeholder resources.
  • A proposed 30-day pilot framework, decision gates and measures for future evaluation.

System architecture

7-layer operating model
L1
Client profileskills + goals + constraints
L2
Job sourcesboards + employers + public sources
L3
AI screeningrequirements + exclusions + fit
L4
Human reviewverification + judgment
L5
Application supportresume + research + interview
L6–7
Outcomes + learningproposed tracking + refinement
LAYER 1
Client Profile

Resume, work history, education, skills, salary, location, interests, exclusions and priority conditions.

LAYER 2
Job Sources

Major job boards, employer sites, government, higher education, healthcare, nonprofit, NEworks and other accessible public sources.

LAYER 3
AI Screening

Salary, employment type, qualifications, credentials, experience level, exclusions and practical fit.

LAYER 4
Human Review

Career planner and client verify important facts and decide whether an opportunity should be pursued.

LAYER 5
Application Support

Truthful resume alignment, employer research, STAR preparation and interview questions.

LAYERS 6–7
Proposed Outcomes + Learning Loop

If piloted, track lead → application → interview → offer → placement and refine the strategy based on results.

Implementation materials already created

Leadership + staff materials

Executive Brief

Leadership discussion

Full Proposal

Scope + concept

Career Planner Guide

Proposed staff workflow

Client Quick Start

Participant guidance

Responsible AI Controls

Safeguards

Pilot Measures

Future evaluation

Development path

Problemfragmented search
Architecture7-layer model
Prototypeworking site
Safeguardshuman review
Pilot designnot yet launched

Proposed pilot measures

Measures defined for a future controlled pilot:

Quality

Are surfaced jobs actually worth discussing?

Time

Does the workflow reduce repetitive search work?

Conversion

If used by participants, what happens from lead to application to interview to offer?

Errors

Where does the workflow produce weak matches or require stronger guardrails?

Responsible AI

The model is designed so AI supports decisions rather than making eligibility, benefit, sanction or service determinations. It calls for minimum necessary data, truthful content, verification of important facts and human review before action.

My proposed role

I created and built the concept and currently serve as its originating developer. For a future controlled pilot, I have proposed serving as Project Lead to standardize the workflow, coordinate testing, support staff, track agreed measures and present recommendations.