BIG IDEAS. BOLD MOVES. MEANINGFUL IMPACT.

Here's What We Have Created

CASE STUDIES

Real Problems, Real Systems

Every engagement here started the same way: a company had a real operational gap, whether that was no central system to run decisions off of, an org structure that hadn't kept pace with growth, or a process too inconsistent to scale safely, and needed someone to design the fix and see it through to something people actually use every day. The work below spans operations systems, organizational design, training frameworks, and AI-driven tools, but the throughline is the same in each: identify what's actually needed, build it with the right guardrails in place, and hand over something functional, not just a plan on paper.

Building a Real-Time Operations Dashboard for Executive Decision-Making

Case Management Consulting Firm (VA Disability Claims)

Context

The client manages VA disability claim casework through a network of roughly 30 independent contractor case managers. There was no central database or dashboard in place, so company executives had no consolidated, real-time view of intake volume, assignment speed, caseload distribution, or revenue.

Challenge

I identified the need for a system that could serve as the company's central database and give executives a live, visual picture of operations to support decision-making. That meant handling two distinct client channels (centrally routed intake and case-manager-direct intake), enforcing a compliance gate before any case manager could be assigned new clients, tracking payments and revenue by source, and automating the auditable fields (like assignment dates) that had previously depended on manual entry.

Approach

Working iteratively in short, verified phases, I designed the system architecture and data model, then built it out in Airtable: a Clients table with automation-stamped audit fields, a Case Managers table with compliance and capacity tracking, and a Finance Transactions table tied to automated notifications. I designed the DocuSign-to-Zapier intake pipeline, including a routing mechanism that identifies which channel a new client came through based on the DocuSign recipient address, and built role-based dashboard interfaces so intake staff, case managers, and finance each see only what's relevant to their role.

What Was Delivered

A live database and dashboard system holding over 800 client records, including:

An automated intake pipeline that creates and routes new client records without manual entry
A compliance gate that only allows assignment to case managers meeting all required conditions (active status, accepting new clients, required agreements signed)
Executive-facing dashboards tracking intake volume, average days to assign, aging unassigned cases, caseload by case manager, and revenue by source and by case manager
A portal experience for case managers to view and manage only their own assigned clients

Case managers are onboarded as system users, notification automations are active, and payment tracking is fully integrated, giving executives one live source of truth for operations and revenue.

Redesigning Org Structure and Accountability Lines

Client Management Consulting Firm, Operations Restructuring

Context

As the client scaled its case management network, its existing referral and operations structure had no clearly defined compliance function, and reporting lines between leadership, case managers, and support staff had not been formalized in writing.

Challenge

The client needed a revised Client Referral Processing SOP and Director of Operations SOP, plus a new organizational structure that could support growth without creating supervisory overreach into independently contracted case managers, whose relationship to the company needed to stay non-supervisory to preserve their contractor status.

Approach

I worked directly with the client's leadership to separate two functions that had previously been blended: the design of frameworks, systems, and training (my role as Solutions Architect) versus the execution and monitoring of those systems day to day (a new Director of Compliance and Performance role). I defined the new director role's scope carefully, including monitoring-only dashed-line relationships to case managers and affiliated physicians, and routed all resulting SOPs and contracts through the client's legal advisor before implementation.

What Was Delivered

A new org structure with a Chief Solutions Officer role absorbing operations, a Director of Compliance and Performance and Referral Manager reporting into it, and a dashed-line advisory ring of legal, accounting, and bookkeeping to the President.

Supporting SOPs define scope, reporting lines, and non-supervisory boundaries in writing, and the structure is in place and guiding how the company operates.

A Conversational Agent for Lead Qualification

AI Solutions Build

Context

Small service businesses and consultants often lose prospective clients to a plain contact form: visitors either abandon a long form, or submit vague inquiries that require a manual back-and-forth just to figure out what they actually need before a real conversation can start.

Challenge

The goal was a lightweight, embeddable agent that could hold a short, natural conversation with a website visitor, extract the handful of details a business actually needs to qualify a lead (name, email, what they need, timeline), and hand off a clean, structured record, without exposing raw model output or a clunky multi-step form to the visitor.

Approach

I designed the agent around a dual-channel response pattern: the model is prompted to reply conversationally in plain language while appending a hidden, structured JSON block after every turn containing whatever lead details it has learned so far. The interface strips that block before displaying the message, uses it to update a live "current lead" summary panel in real time, and automatically saves the lead once the minimum required fields (name, email, need) are present. This avoids a separate extraction step or follow-up API call, the qualification data comes out of the same conversational turn the visitor sees.

What Was Delivered

A single self-contained chat widget with:

A conversational front end that asks one or two questions at a time rather than presenting a form
A live lead summary panel showing exactly what's been captured and what's still missing
Automatic, gated saving, a lead is only stored once it's genuinely qualified, not on every message
A running history of previously captured leads, visible in the same interface

The agent is deployed and actively qualifying and capturing leads for its client.

Building a Scalable Academic Support System for a Microschool Leader

K-6 Blended-Learning Microschool

Context

A large-scale training and change-management initiative had generated valuable resources scattered across multiple systems: SharePoint libraries, project portals, Teams and Slack channels, workshop documents, job aids, workbooks, FAQs, and event materials. The information existed, but participants, facilitators, and leadership all struggled to find the right answer quickly during live onboarding sessions, learning labs, and hands-on events.

Challenge

Participants needed timely answers in the moment. Facilitators needed fast access to approved resources without digging through folders. Leadership needed visibility into recurring questions, readiness risks, and unresolved barriers across the program. All of this had to operate inside a sensitive, high-accountability environment, meaning any solution needed real guardrails around source accuracy, privacy, and knowing when to hand a question to a human.

Approach

I designed a controlled retrieval and guidance assistant built around three distinct, role-based experiences: a Participant mode for real-time Q&A during events, a Facilitator mode for retrieving job aids, workbooks, and escalation paths, and a Leadership mode surfacing recurring questions, readiness concerns, and workflow friction. Rather than a general-purpose chatbot, the assistant only draws from curated, approved training materials, categorizes incoming questions to surface patterns, and is built to explain its answers and escalate clearly whenever a question needs subject-matter or human review instead of guessing.

What Was Delivered

A live assistant supporting:

Real-time Q&A during Teams calls, learning labs, and onboarding events
Role-readiness coaching and guided next steps
Retrieval of approved job aids, workbooks, and resources
Categorization of questions to reveal patterns and knowledge gaps
Clear escalation guidance for issues requiring subject-matter expertise
Workshop pulse summaries and drafting assistance for learner follow-ups

The assistant is in active use across the program's training and change-management events, cutting down time spent searching for information, promoting more consistent guidance across facilitators, and giving leadership a structured, real-time view into where additional communication or reinforcement is needed.

Centralizing Scattered Training Materials Into a Single Source of Truth

Training Resource Hub, Large-Scale Change Management Program

Context

Ahead of building the Training Navigator Assistant, the same program's training materials, forms, job aids, videos, and resources were spread across disconnected folders and channels with no single place anyone could point to as the current, correct version of anything.

Challenge The program needed one central home for every training resource that facilitators and participants could rely on, but static file storage alone wasn't enough: forms and materials needed to stay easy to update in real time, and users needed quick, reliable downloads without hunting through nested folders or outdated copies.

Approach

I built a SharePoint site to house all training materials in one place, then linked it to Google Drive so the underlying files could be updated in real time without anyone having to re-upload or re-share a new version each time something changed. This gave the hub two things at once: SharePoint's structure as the organized front door, and Google Drive's live editing behind it, so a form, resource, or video updated in Drive reflected immediately wherever it was linked.

What Was Delivered

A centralized training resource hub with:

All training materials, forms, job aids, and videos organized in one SharePoint location

Real-time syncing with Google Drive so updates go live without manual re-uploading

Easy, reliable downloads for forms and resources across the program

This hub became the underlying source material for the Training Navigator Assistant, giving it a single, current, well-organized set of approved resources to draw from rather than the scattered systems the program had relied on before.

Redesigning Org Structure and Accountability Lines

Client Management Consulting Firm, Operations Restructuring

Context

A microschool leader is often responsible for nearly everything: instruction, student progress, parent communication, staffing, scheduling, enrollment, and compliance. At this K-6 blended-learning microschool, students learned in mixed-age groups with widely varying academic needs, and the challenge was preserving individualized learning and a relationship-centered culture while building enough structure for the school to grow sustainably.

Challenge

Teachers needed a practical way to use assessment data, form flexible skill groups, plan differentiated instruction, monitor progress, and communicate clearly with families, all without creating an administrative burden that would fall entirely on the school leader.

Approach

I designed a leadership support framework across four connected areas. Academic visibility was built on periodic assessments (including NWEA MAP), classroom observation, and daily student work, using flexible skill-based grouping rather than grouping by age alone. Differentiated instruction systems for mixed-age classrooms included small-group rotations, tiered learning groups, daily skill tracking, standards-aligned differentiated lessons, and intervention and behavioral supports. AI-enabled leadership tools turned assessment results into student skill profiles, suggested instructional groups, priority learning gaps, and parent-friendly updates, without replacing teacher judgment. And clearer role definitions across the instructional leader, front-office support, teachers, intervention staff, and the Founder/Director made sure responsibility for daily operations, student support, and family communication didn't all default to one person.

What Was Delivered

A working leadership and instructional framework that gives the school:

More intentional, skill-based grouping and targeted instruction
Greater visibility into individual student needs and progress
Faster, more consistent parent-facing communication
Clearer staff roles and daily systems
A structure that scales without losing the school's small, relationship-driven culture

The framework is in active use, giving the school's leadership team the operational clarity to grow sustainably while keeping instruction personalized.

Creating an AI Operations Hub for a Microschool Leader

K-6 Blended-Learning Microschool

Context

Microschool leaders wear many hats: leading instruction and supporting students while also managing parent emails, social media, newsletters, enrollment inquiries, scheduling, and staff communication. At this relationship-centered school, families expected timely, warm, personal updates, and the leader still needed protected time for students, teachers, and academic planning.

Challenge

The challenge was reducing repetitive administrative work without losing the personal voice and trust families rely on. A generic automated tone would have been worse than the delay it was meant to fix, and any system touching student information had to keep sensitive details out of public-facing content by design, not by hoping someone remembered to check.

Approach

I designed an AI-supported operations system built around a single shared record of what's happening in the school, rather than a set of disconnected tools. A classroom update enters the system once and can generate a parent email, a newsletter section, a social caption, and an internal staff note, but all four drafts stay linked to that one source update so they stay consistent with each other and with what actually happened.

The system has five working parts:

Email Sorting Agent reviews incoming messages by topic, urgency, and required next action, and routes anything matching a defined escalation list (safety concerns, complaints, custody or legal matters, special-needs accommodations) directly to the leader instead of the drafting queue.
Parent Communication Assistant drafts responses in the leader's voice, trained on a curated set of the leader's own past messages across different tones (routine updates, corrections, celebratory notes) rather than a generic assistant tone.
Newsletter and Social Content Agents pull from the same approved classroom updates to produce a recurring newsletter and a posting calendar, so a single piece of content doesn't have to be written twice in two different voices.
Academic Insights Assistant turns assessments and teacher notes into parent-friendly summaries, built on the same no-clinical-detail intake standard used elsewhere in the school's data handling: paraphrased, scheduling-and-support framing, never a diagnosis or a file upload of clinical records.
School Operations Assistant handles checklists, staff updates, and reusable templates.

Every draft, regardless of which agent produced it, lands in one review queue. The leader approves, edits, or rejects from a single place rather than checking multiple tools, which is what actually protects the time savings the system is built for. Nothing goes out without that approval step, and no student-identifying detail moves from a private channel into a public one without a separate explicit check.

What Was Delivered

An AI Operations Hub that gives the school:

One shared system of record so a classroom update produces consistent, aligned communications instead of duplicated manual writing
A single review queue that keeps the leader as the final approver without turning review into a second full-time job
Defined escalation rules that route sensitive topics to the leader directly, before any drafting happens
A voice-matching process built from the leader's actual message history, not a generic assistant tone
More protected leadership time for relationships, coaching, and decision-making

The hub is in active use, giving the school a consistent communication rhythm while keeping the leader as the final decision-maker on every family-facing message.

AI-Powered Student Data and Instruction Planning

K-6 Blended-Learning Microschool

Context

Teachers and instructional leaders had access to NWEA MAP assessment data, but interpreting raw RIT scores, growth data, and domain-level results into an actual instructional plan takes expertise most classroom teachers don't have time to apply consistently, especially in mixed-age, mixed-grade classrooms. The raw exports themselves were also a barrier: individual student reports came through as image-only PDFs, and class-level data arrived in a format standard AI tools couldn't read natively.

Challenge

Any tool touching student assessment data needed to be right, not just plausible. Teachers needed reliable instructional levels, skill gaps, and grouping recommendations without the risk of an AI system confidently inventing a score or a gap that wasn't actually in the data, and the tool had to work within a Custom GPT platform's real limitations, no code interpreter, no native spreadsheet reading.

Approach

I designed a Custom GPT built around six connected layers of instructional intelligence: assessment interpretation, instructional planning, instructional delivery, classroom-level grouping, progress monitoring, and growth tracking, all governed by strict evidence-integrity rules that tag every finding as Verified, an Interpretation, or Insufficient Evidence, with fabricated scores never permitted. I solved the underlying data problem first: building an OCR preprocessing step for the image-only individual student PDFs, and working around the platform's lack of spreadsheet support to make class-level data usable. I then tested the system extensively, verifying individual student diagnostics, multi-length instructional plans, lesson and worksheet generation, and full classroom grouping against the original data by hand before trusting any pattern it produced.

What Was Delivered

An assistant that:

Turns raw MAP assessment data into individual student skill profiles with priorities traced directly to the evidence behind them
Builds 60, 90, and 180-day instructional plans that adapt at checkpoints as new assessment data comes in
Generates standards-aligned lessons and worksheets with fully verified answer keys
Groups an entire mixed-grade classroom into skill-based instructional groups from a single class data export, without misplacing strong students based on one weak domain
Produces parent-friendly summaries that translate assessment results into plain language

The assistant is in active use, giving teachers and instructional leaders a fast, evidence-checked starting point for turning assessment data into concrete instructional decisions, applying AI directly to instruction rather than just the administrative work around it.