Case Study
A custom AI system for an executive advisory practice
Dr. Joleen Cowan is an organizational psychologist and executive advisor in California, and the founder of Leadership Archway. Her practice runs on a proprietary methodology she developed, and delivering it involved a large amount of manual work for every single client. ManiWebDev built her a custom AI system across four parts: a client assessment platform, an AI document pipeline that drafts her client deliverables, an AI assistant named Archie that qualifies website visitors around the clock, and an automated client program that runs on its own schedule. She reviews and approves. The system does the rest.
The challenge
Leadership Archway was built on something genuinely rare: a methodology developed by a practitioner with a PhD in organizational behavior, grounded in real research rather than repackaged advice. The quality was never the question.
The question was reach. Delivering that methodology meant collecting each client’s information by hand, working through it by hand, and writing every deliverable from scratch. It produced excellent work, and it meant the number of people she could help was capped by the number of hours she personally had.
The brief was clear: help more leaders, without losing the human connection that makes the work matter.
That second half is where most automation projects go wrong. The goal was never to put an AI between her and her clients. It was to remove the transcription, the admin, and the blank-page problem, so her time went to judgment and relationships instead of paperwork.
What we built
Four systems, delivered in sequence so each one was live and working before the next one started.
- 01
Client Assessment Platform
A private platform that delivers her proprietary assessment to clients and handles everything that used to be done by hand afterwards.
Clients move through it one question at a time on any device. Every submission lands in a secure admin panel with status tracking, internal notes, and bulk export. The methodology stays hers. The infrastructure around it stopped being manual.
- 02
AI Document Generation
The written deliverables clients receive are now drafted automatically from each submission, in her structure and her branding.
Every document runs through a generate, edit, and approve flow. Nothing reaches a client until she has read it and signed off. The AI removes the blank page, not the expert.
- 03
Archie, the AI Assistant
A chat assistant on leadershiparchway.com that engages visitors around the clock, understands what they need, captures their details, and sends her an instant notification.
Archie is scoped deliberately narrow. It qualifies and routes. It does not advise, diagnose, or imply an engagement has begun, and the site terms state that clearly to every visitor.
- 04
Automated Client Program
A multi-week email program that runs itself for each client, scheduled in that client's own timezone.
Clients get a magic-link portal with no password to remember, where they complete their check-ins. Sessions arrive with calendar attachments that add straight to any calendar app. An admin panel shows exactly where every client sits in the program.
Decisions that shaped the build
The expert stays in the loop. Every AI-drafted document requires her approval before a client sees it. It would have been easier to let the system send directly. For a practice where credibility is the entire product, easier was the wrong answer.
The methodology stays private. The platform is built around her framework without exposing it. Her intellectual property does not live in a third-party tool, and it is not visible to anyone using the system from the outside.
A narrow AI assistant beats a clever one. Archie was deliberately kept to one job: understand the visitor and route them. An assistant that tries to advise is an assistant that eventually says something the practitioner would never have said.
Off serverless, onto a small server. The backend started on AWS Lambda and moved to a Lightsail instance running PM2 behind Nginx. Scheduled sequences and cron work are a poor fit for serverless. A small always-on server removed an entire layer of complexity that was buying nothing.
No passwords. Executives will not create an account to fill in a form. They click a link in their email and they are in.
The stack
- Frontend
- Next.js
- Backend
- Node.js, Express
- Database
- MongoDB Atlas
- AI
- Claude and Groq models
- Infrastructure
- AWS Lightsail, PM2, Nginx
- Scheduling
- node-cron
- Documents
- Server-side PDF generation
What the client said
From building an incredible website to creating Archie (my AI assistant) and developing an automated intake process, he’s helping Leadership Archway scale so we can reach and support more people without losing the human connection that matters most.
What I appreciate most is that Usman combines technical brilliance with genuine partnership. He makes complex things simple, and he’s an absolute pleasure to work with.
Common questions
What does a custom AI system actually do for a coaching or consulting practice?
It takes the repeatable work out of the expert's hands without taking the expert out of the process. Intake, scoring, first-draft writing, scheduling, and follow-up sequences all run automatically. The consultant reviews and approves instead of starting every engagement from a blank document.
Why build custom instead of using an off-the-shelf platform?
Consultants who work from a proprietary methodology hit a wall with template tools. A form builder or a funnel platform can collect answers, but it cannot hold the logic a real methodology runs on, produce deliverables in the practitioner's voice, or keep that methodology private. Custom means the system fits the practice instead of the practice bending to fit the software.
Does the AI send anything to clients without human review?
No. Every generated document passes through a generate, edit, and approve step. The practitioner is always the last checkpoint before anything reaches a client. For a practice where credibility is the product, that is not optional.
Is the client's methodology or data exposed by the AI?
No. The methodology stays private to the practice. The system is built around it without publishing it, and client data sits in a private database the practitioner controls.
How is a project like this delivered?
In phases. Each system goes live and proves itself before the next one starts, so the practice sees value early instead of waiting months for one large launch. This project was built in phases through 2026.
Is your practice running on manual work?
If your methodology lives in documents and spreadsheets and you are the bottleneck, that is a system waiting to be built. Tell me what your intake looks like today and I’ll tell you honestly whether it is worth automating.
Start a conversation