◆Work

Real integrations. Plainly described.

◆Platform build / Franchise advisory

MatchRight: a full decision platform, built end to end

From a consulting method on paper to a live product people pay for and finish

Client: MatchRight Advisors / matchrightadvisors.com

◆The problem

MatchRight Advisors helps people decide whether buying or starting a business, including a franchise, is right for them. The method was proven in person: the MatchRight 7-Step Decision Making System. It lived in documents and conversations, which meant it could only reach as many people as the founder had hours. They needed it to work as a product someone could start on their own, pay for, complete, and come out of with a written answer.

◆The build

We built the whole platform: the public website, the member program, the scoring, the AI reports, the payments, and the internal tools the MatchRight team uses to run it. One codebase, one database, shipped in stages while real users were going through it.

◆What we built
The program itself
Seven decisions in sequence. Each one has a video, a scored assessment and a workbook, with progress saved as people go. The first decision is free, so people experience the method before they are asked to pay.
Scoring that matches on both sides
Results are scored in the browser for speed and again on the server as the record. A build check keeps the two copies identical, so a member never sees one result and gets a report about another.
AI reports grounded in the method
After each decision, an AI model writes an interpretation of that person's result. At the end it produces the MatchRight Readiness & Alignment Report. The model works from the scoring and the instructions MatchRight's own team wrote, not from general knowledge.
An editor for the people who own the method
MatchRight staff change how the AI interprets each decision in a plain-language editor: one instruction paragraph per decision, with draft, review and publish. No code and no developer in the loop.
Onboarding, paywall and payments
A short fit profile, the free first decision, then a paywall with live card payments, friend credits and company access keys. Members sign in with email, Google or a phone number.
Company and team side
Employers can sponsor people through the program. Each company gets its own branded workspace, seats and access keys, and administrators who manage their own people.
Internal dashboards
A team dashboard on its own sign-in address shows who is enrolled, where each person is, and which decisions stall people. A pilot group can be invited by email or link and tracked separately.
Reminders and follow-up
Lifecycle emails bring people back to the next decision. Contact, consulting and webinar forms route to the right inbox with a branded confirmation.
Readable by search engines and AI
Every public page carries structured data, and the site publishes a plain-text summary for AI models, so the company is described accurately when someone asks an assistant about it.
Status

Live, with paying members. Still shipping.

◆Website and AI visibility / Franchise consulting

Franchise Growth Partners: a website AI models can read

A full site overhaul for search engines and for ChatGPT, Claude and Perplexity

Client: Franchise Growth Partners / franchisegrowthpartners.com

◆The problem

Franchise Growth Partners is a franchisor consulting firm that helps business owners franchise their companies. The firm had decades of real experience and a website that did not show it to machines. Search engines and AI assistants could find the company by name, but had little to go on when someone asked a general question like who helps a business become a franchise. Some AI crawlers were not clearly permitted to read the site at all.

◆The build

We audited the site the way a search engine and an AI model see it, then rebuilt what they read. Every page got a new title and description. Priority pages were rewritten around the questions buyers actually ask. We published a plain-text company summary written for AI models and opened the site to the major AI crawlers. All of it was done directly on their existing website, with no rebuild.

◆What we built
A full audit, written as a plan
Crawler access, page titles and descriptions, content structure, how clearly the company and its services are defined, and the signals AI systems use when deciding which firms to name. The findings came back as a prioritized list of changes, not a generic report.
Titles and descriptions on every page
Every page rewritten and published, so each page says plainly what it is about.
Priority pages aimed at real questions
Priority pages reworked around franchise consulting, franchise growth, advisory services and location, so the firm shows up for the category and not only for its own name.
A summary written for AI models
A plain-text file at the root of the site that tells a language model who the firm is, what it does, who leads it and where the important pages are.
Crawler access
The site now explicitly permits search engines and AI crawlers. A later audit confirmed the AI crawlers we tested could read it.
Tracking how AI describes the firm
An ongoing check of how the company appears in AI answers over time, so changes can be judged by what assistants actually say.
Status

Live. Visibility tracking is ongoing.

◆Data and AI access / Consumer products

From a COBOL system to a company brain

Order data moved out of a legacy ERP into a secure modern database employees can ask questions of

Client: A product company (name withheld)

◆The problem

The company runs on a COBOL-based ERP. It works, and the whole business depends on it, but the data inside is hard to reach. Values are stored as codes, reports come out through an old export tool, and the online store and shipping system each keep their own version of the same orders. Answering a simple question, like where an order stands or how a salesperson is doing this year, meant asking the few people who know the system.

◆The build

We moved the company's data out of the legacy system into a secure modern database, without touching how the ERP itself runs. Orders from the ERP, the online store and the shipping platform are matched into one record. Every code is stored with its name. On top of that sits a connector that lets employees ask questions in ChatGPT and get answers from the company's own data.

◆What we built
A modern copy of the legacy data
Order, customer and salesperson data loaded from the ERP into a structured database, moving from manual exports to a scheduled feed.
One reconciled order record
Each order is matched across the ERP, the online store and the shipping system. Where the systems disagree, the record flags it. It never guesses.
Codes turned into names
Status codes, order types and salesperson numbers are stored with verified labels, so an answer says a name, not a number.
An AI connector with permission roles
Employees ask questions in plain English through a connector built on the Model Context Protocol. Access is governed by roles in the database for sales, accounting, operations and leadership. The first users are testing with full read access before roles are assigned person by person.
Safety by design
Everything is read-only. The connections to the store and shipping systems can only fetch data, enforced in code, and free-form queries are checked before they run.
Status

In progress. Next: more of the ERP's history, per-person roles, and workflows employees set up for themselves.

◆Accounting / CPA

CPA Firm Client Intake Automation

From hours of manual intake to a guided portal and automatic data extraction

◆The problem

A mid-size CPA firm in Thousand Oaks was drowning during tax season. Each new client required hours of manual intake work: collecting documents via email, entering data into their practice management software, verifying information, creating client folders, and sending welcome packets. Their admin staff was working overtime just to keep up. Errors from manual data entry were causing rework, and clients were frustrated by the slow onboarding process.

◆The build

We built an automated intake pipeline. New clients receive a branded portal link where they upload documents (tax returns, W-2s, 1099s, etc.). AI extracts data from uploaded documents automatically - names, SSNs, income figures, deductions - and populates the practice management system. The system runs validation checks, flags discrepancies, creates organized client folders, and sends automated status updates to clients. Staff only intervene when the AI flags something unusual.

◆Legal

Law Firm Document Processing

Case files read, sorted and summarized automatically, with attorneys reviewing what matters

◆The problem

A personal injury law firm in Calabasas was spending enormous amounts of time on document intake and review. New case files arrived as stacks of police reports, insurance correspondence, and court documents - sometimes hundreds of pages per case. Paralegals spent hours per case just reading, categorizing, and summarizing documents, and document processing was their biggest bottleneck.

◆The build

We built a document processing pipeline that handles the entire workflow. Documents are uploaded (or received via email) and AI reads every page, classifying documents by type (billing, correspondence, legal filings). It extracts key information - key dates, providers, amounts billed - and creates structured summaries. The system generates a case timeline automatically and flags critical items that need attorney review. Everything integrates with their case management software.

◆Real Estate

Real Estate Lead Automation

Inquiries answered quickly, qualified and followed up automatically

◆The problem

A real estate team in Simi Valley was losing leads to faster-responding competitors. Website inquiries sat in an inbox before anyone followed up. By then, the prospect had already talked to three other agents. The team knew speed mattered, but they couldn't watch their inbox all day or respond to every inquiry quickly while also showing properties and managing transactions.

◆The build

We built an AI lead response system. When a new inquiry comes in - from the website, Zillow, Realtor.com, or any source - the AI responds quickly via text and email. It has a natural conversation with the prospect: asks about their budget, timeline, neighborhoods of interest, and whether they're pre-approved. It qualifies leads based on the team's criteria, schedules showings for hot leads, and nurtures cold leads with automated follow-ups. All interactions sync to their CRM with full conversation history.

◆The next step

Bring us the bottleneck. We will find the right way through.

One working session to understand the business, clarify where AI fits, and identify the most practical next move: strategy, a website or app, an agent, automation, or a custom integration.

Jake Read, founder of Withn
Jake Read, founder of Withn
Or email jake@withn.tech

Let's Talk

START YOUR
AI JOURNEY

Ready to integrate AI into your business? Reach out directly.

Contact Details

jake@withn.tech(805) 390-8416

Service Area

Ventura County first: HQ in Westlake Village, in person across Thousand Oaks, Oxnard, Ventura, Camarillo, Simi Valley and the rest of the county. Clients across the United States served remotely.