Support and CX operations

Johnny Khabushani

I run customer support in enterprise B2B SaaS, and I build the AI the function actually needs. Ten years in the queue and the org chart. This past year I shipped it for real: Throughscan, an independent support-ops audit that is live and taking payment, plus a paid iOS app, an internal support assistant, and document pipelines.

Ten years in enterprise B2B SaaS  ·  Production AI, shipped and running

Scroll

The prototypes behind Throughscan.

Each of these answers one question I kept asking inside a support org and could never get a straight number on. They run entirely in your browser, nothing uploads anywhere. The full audit at throughscan.com now does both, and the rest of the read, on your own export.

Browser tool

Recontact Analyzer

A resolution rate is measured the moment you close a ticket. This measures what happens after: how often a resolved issue comes back as the same account opening the same kind of ticket inside a window you set. Upload a ticket export and see the honest number.

Open the tool →
Browser tool

Knowledge Base Auditor

Before you ground an AI on your help center, you need to know what state it is in. Paste a Zendesk help-center URL and it reads every published article and flags what is quietly rotting: pages years out of date, near-duplicates, and articles too thin to answer the question.

Open the tool →

Production systems I built and run.

Customer Support

Throughscan

An independent support-ops audit I built as its own product and business, now live and taking payment. A support leader uploads one ticket export and gets the read most dashboards hide: a plain-language diagnosis, their real median and p90 response times, what customers actually contact them about, and starter macros and a knowledge base drafted from their own history. Cloudflare Worker, Claude, and Stripe, with every figure computed in code or labeled a draft.

How it's built →
Customer Support

Institutional-Knowledge Assistant

An internal assistant I built at my day job. First I systematized that two-person team's knowledge, and with no added headcount we absorbed roughly 30% ticket growth over two years, driven by client and product expansion, at a 72% single-reply resolution rate. The assistant came after: I loaded a decade of judgment in once, so the answers I used to repeat now generate themselves, grounded and human-reviewed.

How it's built →
Consumer AI

Work the Steps

A paid iOS app with a Claude-powered conversational guide: cost guardrails, crisis-safe escalation, on-device encryption, and App Store compliance. Live on the App Store, a subscription with a recurring AI cost and the unit economics worked out before launch.

How it's built →

An operator who builds.

I'm Johnny Khabushani. I spent ten years inside enterprise B2B SaaS running support, success, and implementation: the migrations, the workflows, the escalations, and the systems underneath all of it. In the past year I started building production AI for the function, directing it end to end and shipping it: Throughscan, an independent support-ops audit that's live and taking payment, plus an internal support assistant, a paid iOS app, and document-extraction pipelines.

That combination is the point: AI does the implementation, and operations judgment decides what gets built, what will break, and what it should cost. Everything in the Lab is something I designed and run in production, with real users and real unit economics behind it.

If you're leading or building a support or CX function and you want AI that actually holds up, that's the background I bring to it.

Always glad to compare notes.

If you're working on support, CX, or AI that has to hold up in production, say hello. I read every message.

I reply within one business day.

Message sent. You'll hear from me shortly.