TTelevika · Product Team

How we rebuilt our product process with AI

One end-to-end pipeline — from hearing the customer to handing off to delivery — where AI does the heavy lifting at every step and people make the calls.

Prepared for the CEO· 8 steps from a real workflow· July 2026
Why this matters

Our whole product process used to be manual, slow and scattered.
Now it's one pipeline that runs every day.

AI reads thousands of reviews, watches the competition, audits the product, synthesizes insight, prototypes and drafts PRDs — so the team spends its time on judgment and decisions.

What follows are real samples of each step — not everything we do. We'll open each one live.

Before · scattered & manual
Reviews in a spreadsheet QA by hand Competitor notes Ideas lost in chats PRDs in random docs
Now · one connected pipeline that runs every day
The whole map

The 7-step product pipeline

1
Voice of Customer & Feedback Synthesis
listen at scale
product-audits · feedback
2
Competitive & Market Research
read the whole market
competitor-intelligence · Mobbin
3
Product Auditing & QA
audit that never sleeps
product-audits · E2E
4
Insight → Opportunity Synthesis
noise into strategy
Claude Artifact · Product Map
5
UX & Design Optimization
live, clickable prototypes
live prototypes · sketch boards
6
PRD Drafting
idea to spec, same session
html-docs · Figma mockups
7
PM Operations
chat becomes tracked work
decisions.jaan.to · roadmap
8
Impact Review & A/B Testing
close the loop, measure the bet
abtest-dashboard
1

Voice of Customer & Feedback Synthesis

LISTEN AT SCALE

AI auto-classifies thousands of reviews and cancellations, and surfaces the strategic patterns behind them.

2,000
feedback items analyzed
4,323
cancellations, reason-mined
52%
of cancels driven by price — the top lever
Sentiment + categorization by Claude Usage trends: watch-time & unique users over time
product-audits.vercel.app/#/filimo/user-feedback
.../filimo/user-feedback
Televika · Monthly Watch-Time Trend
2

Competitive & Market Research

READ THE WHOLE MARKET

We classify competitors' reviews at a scale of hundreds of thousands to find exactly where the market fails — and where we can win.

224,441
competitor comments classified
18
rivals, local & global
77%
complain about rivals' subscription model
86% of rivals' users want a bigger catalog Global rivals' UX patterns straight from Mobbin MCP
competitor-intelligence · Mobbin MCP
televika-foreign-comment-analysis-d.vercel.app
Mobbin MCP · Design Benchmarkingreference patterns pulled from real apps
benchmark → play
Netflix · Paywall
Disney+ · Player
YouTube · Shorts entry
Spotify · Onboarding
Max · Checkout
Prime · Cancel flow
Televika vs. benchmark
Onboarding steps4 / 6
Checkout clarityB+
Shorts discoverygap
3

Product Auditing & QA

AN AUDIT THAT NEVER SLEEPS

AI auto-tests every critical path each morning — and a 24/7 monitor pings Telegram the instant the site goes down.

7:15
AM daily E2E — automated (Tehran)
7
E2E tests · desktop + mobile
24/7
uptime watch → instant alert
Login · Payment · Player · Search · Homepage Site down? auto-alert to Telegram via n8n
product-audits · E2E + uptime alarm
.../filimo/e2e-tests
Uptime monitor → Telegram alarm (n8n)
4

Insight → Opportunity Synthesis

NOISE INTO STRATEGY

AI folds hundreds of scattered issues into one product heatmap — every block sourced, tagged Fix / New / Idea, and prioritized.

166
issues logged → 5 critical areas
34,900
competitor reviews + 38 sourced benchmarks
58
tasks in flight
Mapped along the user journey: awareness → revenue Every recommendation traces back to its source issue
claude.ai · Televika Product Map (Artifact)
claude.ai · Televika Product Map — user-journey heatmap
5

UX & Design Optimization

LIVE, CLICKABLE PROTOTYPES

From idea to a live, clickable prototype — we build UX concepts on the real product so they're tested before a line of code.

Live
clickable "Most-popular" prototype on the real UI
Hint Discovery sketch board with a method map Editable & maximizable for review
most-popular-prototype/v2 · html-docs/shorts-ideas
.../most-popular-prototype/v2
.../entry-points-ideas-for-shorts
6

PRD & Documentation System

IDEA TO SPEC, IN A LIVING SYSTEM

PRDs and docs live in one AI-built system — specs, visibility rules and Figma mockups, with every experiment variation documented side by side.

Visibility rules & product logic spelled out Samples: subscription promo · per-month pricing experiment
html-docs/subscription-promo · product-audits/docs/per-month-pricing
html-docs-29u.pages.dev/subscription-promo-touchpoints
.../filimo/docs/per-month-pricing-experiment
7

PM Operations

AN AI-RUN TASK SYSTEM WE CONTROL

Our roadmap is a task-management system that Claude runs and keeps current — lighter and faster than off-the-shelf PM tools, and flexible enough to reshape to exactly how we work.

Simpler & less time-consuming than typical PM tools Built on Galaxy MCP — we change & improve it freely
decisions.jaan.to/roadmap
decisions.jaan.to/roadmap
8

Impact Review & A/B Testing

CLOSE THE LOOP — MEASURE EVERY BET

We close the loop: every bet becomes a measured A/B test. AI helps design, run and read experiments — so we ship what wins and kill what doesn't.

Growth % + export on every experiment Each test linked to the insight that spawned it
abtest-dashboard.vercel.app
abtest-dashboard.vercel.app
The payoff

One complete pipeline, powered by AI — from hearing the user to measuring the impact.

👂1
Voice of Customer
🔭2
Market Research
🛡️3
Audit & QA
💡4
Insight → Opportunity
🎨5
UX & Design
📝6
PRD & Docs
⚙️7
PM Operations
📈8
Impact & A/B
🔭

Scale

224,441 competitor comments, thousands of reviews, daily audits — coverage no headcount could match.

Speed

From insight to a linked backlog and a PRD in the same session — not the same quarter.

🔗

Traceability

Every recommendation is footnoted to its evidence. AI here isn't a black box — it's an auditable chain.

We don't just use AI — we build our own AI tooling and compound it. That's the difference between "adopting AI" and "engineering with it."
Navigate with ← → or click · F for fullscreen