nikam.dev ← All cases Telegram →
nikam.dev / Cases / AI product from zero to production
Live · AI · Voice · Payments

AI product - from idea to paying users.

Konsilion. B2C SaaS with 105 AI advisors: business strategists, philosophers, investors, athletes. A board of several advisors, «Stratagem» - deep reports with live web search, ideas by voice, advisors that react to the news and argue with each other. 12 languages, auto-renewal and one-tap cancellation, payment in Stars and rubles. One Pro subscription covers all 105.

What
B2C SaaS · AI chat with voice
Audience
Global, 12 languages
Payment
CloudPayments + Telegram Stars
Status
Live · 105 advisors
Open konsilion.nikam.dev →
- The problem

AI products often don't reach revenue - because unit economics weren't counted from day 0.

The demo is beautiful, users come. A month later the founder opens the API bill and discovers each free user costs $3 a day. A subscription can't cover this. The product dies not from a bad idea, but from not pricing the cost of each AI request.

Pain 01
AI request cost not priced into the subscription
Launch without understanding «how many AI requests an average user spends per day». Subscription $9.99/mo, cost $50/mo - on each active user. The more successful the product, the bigger the loss.
Pain 02
A hundred characters, one voice
A line-up of AI characters degrades fast: they all answer in the same polished «assistant voice». The stoic, the marketer and the hockey coach give the same advice - the product loses the very point it was bought for.
Pain 03
Payments for an international audience - a headache
A user in Russia can't get a card through, someone else's foreign acquirer won't work, another needs Telegram Stars right in the chat. Blocks, subscription cancellations with no clear reason. The founder spends time on payments instead of the product.
Pain 04
Character evolution - chaotic
Each new advisor brings its own prompt, its own knowledge, its own edge cases. After 20 characters the code is spaghetti, and updating everyone's knowledge at once is a separate project.

B2C SaaS with a hundred AI advisors - with honest unit economics from day 0.

Not «a hackday demo». A paid product with a subscription. Each advisor is a character, a knowledge base and a style; the user consults one, assembles a board of several, or orders a «Stratagem» - a report with live web search.

105 advisors - each with their own knowledge base
Each has a character, principles, a knowledge domain (a RAG base: sources, decision cases, mental models) and their own manner of speech. Anti-convergence is built in: different advisors' answers must differ, similarity is caught by an automatic quality guard.
«Stratagem» - a report written by several advisors
The AI picks 3-5 relevant advisors for the question; each writes their own section - with live web search and cited sources. Then the board dissects the disagreements. The report runs on credits and survives to the end even if the tab is closed.
Ideas - by voice, on the go
Dictate a thought - it's transcribed to text and lands in the idea vault. Then a board session can be run on it: discuss, hard stress-test, develop, or turn into an action plan.
Advisors live
They react to fresh news in their own style and argue with each other in pairs; the reactions form a weekly feed. Updated knowledge is translated to the user's language automatically.
Memory and an onboarding interview
Konsilion interviews the user on first entry: business, goals, context. Facts from conversations are remembered (and encrypted), advice becomes personal. The memory is visible to the user and managed by them.
Auto-renewal + one-tap cancellation
Subscription auto-renews via Telegram Stars and CloudPayments - no portal needed. Cancellation is one button on the subscription page - no chatbot retention. Higher trust, lower churn.
Coupons and a partner programme
Coupons with validity windows and usage limits, partner codes with a commission - the partner sees their stats and accruals in a dashboard. A marketing lever without rewriting billing.
23 advisors - free
You can try without a card or a subscription: 23 advisors are open to everyone. Payment runs on two rails: CloudPayments for ruble cards, Telegram Stars right in the chat - for everyone else.

Architecture that scales from 10 to 10 000 paying users.

Not «MVP we'll rewrite later». Built in from day 0: cost accounting for every AI call, a provider-agnostic gateway, a structured advisor model, two-rail billing.

/ 01
Cost accounting per call
Every LLM call lands in the ledger: model, tokens, price, latency. Daily limits per tier, spend alerts, telemetry. The economics are visible before they become a problem.
/ 02
Provider-agnostic gateway
The model fits the task: a fast one for chat, deep ones for «Stratagem» sections, a live-web model for search. A provider changes its pricing - migration without rewriting the product.
/ 03
The advisor model
An advisor = config + prompt + knowledge base + index. A new one is added with files, without code changes. Knowledge is versioned and auto-translated into 12 languages.
/ 04
Two-rail payments
Telegram Stars + CloudPayments. Stars work right in the chat even where a card won't. Webhooks handle all events: renewals, cancellations, refunds - with a 21-day refund window.
- Outcome

What came out after launch.

105 advisors
In the catalogue - each with their own character, knowledge base, style. All 105 covered by one Pro subscription; 23 free for everyone.
12 languages
The interface, all 105 advisor bios and their updates translate automatically. A user in Jakarta and in Berlin sees the product as native.
3-5 advisors
Assembled by «Stratagem» for one question - a report with live web search, where each section is written by a relevant advisor and the disagreements get their own pass.

How it looks.

A board of great minds at hand: ask your board, assemble a «Stratagem», a morning briefing - and a catalog of 105 advisors.

Catalog of 105 advisors
105 advisors - by interest and match with your profile
Dashboard - ask your board
Dashboard: ask your board, quick scenarios, morning briefing

AI product in your head or in pre-MVP?

I take you through all four traps where AI products usually die: unit economics, prod vs demo, payments, extensibility. If you have an AI product idea and need a founder-engineer in the loop to get it to paying users - let's talk.

Message on Telegram → nick@nikam.dev