WEB_BOOT_V.2.5
INITIALIZING
nikita sych / product designer
nikita sych / product designer
sych

MIRA AI PAYWALL FOR $500

A successful take-home assignment that almost landed me a role on the product team.

New Mira AI paywall, Control B
New Mira AI paywall, Control C

context

A month ago, a recruiter from The Open Platform reached out and invited me to interview for a Middle+ Product Designer role at Mira. I already knew the product, and the opportunity really caught my interest. We quickly had a screening call, and I received a take-home assignment.

task

Design a paywall concept for Mira AI — a native AI agent in Telegram with 600K+ MAU.

  1. Analyze the current iteration: what works and what could be improved
  2. Research the market and competitors:
  3. Propose a paywall concept

The solution may include any number of screens. It should be designed within the Telegram Mini Apps guidelines for iOS. You may improve the current version or create a new one from scratch.

discovery

what we’re designing

A new paywall for Mira AI — a native AI agent inside Telegram. The scope is at least one Mini App screen.

why

We [the Mira team] don’t really like it [the current paywall]. We see a lot of problems with it, but it can serve as a starting point.

I needed to identify problems in the current flow and main screen, fix them, and propose hypotheses for improving key metrics and the user experience.

success criteria

Primary metric — purchase conversion
Secondary metrics: average order value, churn rate, and LTV

segments and scenarios

  1. When tokens run out or the balance is insufficientThe user is working on an important task, but the free limit is not enough. At this point, their motivation to continue is high. This is the primary paywall scenario.
  2. RenewalThe subscription has expired, and the user needs to decide whether to renew it.
  3. Limited-time offerThe user knows the product but is still on the free plan. A discount or promotion can trigger FOMO and encourage a purchase.
  4. Intentional purchaseThe user already knows the product and opens the pricing page without an external trigger to match the plans to their needs and choose one.

pipeline

  1. Analyze the current paywall and funnel
  2. Run a benchmark
  3. Generate and prioritize hypotheses
  4. Design the hypotheses
  5. Prepare a validation plan
  6. Present the solution

important

Most user interactions with Mira happen through a TMA (Telegram Mini App). It mimics a native app but is technically a web application. Settings, memory, content-generation parameters, and other features live in the Mini App because Telegram’s chat interface does not support quickly implementing strong UX patterns from other well-known AI products (Hermes, OpenClaw, Claude, ChatGPT, and others).

The low entry barrier is both an advantage and a constraint. Looking ahead, this is the trade-off I would focus on (and I’m confident I could do it successfully).

Current paywall analysis

Control A — Mira AI screen
Current Mira AI paywall
  1. The “Professional AI models” block clashes with the rest of the interface.
  2. Four benefits are presented as two, making the value proposition feel stretched and unconvincing.
  3. The plan-switcher emojis create visual noise because they convey nearly the same emotion. When everything is emphasized, nothing is.
  4. Scheduled tasks are already available on the free plan with limits. The benefit should either be reframed or removed from the spotlight.
  5. “For everyday AI work” and “For heavy users” can work both ways: some users will identify with a label, while others may feel excluded. The bright purple also creates an unnecessary visual anchor.
  6. FOMO could be stronger if the old pre-discount price were shown.
  7. “Gift to friends” is a secondary scenario but takes up half the screen alongside the primary subscription button.
  8. The copy says users can pay with Stars or a card, but tapping Subscribe opens Stars payment only—most likely a bug.

benchmark

Telegram — Mira AI screen
Telegram
Claude — Mira AI screen
Claude
ChatGPT — Mira AI screen
ChatGPT
Perplexity — Mira AI screen
Perplexity
Victor — Mira AI screen
Victor
Cal AI — Mira AI screen
Cal AI
Poke — Mira AI screen
Poke
Grok — Mira AI screen
Grok

general takeaways

  1. Grok’s paywall is nearly perfect.
  2. I like how Perplexity emphasizes its main advantage: access to powerful models.
  3. ChatGPT uses contrast and isolation — both are effective techniques.
  4. No one offers weekly or three-month plans. These plans tend to hurt revenue more than help it. Weekly is usually replaced by a one-to-three-day trial, while a three-month plan does not clearly match any segment’s needs.
  5. The first paywall screen should be self-sufficient: users should be able to complete every available action without leaving it.

pricing

To design the paywall well, I first needed to understand the value proposition in detail. I asked both Mira and Claude about the benefits. Their answers differed, and many points did not match reality, so I worked only with the benefits I could verify. This produced the following table:

FeatureFreeProPro Max
Tokens02,0007,000
MessagesUp to 30 dailyUp to 1,000Up to 10,000
Image Generation–Up to 60Up to 200
Video Generation–Up to 10Up to 100
Access to top AI models–++
50% off on token purchases–++
No watermarks–++
Autonomous work 24/7–++
Less Content Moderation–++
1000+ app integrations–++
Scheduled Tasks (removed because it is included in every plan and overlaps with autonomous work 24/7)+++

important

Pro and Pro Max differ only in token allowance; every other benefit is identical across the two paid plans.

hypotheses

If the paywall is adapted to the user’s context, purchase conversion will increase because the self-reference effect can make the offer feel more personal and valuable.

Three personalized paywall variants
Three context-specific headings: Default (no context), Interrupt, Renew

If we keep only monthly and annual plans, conversion and LTV will increase while churn decreases: the choice becomes simpler, and removing weekly plans reduces short-term subscriptions that are canceled after a one-off task.

Simplified Mira AI pricing structure
Simplified pricing structure. Before / After

Fixing a bug that critically affected purchase conversion. The native payment widget allowed users to buy a subscription only with Stars, making card payments impossible. Update: it was fixed the day after I presented the assignment.

Fixed subscription payment flow
Subscription payment fix

If the benefits communicate their value more clearly, users will be more likely to understand the paid plan and purchase it because they can match Mira’s capabilities to their needs.

I explored this hypothesis in two variants

Prem1 — Mira AI screen
1 — Condense the benefits and briefly show what’s included.
Prem2 — Mira AI screen
2 — Side-by-side plan comparison using contrast and isolation.

result

New paywall demo
Control A — Mira AI screen
Before. Control A
Control B — Mira AI screen
After. Control B
Control C — Mira AI screen
After. Control C

how to validate the solution

Run an A/B/C test among users who opened the paywall in the same context: interrupt, renew, or default. Within each context, traffic should be split randomly and evenly across variants.

First, test whether reducing choice improves conversion and shifts users toward longer plans. Compare the current weekly / monthly / three-month / annual structure against the same interface with only monthly / annual plans.

Then compare Control B and Control C on the winning pricing structure and track the metrics below.

Primary metric — purchase conversion
Secondary metrics: average order value, churn rate, and LTV

Accept the new version if it delivers a statistically significant conversion lift without reducing LTV, increasing churn, or lowering the average order value.

what's next

The Mira team really liked the assignment, and I was invited to speak with Dasha, Mira’s CEO, and Ilya Grishin, TOP Design Director.

They were both warm and open. Each conversation lasted more than an hour. After the technical interview with Ilya, HR started a security check — the final stage before an offer. I expected the offer within a few days.

Around that time, Telegram’s core team decided to bring Wallet in-house and make it a built-in section of the app. The Wallet team was left without a project.

Three days later, I was told that the position had been filled internally: a Wallet designer was transferred to Mira AI. Still, I was paid a full $500 for the assignment!

I was disappointed because I had already fallen in love with the product, but in the end I somehow lost and won at the same time.

thanks for watching

If you have a question, an idea, a suggestion, or just want to chat, message me on Telegram @ihaveui