"I know English, but I can't speak it": an AI speaking tutor from a small Tashkent team that explains in your native language climbs the education charts in Uzbekistan

Bandmate is an AI English speaking app from a small company in Uzbekistan: you talk in English to a mascot named Bandi, and when you make a mistake it explains in Uzbek or Russian. Launched in August, it reached as high as #19 in the Uzbek App Store's free Education chart in early November. The numbers are small, and it charted because the market's barrier is low. The approach isn't new. What's worth noticing is that it brought a category that already exists globally into a small-language country, and the cost of doing so was low.

Oct 12, a Reel with a promo code

On Oct 12, Tashkent UGC creator Nabiyeva Sarvinoz (@novasarvii_) and @bandmate.uz co-posted an Instagram Reel. The Uzbek caption reads: "Bandi bilan ingiliz tili oson va qiziqarli!" (English with Bandi is easy and fun!), with "Promokod: SARVIBES" below, and the tags include #rek, which means "ad."

Three days before that, on Oct 9, the company posted an XP contest on its website: the three people with the most XP by Nov 9 would each get 3, 2, and 1 months of Pro, with total prizes worth 1.19M so'm (about $95). The XP rules matter most: 10 XP per minute of English spoken, 20 XP per completed lesson, 80 XP per IELTS speaking practice, 3 XP when an invited friend completes 150 lessons, and 200 XP when a lapsed user comes back. This is really a month-long acquisition and win-back campaign, dressed up as a leaderboard.

Over the next ten days, its ratings picked up. The Uzbek iOS rating count was 10-28 at 63 and only rose by 1 on 10-29; after that, 10-30, 11-01, 11-03, and 11-06 each added about 10 a day, reaching 119 by 11-06 (App Store Uzbekistan storefront). Whether the contest or the creators drove the acceleration can't be told from the outside.

What it is

The last line of the store page states the problem it solves: "I know English, but I can't speak it." When you open the app, Bandi spends three minutes asking your level, how much time you have each day, and why you're learning English, then lays out a route from Starter to C2. Your daily core is holding the mic button and talking to Bandi.

The website gives an example: you say "I goed to the market yesterday," and Bandi replies "Nice! Just one thing: we say 'I went'." If you choose Uzbek explanations, it adds "'Go' ning o'tgan zamoni — 'went'" (the past tense of go is went). Explanations can be in Uzbek, Russian, or pure English, so both Uzbek and Russian speakers can use it.

Beyond conversation: story-style courses, timed IELTS Speaking Part 1–3 practice with score estimates, pronunciation drills, weekly leagues, and streaks. The leagues and streaks are modeled on Duolingo.

Pricing: the free tier allows 3 conversations per day, each up to 3 minutes. Pro costs 199,000 so'm per month (about $16), or 12 so'm for 1,699,000 months on the website (11-06). Payment can go through Apple in-app purchase, or inside the app with local Uzcard / Humo cards (via Click), Payme, or Uzum Bank (bandmate.uz/payments, updated 10-12).

How big it is

Look at it inside Uzbekistan's Education category. Same week, new ratings: Duolingo 126, local veteran Englify 90, Learna 52 (Turkish Codeway, paid acquisition), Bandmate 45, Praktika 7, Loora 5, ELSA 1 (App Store Uzbekistan store, 11-06). Tiny numbers, but in the small "AI speaking" niche it ranks behind only Learna in Uzbekistan, and it's growing faster than global products like Praktika and Loora, which have raised funding.

Can't tell how much money it makes. Some users pay with local payment methods inside the app, and that money doesn't go through Apple, so it never shows up on the grossing charts.

Where users come from

What's visible are four channels, each very cheap:

1. In-app contests: the XP contest mentioned above, with Pro given out for a month, worth about $95 in total.
2. Local micro-influencers with codes: we found Sarvinoz's Reel with #rek, plus a short video tagged #ielts #bandmate. How many influencers were hired and how much was paid, we don't know.
3. Bulk SEO: bandmate.uz's sitemap has 1,280 URLs, 1,270 of them blog posts, written in Uzbek (both Latin and Cyrillic scripts), Russian and English. Titles cover local long-tail queries like IELTS speaking past papers, IELTS exam fees, mapping CEFR to national certificates, and score calculators. Most were published in a cluster over the last days of September and first days of October. Almost certainly bulk-generated with an LLM (inferred).
4. Store keywords: the app name includes "Speak English"; the Russian store name was swapped to the Uzbek "Inglizcha gapiring" (speak English). The website's links to Google Play carry a landing_id, which suggests they track conversions per landing page, and they may run separate landing pages for ads (unverified).

Not findable on TikTok, nobody mentions it on X, no media coverage. Whether it buys ads on Meta or Telegram, we couldn't find out.

Why this one, and why Uzbekistan

• Global products don't speak Uzbek. Praktika, Learna, Loora and ELSA all have English-only interfaces and explanations. For someone at A2 level, a correction you can't understand is a correction you can't use.
• IELTS is a must-have. More than 50k people in Uzbekistan take IELTS each year (per local competitor BandUp's website, unverified), and speaking is the most common weak spot. Bandmate puts "start speaking from zero" and "IELTS speaking score estimate" in the same app.
• Local payments take the friction out of paying. Many Uzbeks only have Uzcard or Humo, not Visa, so foreign subscriptions are hard to pay for. Conversion only really takes off if Click or Payme works.
• Low bar to chart. 45 ratings in a week is enough to land in the top 20 of the free Education chart. One event and a few influencers is all it takes.
• The team already had assets. The developer, FERATECH, also built Salom AI, an Uzbek-language AI assistant, and a per-minute voice agent, so Uzbek voice was already covered.

Who it's competing with

The category is crowded, just in different places:

The moat is thin: if Englify adds voice conversation, or Learna adds Uzbek explanations, they catch up fast.

Another country: the neighbors are still empty

Whether this setup transfers depends on the neighbors. On the Azerbaijan App Store, Learna is #1 on the Education grossing chart, adding 164 ratings a week. The AI speaking apps behind it (YayTalk, BetterSpeak, Loora) get single digits a week, and none explain in Azerbaijani. Armenia is similar: Learna is #1 on the free chart (App Store Azerbaijan and Armenia stores, 11-06). Azerbaijan did see a local IELTS practice app, "Bandaz", appear in September, with only 1 ratings.

So: the money has already been validated by a fully English global product (Learna is #1 grossing in Azerbaijan), and the local-language layer is still empty. We didn't check Kazakhstan, Kyrgyzstan, Georgia and Mongolia one by one; they should look similar (inferred).

If it were you

One-line verdict: worth trying. Don't copy Bandmate itself, copy the recipe: a global AI speaking model that works × a small-language market big global players can't be bothered to localize × native-language explanations × local exams × local payments. The opportunity is real, and so are the weaknesses: the market is small, the ceiling is low, and nobody knows how much money Bandmate itself makes.

Who it fits: people from that country, or teams with a native speaker; people who can plug in local payments (Click / Payme or the local equivalent wallet); indie devs who can ship voice conversation with correction in a few months with one or two people.

Who it doesn't fit: people who don't speak the local language and plan to localize with translation software (the explanations are the selling point; clumsy ones kill it); teams aiming for a big market with millions of users.

How to do it:

1. Picking a country: many people take IELTS or a national English certificate, global products have no local-language support, and payments run on local cards. Azerbaijan is the most recent example we can see.
2. Build it roughly like this: voice conversation, a native-language explanation for each error, a path from zero to C2, IELTS speaking score estimates, leagues and check-ins. The free tier has to be genuinely useful: a few sessions a day, a few minutes each, is enough.
3. Price against the global products' local-currency prices, allow one-time monthly payment through local wallets, and don't force auto-renewal.

How to get users: use its three cheap plays. A monthly XP contest, with rewards for "invite a friend to finish 3 lessons" and "come back"; recruit 5 to 10 local micro-influencers, each with their own discount code so you can track who brought whom; bulk-generate local-language long-tail articles on exams (fees, past papers, score conversion) to hold search.

How soon: there's no hard deadline, but move before Learna or Praktika start localizing explanations into minor languages. Starting within six months to a year looks safe (estimate).

Signs you misread it: ① After the Nov 9 contest ended, Bandmate's ratings dropped back to one or two a day, which means its growth came from events, not from localization itself; ② Learna or Duolingo launched Uzbek or Azerbaijani lessons; ③ You've spent two or three months in the target market and paid conversion is still too low to cover server and voice costs.

Data and sources

• Ratings and ranks: App Store Uzbekistan, Azerbaijan and Armenia stores, 10-28 to 11-06; Google Play page, 11-06.
• Prices, XP contest rules, payment methods, product description: bandmate.uz homepage, /contest (published 10-09), /payments (updated 10-12), checked 11-06.
• Influencer Reel: Instagram @novasarvii_, 10-12.
• SEO blog count: bandmate.uz sitemap, 11-06.
• The IELTS candidate figure is from BandUp's website, unverified; no revenue or ad-spend figures are public.