年間250万ドル（$2.5M）のAIカスタマーサポートアプリをどう作ったか
月商 $190,000 · 創業者2人 · 従業員8人 · Amsterdam
Deepak Singla（Founder, Fini AI）· 創業 June 2022 · 記事掲載 July 18th, 2025 · usefini.com · 元の記事
創業者の個人的なネットワーク（友人、YC卒業生、Uber同窓生）を活用し、ニーズのヒアリングから初期顧客を獲得。創業者のLinkedInでのコンテンツ発信（AIエージェント構築の知見や失敗談、成功事例）が信頼を築き、リード（見込み客）獲得につながった。月商19万ドル（約2,850万円）、従業員8人は2025年7月記事掲載時点の数字
売値の出典: 公式サイトの料金ページに「Pricing」の記載があるが、具体的なプランと金額は確認できず。記事の本文（2025年時点）に初期顧客が月額500ドルで利用したと記載（2026-10-14 取得）
I’m Deep, co-founder of Fini, an Agentic AI support platform for enterprise and growth-stage B2C companies, especially in fintech, e-commerce, and healthcare. Fini automates 80% of support issues, saving 50% in costs and boosting CSAT by 10%. Before Fini, I led customer experience at Uber and saw how human-led support doesn’t scale. So we built Fini: AI agents that think, reason, and act like humans would; handling complex queries across chat, and email. What makes Fini unique is our RAGless architecture, Agentic AI Knowledge Store (knowledge built for AI agents), and Agentic execution framework, which lets our AI agents take real actions (like issuing refunds, checking KYC status, or pausing subscriptions) with 98% accuracy. Today, we help companies like Atlas, Column Tax, and TrainingPeaks resolve over 500,000+ support tickets every month My image
The idea for Fini came from firsthand frustration. At Uber, I led Product Growth and saw hundreds of agents handling repeat questions, things that should’ve been automated. We started Fini in June 2022, just as LLMs were improving, and knew this was the future of support. We didn’t get it right at first. We built a product, launched it, and no one used it. Then we scrapped everything and sat inside a fintech support team for 4 weeks. That’s when it clicked: great AI agents need a strong, AI-native source of truth. We rebuilt Fini, and our first customer, paying just $500/month, saw 70% resolution in weeks. That’s when we knew we were onto something real.
In building Fini AI, the founders focused on creating AI agents that could handle complex support queries across chat and email. They employed a RAGless architecture, Agentic AI Knowledge Store, and Agentic execution framework to achieve 98% accuracy in actions like issuing refunds or checking KYC status. The first version took several months to build, with an initial prototype that saw a 70% resolution rate within weeks for the first paying customer. The experience of building Fini AI was challenging, with the founders initially launching a product that didn't gain traction until they pivoted to focus on AI-native sources of truth based on real customer needs and feedback.
For our launch, we leaned on our network: friends, YC alums, and Uber alumni. At first, we had no product, just a question: “Can we solve this problem for you?” We did things that didn’t scale and made our first $500 after 9 months. But that one customer loved it, and from there, it snowballed to 10, mostly through referrals. Alongside, we built a strong founder brand on LinkedIn by sharing our journey and insights on AI support. What I’ve learned: building a startup is 90% about showing up every single day. Forget everything else. Just talk to customers, solve their problems, and make 10 people really love what you’ve built. That’s all you need to start something great.
Our growth has been entirely driven by two things: 1. Deep customer obsession 2. A strong founder-led content strategy We have always obsessed over making our customers really successful. That has led to organic referrals, our first few customers brought the next 10 and even today that's how we get a large portion of our customers. From there, we layered on a LinkedIn-first content strategy, where we shared lessons from building AI agents: things we were learning, real examples, mistakes, and wins. Because we sell to enterprise CX leaders, trust matters, and founder content builds trust. One tactic that worked especially well: we posted a teardown of why traditional AI chatbots fail in fintech, with examples from real support tickets (suitably anonymized). That one post brought 4 inbound leads from fintech teams who resonated with the pain and asked, “Can you do this for us?” We later turned that post into a full-blown blog and sales deck. We also run highly personalized outbound. We literally go to the help center for each of our target company, identify the gaps and send them personalized outreach on what they can improve. That relevance-first approach gives us a 7-10% reply rate. What aspiring entrepreneurs can learn: 1. Start narrow. Go deep with a small group of customers. Make them love you—they’ll become your loudest champions. 2. Build in public. Share your learnings, wins, and failures openly. It builds trust, attracts your tribe, and creates momentum. 3. Sell before you scale. Don’t wait to polish your product. Start selling as soon as you can deliver value, even if it’s manually behind the scenes. 4. Do things that don’t scale (at first). Sit with your users, write custom prompts, handle edge cases yourself. That’s how you discover what really matters.
Starting Fini has been the most humbling and transformative journey of my life. I’ve learned that the best products don’t win, relentless execution does. We made plenty of mistakes early on: we built what we thought was right instead of sitting with real users. We spent 6 months building the wrong thing. Only when we embedded with a customer and solved one specific pain well did everything change. A good decision we made? Staying close to the problem. Even today, I do weekly support calls with customers, not to sell, but to learn. That feedback loop drives our roadmap more than any investor or blog post. We also got lucky with timing. We started in 2022, just as LLMs were becoming practical. But luck only compounds if you show up every day. That’s been our superpower: showing up, shipping fast, and listening hard. I had to learn to write well, sell without selling, and let go of perfection. A helpful skill? Clarity. If you can explain your product in simple words and pitch without buzzwords, you’re 10x ahead. One thing I see founders get wrong: They chase scale before love. You don’t need 1,000 users at the start. You need 10 who’d be angry if you shut down tomorrow. My best advice: Talk to your customers more than you talk to your team. Be obsessed with their pain. And remember, you only lose if you stop showing up.
IDEA REVENUE PA Track your peptide protoc... · pepaiapp.com Track your peptide protocol in one simple app. $30K /mo Remote living resource hub · nomadlist.com 'Community platform for digital nomads worldwide' — Unblocked Games Site for ... · duckmath.org Unblocked Games Site for kids in school $15K /mo A Ai-driven job application... · aiapply.co AI-powered job application toolkit for job seekers. $15K /mo F The only fully Instagram ... · followbuddy.com The only fully Instagram compliant IG unfollower tracking tool. $17K /mo Free music for creators · freetouse.com A music library with high-quality music for UGC $85K /mo GW Ai-powered job search ass... · globalwork.ai Unlock the latest job search strategies with our AI-powered platform $350K /mo newDocumentReady(function() { // Save-bookmark hover — delegated on document (guarded once) so rows // appended by infinite scroll get the behavior too. mouseover/mouseout // bubble (mouseenter/leave don't); closest('tr') scopes to the hovered row, // and the relatedTarget check suppresses flicker when moving cell-to-cell // within the same row. if (!window.__saveBookmarkHoverInit) { window.__saveBookmarkHoverInit = true; document.addEventListener('mouseover', function(e) { var row = e.target.closest('tr'); if (!row) return; var saveBookmark = row.querySelector('.save-bookmark'); if (saveBookmark) saveBookmark.classList.remove('invisible'); }); document.addEventListener('mouseout', function(e) { var row = e.target.closest('tr'); if (!row) return; if (e.relatedTarget && row.contains(e.relatedTarget)) return; var saveBookmark = row.querySelector('.save-bookmark'); if (saveBookmark) saveBookmark.classList.add('invisible'); }); } // ---- Tag-pill hover tooltip (LLM-generated context for each Tagging) ---- if (!window.__tagTooltipInit) { window.__tagTooltipInit = true; const tip = document.createElement('div'); tip.className = 'tag-tooltip-popover'; tip.style.cssText = [ 'position:fixed','z-index:9999','max-width:340px','display:none', 'background:#0f172a','color:#fff','font-size:12px','line-height:1.4', 'padding:8px 10px','border-radius:6px','box-shadow:0 8px 24px rgba(0,0,0,.18)', 'pointer-events:none','white-space:normal' ].join(';'); document.body.appendChild(tip); const showTip = function(e) { const target = e.currentTarget; const text = target.getAttribute('data-tag-tooltip'); if (!text) return; tip.textContent = text; tip.style.display = 'block'; const r = target.getBoundingClientRect(); const tr = tip.getBoundingClientRect(); let top = r.bottom + 8; if (top + tr.height > window.innerHeight - 8) top = r.top - 8 - tr.height; let left = r.left; if (left + tr.width > window.innerWidth - 8) left = window.innerWidth - tr.width - 8; if (left < 8) left = 8; tip.style.top = Math.round(top) + 'px'; tip.style.left = Math.round(left) + 'px'; }; const hideTip = function() { tip.style.display = 'none'; }; document.addEventListener('mouseover', function(e) { const el = e.target.closest('[data-tag-tooltip]'); if (el) showTip({ currentTarget: el }); }); document.addEventListener('mouseout', function(e) { const el = e.target.closest('[data-tag-tooltip]'); if (el) hideTip(); }); } });
Who is the owner of Fini AI? When did Deepak Singla start Fini AI? How much money has Deepak Singla made from Fini AI? {"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"Who is the owner of Fini AI?","acceptedAnswer":{"@type":"Answer","text":"Deepak Singla is the founder of Fini AI."}},{"@type":"Question","name":"When did Deepak Singla start Fini AI?","acceptedAnswer":{"@type":"Answer","text":"2022"}},{"@type":"Question","name":"How much money has Deepak Singla made from Fini AI?","acceptedAnswer":{"@type":"Answer","text":"Deepak Singla started the business in 2022, and currently makes an average of $2.28M/year."}}]}