No Sales Team, $20 Million in Revenue: How Did AI Employee Viktor Win Over 30,000 Companies?

marsbitDipublikasikan tanggal 2026-06-19Terakhir diperbarui pada 2026-06-19

Abstrak

The AI employee Viktor, developed by a team with DeepMind background, has achieved $20 million in annual revenue without a traditional sales team, serving over 30,000 companies. Its core innovation lies in positioning itself as a "Tier 3 AI Coworker" capable of "end-to-end execution and delivery of results," moving beyond the "draft and wait for human completion" model of typical AI assistants. Users can simply mention Viktor in Slack or Microsoft Teams using natural language commands, and it autonomously performs tasks like pulling sales data from a CRM, generating reports, or even cross-tool operations like creating board meeting PPTs by aggregating data from six different sources. Key to its growth is a pure Product-Led Growth (PLG) model, eliminating complex implementation cycles and per-seat licensing. Instead, it charges based on task credits or consumption, lowering the trial barrier with a $100 free credit offer and no credit card required. This enabled viral, bottom-up adoption within organizations. Viktor's interaction paradigm removes the barrier of prompt engineering, allowing non-technical employees to delegate complex workflows seamlessly. It also features proactive, automated task execution (e.g., overnight bookkeeping, scheduled reports) based on triggers, effectively embedding AI as an automated "process layer" within business operations. However, its expansion into Microsoft Teams—a platform with 320 million users—highlights challenges. Large enterprises ...

The expansion of traditional enterprise software often involves large sales teams and lengthy implementation cycles. From initial contact to final deployment, it typically takes months, involving multiple demos, compliance reviews, and custom development. But AI employee Viktor defies this common sense.

Before delving into the commercial data, it's essential to clarify what Viktor actually is. This product was founded by a research and development team with a DeepMind background. Its core philosophy is to create a "Tier 3 AI Coworker," rather than a simple Copilot. In the view of the Viktor team, most current AI tools remain at the stage of "draft and wait for human completion," while Viktor's goal is "end-to-end execution and delivery of results."

In simpler terms, Viktor is like an indefatigable digital employee. You don't need to teach it how to use various software, nor write complex instructions. You just need to @mention it in a Slack or Teams chat window as you would a colleague, telling it "help me check last week's sales data for the East China region and generate a brief report with charts." It will then pull data from the CRM system, generate charts in a spreadsheet tool, and send the final report back to the conversation window. Besides passive responses, it can also proactively perform tasks triggered by specific times or events, such as automatically reconciling accounts late at night, or gathering data across 6 different tools to create a board meeting PPT.

According to its official disclosure, it is precisely such a product that, on the Slack platform without a sales team or implementation projects, achieved $20 million in annualized revenue, serving over 30,000 companies. Recently, Viktor officially integrated with Microsoft Teams, opening up a free trial to an ecosystem of 320 million users. As AI employees abandon prompt engineering and move towards "zero-barrier @mentioning," has the tipping point for enterprise automated office work arrived? This is not just a question of product feature updates, but concerns the fundamental restructuring of the business model for enterprise-level AI applications.

$20 Million Without a Sales Team: The Victory of the PLG Model in Enterprise AI

The enterprise SaaS industry has long adhered to a "sales-driven" approach. To secure large clients, companies need to build extensive sales teams, configure customer success managers, and undergo lengthy POC (Proof of Concept) and implementation cycles. This model has extremely high customer acquisition costs and heavily relies on relationship maintenance. Viktor's performance on Slack, however, showcases a completely different path.

Official disclosed data shows that Viktor, without forming a sales team, without implementation projects, and without per-seat licensing contracts, achieved $20 million in annualized revenue and served 30,000 companies. This pure PLG (Product-Led Growth) model, while having precedents in the traditional SaaS era, is extremely rare in complex enterprise AI applications. AI products typically require substantial context configuration and scenario debugging, making it difficult to achieve out-of-the-box usability. The core reason Viktor can achieve self-propagation lies in its reduction of the configuration barrier to a minimum.

The traditional SaaS per-seat billing model often makes enterprises worry about "idle waste" during procurement. Buying 100 accounts might result in only 20 people using them frequently, leaving the remaining 80 accounts as sunk costs. Viktor tends to charge based on credit or task consumption, a model more aligned with the actual logic of AI task execution. Companies no longer pay for the "potential number of employees using AI," but for the "actual workload completed by AI."

This billing method lowers the trial-and-error cost for enterprise procurement, allowing department-level managers or even frontline employees to start trying directly with a credit card or free credits, bypassing lengthy IT procurement approvals. The validation of this business model confirms a judgment: the core barrier for enterprise AI products is not the coverage capability of sales channels, but whether the product itself can prove its value within an extremely short trial period.

Viktor's strategy of offering $100 in free credits without requiring a credit card is precisely aimed at maximizing the shortening of this "value verification" cycle. When employees find that simply @mentioning Viktor can complete tasks that originally took hours, such as account reconciliation, natural product self-propagation occurs. According to public reports, Viktor recently completed a $75 million Series A funding round led by DN Capital, which also reflects the capital market's recognition of its PLG model from the side. However, it should be noted that the specific calculation method for the $20 million ARR has not been publicly detailed by the company. Whether it's based on credit consumption, action billing, or a hybrid model conversion, is unknown to the public. This non-transparent billing approach may help lower the trial barrier in the early stages, but could become an obstacle for ROI calculation during large-scale enterprise procurement.

Leveling the Prompt Barrier: From "Draft and Wait" to "End-to-End Delivery"

The key reason Viktor can achieve zero-configuration self-propagation lies in the dimensionality reduction of its interaction paradigm. The effectiveness of traditional AI tools heavily depends on the user's prompt-writing ability. An OmniTools article titled "After Three Years of Observation, I've Categorized All AI Users into 10 Levels" previously analyzed this phenomenon in detail: from structured prompts to encapsulated Agent skills, AI users are divided into multiple tiers, with prompt engineering becoming an invisible barrier.

In real enterprise scenarios, this barrier is particularly critical. Finance staff, HR specialists, and operations managers don't have the time, nor the obligation, to learn how to engage in complex "prompt games" with AI. If the effectiveness of AI depends on employees' prompt-writing skills, then AI will forever remain an efficiency tool for a few geeks, unable to become a universal infrastructure for enterprises.

Viktor's positioning is "Tier 3 AI Coworker," not simply a Copilot. The logic of a native Copilot is "draft and wait for human completion." It excels at summarizing documents and drafting emails, but the final step still requires human intervention. For example, you ask a Copilot to write a customer follow-up email; after it writes, you need to copy it to the email client, manually enter the recipient, and send it. Viktor's logic is "end-to-end execution and delivery of results." Users only need to describe the goal in natural language; during runtime, the Agent autonomously decides the execution steps, calls necessary tools to complete the loop. Following up with a customer, Viktor can directly connect to the email system, automatically fill in customer information and send the email, even schedule the next reminder based on the customer's reply.

This mechanism directly levels the tier barrier brought by prompt engineering. AI effectiveness no longer depends on employees' prompt-writing techniques, but on the clarity of business objectives. This interaction style pushes AI from an "assistance tool" to an "executor," allowing non-technical personnel to enjoy AI benefits with zero friction.

However, this does not mean Viktor is entirely free from the risk of misinterpretation. When users describe goals with vague natural language, the AI's runtime autonomous decision-making mechanism might generate execution paths that don't align with user expectations. For example, a user saying "clean up the sales pipeline" might lead Viktor to automatically mark some long-unfollowed leads as "lost," which in the enterprise sales process might require more complex approvals. Zero-barrier lowers the usage threshold but also demands higher accuracy in business goal descriptions.

Late-Night Auto-Reconciliation & Cross-Tool PPT Generation: How AI Sinks into the "Process Layer"

If @mentioning is a passive response to human instructions, then Viktor's auto-trigger mechanism demonstrates the proactivity of an AI employee, which is also its core feature distinguishing it from traditional chatbots. According to Viktor's official disclosure, its product supports auto-trigger scenarios without manual @mentioning, such as late-night closing, reconciliation and error marking, screening applicants and scheduling calls, generating board meeting PPTs across 6 isolated tools, and running routine operational tasks at 5 AM.

These scenarios reveal an important trend: AI is sinking from the "conversation layer" to become the enterprise's "process layer." An OmniTools article titled "With Daily Active Users Surging to 3-4 Times the Industry's Second Place, How Did Tencent's WorkBuddy Tear Open a Crack in Office Agents?" previously explored how office Agents serve non-developer groups. Whether it's Viktor or WorkBuddy, their core logic is to encapsulate fixed processes that originally required crossing multiple systems and manual steps into atomic tasks executable by AI.

Take financial reconciliation as an example. In the traditional process, finance personnel need to export payment data from Stripe, export accounting data from Xero, perform VLOOKUP comparisons in Excel, identify discrepancies, and manually mark them. This process is tedious and time-consuming, typically taking finance staff 2 hours. Viktor, through managed authentication, connects to 3200+ tools. When the system time reaches the set late-night node, Viktor automatically logs into Stripe and Xero, pulls the day's data, executes comparison logic, and sends a report with marked errors to the finance channel. The entire process requires no human intervention and, according to the official claim, takes only 6 minutes.

Another example is cross-tool board meeting PPT generation. An executive needs a brief containing sales data, product progress, and market feedback. Traditionally, an assistant would need to open the CRM, project management tool, and customer service system separately, copy data, create charts, and finally paste them into a PPT. Viktor can automatically execute this series of actions at 5 AM, directly outputting a complete PPT file in the conversation window.

Supporting this auto-trigger capability is Viktor's organizational-level memory and context awareness mechanism. According to third-party evaluations, Viktor possesses persistent memory. If a finance person corrects a mistake Viktor made regarding UTM format or reconciliation rules once, Viktor will permanently remember it and automatically apply that rule in all subsequent related tasks. It can even read channel history and proactively explain past decision reasons.

This mechanism makes Viktor not just a tool for executing tasks, but a "process layer" that accumulates enterprise best practices and business rules. It reduces the friction costs of manual reminders, handovers, and "emotional management." When a senior employee leaves and a new one joins, the rules and processes in Viktor's memory remain, ensuring business continuity.

From Slack to Teams: How the PLG Model Navigates the Deep Waters of Enterprise Compliance

Viktor's integration with Microsoft Teams is a critical step in its commercialization journey. While Slack is known for flexibility and developer-friendliness, serving as a "testing ground" for lean teams and frontline companies, Microsoft Teams possesses a more complete departmental structure, approval chains, and org charts, representing the "real large organizations." Official data shows Teams has 320 million users. Viktor's entry into Teams marks AI employees transitioning from "geek toys" to formally entering the "core procurement vision of enterprises."

However, moving from Slack to Teams is not a simple platform migration; it's the beginning of the PLG model entering the deep waters of compliance. In Slack, users can install and authorize an App within seconds, a friction so low it formed the basis for Viktor's viral spread. But in Teams, this seconds-long installation is replaced by lengthy IT admin approval queues, security reviews (like SOC 2 compliance requirements), and application governance policies.

IT departments in large enterprises maintain high vigilance towards any third-party application with data read/write permissions. To achieve end-to-end task execution, Viktor must obtain read/write permissions for CRM, financial systems, and even code repositories. Such high permissions mean it cannot bypass the enterprise procurement cycle. The "bottom-up" PLG propagation path validated by Viktor on Slack may be blocked by the "top-down" control of IT departments in Teams.

To address this challenge, Viktor also opened a free trial with $100 in free credits on the Teams side, without requiring a credit card. This is a typical "wedge" strategy, attempting to let frontline employees experience the product's value and generate internal demand before the IT department becomes aware, thereby pushing the IT department to conduct compliance approvals. However, the effectiveness of this strategy within the Teams ecosystem remains to be seen. After all, enterprise procurement decisions depend not only on product experience but also on compliance risks and data asset security.

The Cost of Full Automation: Black-Box Risks and the Trust Game

The vision of "zero-barrier" and "full automation" painted by Viktor undoubtedly hits the pain points of enterprise operational efficiency. However, in actual deployment, this model faces non-negligible trust crises and black-box risks.

To achieve breadth of coverage and end-to-end delivery, Viktor sacrifices granular control over each step of execution. Traditional workflow automation tools (like n8n or Zapier), although cumbersome to configure, make each step's data flow and logical branches visible, allowing operations personnel to clearly locate errors. Viktor's runtime autonomous decision-making mechanism, however, makes the execution process somewhat of a "black box." When AI possesses "read/write permissions" for CRM or financial systems, a single model hallucination or misinterpretation of a natural language instruction could lead to incorrect data being written into production systems, causing data pollution or even business interruption.

What enterprise procurement decision-makers often care most about is the risk of "misoperation." If an AI employee can automatically update customer information in HubSpot or create invoices in Xero without strict Per-user permissions and Audit logs, a single erroneous execution might require significant manpower for data rollback and recovery. For example, if Viktor, while automatically cleaning the sales pipeline, mistakenly marks a batch of high-value leads as "lost," the sales team might lose important customer leads, and such errors might only be discovered days later.

To mitigate these risks, enterprises in practice often have to enable "approval-first default settings." This means Viktor must wait for human confirmation before executing critical write operations. This compromise, while reducing risk, also breaks the vision of "fully automated, unattended operation," reintroducing manual intervention steps. Finding the balance between "efficiency gain" and "misoperation disaster" is a question all AI employee products must answer.

Viktor's auto-trigger mechanism also introduces new management challenges. When AI can automatically execute tasks based on events or time, enterprises need to establish a new monitoring system to ensure AI behavior consistently complies with business rules and compliance requirements. Strict permission management, comprehensive audit logs, and explainable decision paths are prerequisites for the large-scale deployment of AI employees. If these issues are not properly resolved, AI employees may forever remain confined to marginal scenarios at the departmental level, unable to truly enter the core business flows of enterprises.

From Slack to Teams, Viktor has demonstrated the appeal of zero-barrier interaction in the enterprise market and also exposed the compliance resistance of the PLG model within large organizations. For AI employees to truly become enterprise infrastructure, they require not only smarter models and lower interaction barriers but also a governance framework capable of winning enterprise trust. Only when the balance between efficiency and security gradually stabilizes will the true tipping point for enterprise automated office work arrive.

Kripto yang Sedang Tren

Pertanyaan Terkait

QHow did Viktor achieve $20 million in annualized revenue and serve 30,000 companies without a sales team?

AViktor achieved this through a pure Product-Led Growth (PLG) model. By lowering the configuration barrier to near zero and offering a $100 free credit trial without requiring a credit card, it enabled individual employees or small teams to start using the product quickly. The pay-per-credit/task consumption model, rather than per-seat licensing, reduced the financial risk for enterprises and bypassed lengthy IT procurement cycles. Its core value proposition of 'end-to-end execution' allowed users to see immediate productivity gains, driving organic, bottom-up adoption within organizations.

QWhat is the core difference between Viktor's 'Tier 3 AI Coworker' and traditional AI Copilots?

AThe core difference lies in the level of autonomy and task completion. Traditional AI Copilots operate on a 'draft and wait for human completion' logic, where they assist with tasks like summarizing or drafting but require human intervention for the final steps (e.g., sending an email). Viktor, as a 'Tier 3 AI Coworker', follows an 'end-to-end execution and delivery of results' logic. Once given a natural language goal, it autonomously decides the execution steps, interacts with necessary software tools, and delivers the final result, requiring no further human action to complete the task.

QWhat key feature allows Viktor to function as a proactive 'process layer' within an organization?

AViktor's automatic trigger mechanism allows it to function proactively as a 'process layer'. Instead of only responding to direct @mentions, it can be configured to automatically execute tasks based on specific times (e.g., nightly bookkeeping at 2 AM) or events (e.g., a new applicant submission). This enables it to autonomously run complex, multi-step workflows that span multiple, previously isolated tools—like reconciling data between Stripe and Xero or generating a board presentation by pulling data from 6 different systems—without any human initiation.

QWhat major challenges does Viktor's PLG model face when expanding from Slack to Microsoft Teams?

AThe major challenges involve navigating the enterprise compliance and security 'deep water'. In large organizations using Microsoft Teams, the installation of apps like Viktor is subject to lengthy IT administrator approvals, security reviews (e.g., SOC 2 compliance), and application governance policies. Teams' structure better represents formal corporate hierarchies and procurement processes. Viktor's high-level permissions to read/write core business systems (CRM, finance) cannot easily bypass these centralized controls. The 'bottom-up' PLG adoption path proven in Slack may be blocked by 'top-down' IT governance in Teams, requiring a shift in strategy to gain formal enterprise approval.

QWhat are the primary 'black box' risks and trust challenges associated with Viktor's fully autonomous execution model?

AThe primary risks are related to the lack of transparency and control. Viktor's runtime autonomous decision-making can be a 'black box', making it difficult to audit or trace the logic behind specific actions. With high-level permissions to critical systems, a model hallucination or misinterpretation of a user's natural language instruction could lead to erroneous data being written to production systems—like incorrectly marking high-value sales leads as 'lost'—potentially causing data pollution or business disruption. To mitigate this, companies may need to implement 'approval-first' defaults for critical actions, re-introducing human oversight and partly compromising the promise of full automation.

Bacaan Terkait

CPU Kembali ke Meja Permainan, Drama "Peningkatan Posisi" Senilai 170 Miliar Dolar Dimulai

CPU Kembali ke Meja, Drama “Naik Takhta” Senilai 1700 Miliar Dolar Dimulai Pada 1 Juni, Nvidia meluncurkan Vera CPU, menandai pertama kalinya perusahaan tersebut merilis lini produk CPU independen. CEO Nvidia Jensen Huang menyatakan, di era AI Agent, CPU telah menjadi hambatan kinerja kunci di pusat data. Pasar CPU server diproyeksikan tumbuh pesat, dari sekitar 300 miliar dolar AS pada 2025 menjadi sekitar 1700 miliar dolar AS pada 2030, didorong oleh permintaan AI. Dalam beban kerja AI Agent, CPU menangani 70% atau lebih dari total pekerjaan, karena tugas-tugas seperti pemanggilan alat eksternal dan manajemen konteks sangat intensif. Rasio GPU terhadap CPU dalam penyebaran AI juga menyusut, dari 8:1 menjadi sekitar 4:1, bahkan mendekati 1:1 dalam skenario Agent tertentu. Perubahan ini telah menyebabkan peningkatan harga CPU server sebesar 10-15%, kenaikan pertama dalam lebih dari satu dekade. Nvidia, AMD, dan Intel semua berinvestasi besar-besaran di CPU. Nvidia memproyeksikan pendapatan terkait CPU mendekati 200 miliar dolar AS pada 2026. Pertumbuhan pasar ini juga membuka peluang bagi rantai pasokan CPU China, seperti Haiguang Information dan Huawei, yang diuntungkan oleh permintaan industri dan kebijakan substitusi impor. Intinya, sementara GPU tetap penting, kemampuan sinergi antara CPU dan GPU akan menjadi pembeda utama dalam penerapan AI skala besar berikutnya.

marsbit7j yang lalu

CPU Kembali ke Meja Permainan, Drama "Peningkatan Posisi" Senilai 170 Miliar Dolar Dimulai

marsbit7j yang lalu

Kantor Intelijen TechFlow: Direktur AI AMD Secara Terbuka Mengkritik Claude Code "Menjadi Lebih Bodoh dan Lebih Malas", Trump Klaim Gencatan Senjata Lengkap di Selat Hormuz Tetapi Masih Ada 80 Ranjau Laut yang Harus Dibersihkan

**Wired** melaporkan SK Telecom, mitra strategis Anthropic, sedang ditinjau kontrol ekspor AS terkait potensi transfer teknologi model Mythos. Seorang pengguna Reddit mengeluh Gemini memberikan saran menyesatkan dalam skenario penipuan, memicu diskusi tentang batasan keamanan AI. **Z.AI** merilis model GLM-5.2 China yang diklaim setara Claude Opus tanpa menggunakan chip Nvidia. **0G Labs** mencapai tonggak 100 miliar token untuk inferensi AI terdesentralisasi. Kemampuan visual **DeepSeek** memicu perbandingan dengan GPT-4V di platform Zhihu. Di sektor **Web3**, Bithumb menambahkan pasangan perdagangan RE, sementara Upbit menghapus pasangan KERNEL. Peneliti **MIT** membangun sistem operasi sendiri untuk memahami chip. AS klaim peralatan litografi ASML canggih mungkin telah masuk ke China, namun ASML membantah. **Amazon** dilaporkan bernegosiasi menjual chip AI Trainium/Inferentia mereka. **Apple** dikabarkan akan menggunakan proses N2P eksklusif untuk chip A21 Pro. **GitHub** ditemukan memiliki 10.000 repositori yang mendistribusikan perangkat lunak berbahaya. Apple perbarui firmware untuk tutup celah keamanan di Beats Studio Buds. Karyawan **Amazon** diselidiki internal karena mengkritik ekspansi pusat data AI. **Microsoft** dan **Amazon Web Services** menghadapi pengawasan antitrust ketat dari UE. Saham semikonduktor meroket di pasar saham AS, dengan **Intel** melonjak 10.6%, sementara **SpaceX** turun 3.5%. Meskipun ada pengumuman gencatan senjata di Selat Hormuz, asosiasi tanker memperingatkan masih ada sekitar 80 ranjau laut di jalur utama, menghalangi 80 juta barel minyak yang siap dikirim. Iran menunda perjalanan diplomatik ke Swiss, meragukan prospek perdamaian. **Trump** menyebut kesepakatan dengan Iran sebagai "penyerahan tanpa syarat" dan menegaskan kekuasaan presiden yang tidak terbatas. **Valve** mengalami penumpukan pesanan parah untuk Steam Controller, dengan sebagian pengiriman tertunda hingga 2027. **Inti:** Ketegangan geopolitik di Timur Tengah tetap rapuh dengan ranjau dan ketidakpastian diplomasi, sementara perang teknologi dan rekonfigurasi rantai pasokan semikonduktor terus berlanjut dengan langkah-langkah seperti model AI buatan China dan penjualan chip mandiri Amazon.

marsbit7j yang lalu

Kantor Intelijen TechFlow: Direktur AI AMD Secara Terbuka Mengkritik Claude Code "Menjadi Lebih Bodoh dan Lebih Malas", Trump Klaim Gencatan Senjata Lengkap di Selat Hormuz Tetapi Masih Ada 80 Ranjau Laut yang Harus Dibersihkan

marsbit7j yang lalu

Trading

Spot
Futures

Artikel Populer

Cara Membeli WIN

Selamat datang di HTX.com! Kami telah membuat pembelian WINkLink (WIN) menjadi mudah dan nyaman. Ikuti panduan langkah demi langkah kami untuk memulai perjalanan kripto Anda.Langkah 1: Buat Akun HTX AndaGunakan alamat email atau nomor ponsel Anda untuk mendaftar akun gratis di HTX. Rasakan perjalanan pendaftaran yang mudah dan buka semua fitur.Dapatkan Akun SayaLangkah 2: Buka Beli Kripto, lalu Pilih Metode Pembayaran AndaKartu Kredit/Debit: Gunakan Visa atau Mastercard Anda untuk membeli WINkLink (WIN) secara instan.Saldo: Gunakan dana dari saldo akun HTX Anda untuk melakukan trading dengan lancar.Pihak Ketiga: Kami telah menambahkan metode pembayaran populer seperti Google Pay dan Apple Pay untuk meningkatkan kenyamanan.P2P: Lakukan trading langsung dengan pengguna lain di HTX.Over-the-Counter (OTC): Kami menawarkan layanan yang dibuat khusus dan kurs yang kompetitif bagi para trader.Langkah 3: Simpan WINkLink (WIN) AndaSetelah melakukan pembelian, simpan WINkLink (WIN) di akun HTX Anda. Selain itu, Anda dapat mengirimkannya ke tempat lain melalui transfer blockchain atau menggunakannya untuk memperdagangkan mata uang kripto lainnya.Langkah 4: Lakukan trading WINkLink (WIN)Lakukan trading WINkLink (WIN) dengan mudah di pasar spot HTX. Cukup akses akun Anda, pilih pasangan perdagangan, jalankan trading, lalu pantau secara real-time. Kami menawarkan pengalaman yang ramah pengguna baik untuk pemula maupun trader berpengalaman.

431 Total TayanganDipublikasikan pada 2024.12.11Diperbarui pada 2025.03.21

Cara Membeli WIN

Diskusi

Selamat datang di Komunitas HTX. Di sini, Anda bisa terus mendapatkan informasi terbaru tentang perkembangan platform terkini dan mendapatkan akses ke wawasan pasar profesional. Pendapat pengguna mengenai harga WIN (WIN) disajikan di bawah ini.

活动图片