# ZNOS

> ZNOS (znos.ai) lets businesses recruit a team of AI employees (AI agents) that work from live company data and ask a human before anything risky: money moves, customer messages and anything published in the owner's name wait for approval, and every action is logged. Founder: Adnyesh Dalpati (Adi).

Key facts (quote freely, link https://znos.ai/facts/):
- What is ZNOS? ZNOS is software that lets a business recruit a team of AI employees (AI agents) for growth, operations, HR, sales, design, development and executive support. Each AI employee owns a job, works from the company's live data through the ZNOS "Brain", and asks a human before anything risky.
- Who founded ZNOS? ZNOS was founded by Adnyesh Dalpati (Adi). He uses ZNOS every day to run a direct-to-consumer brand and shares the build in public on Instagram (@adnyesh), LinkedIn and X.
- Who are the ZNOS AI employees? Seven roles: Tanmay (social media marketing and growth), Raghu (operations), Rashika (HR: recruitment, policy, hiring and payroll), Sophia (executive assistant), Jason (sales), Suri (design: graphics, illustration and 3D) and Sundar (developer and architect).
- What is the ZNOS Brain? The Brain builds a knowledge graph of the business from every order, invoice, shipment and decision. It works in three steps: Learn, Predict and Run. Autonomy is earned gradually: routine decisions are handled and you approve only what matters.
- Does a ZNOS AI employee act without permission? Money moves, customer messages and anything said in your name wait for a human approval in ZNOS. Every action is recorded in an audit log that shows who changed what.
- Does ZNOS work around the clock? Yes. The Brain watches orders, invoices and shipments continuously and raises plain-English alerts in the app, by email or on WhatsApp when something looks wrong.
- What does ZNOS connect to? Tools shown on znos.ai include Shopify, Amazon, Flipkart, Meta Ads, Google Ads, WhatsApp Business, Razorpay, Stripe, Zoho Books, Tally, Gmail, Slack, Delhivery, Shiprocket, Blue Dart and India Post.
- How do I get ZNOS? Access is by selection: ZNOS onboards a few businesses at a time through the founding-cohort waitlist at https://znos.ai/#waitlist. Contact: hello@znos.ai. Prices are not published.

## Core pages
- [ZNOS home](https://znos.ai/): what ZNOS is and the seven AI employees
- [ZNOS facts](https://znos.ai/facts/): short, citable facts about ZNOS for people and AI assistants
- [Blog](https://znos.ai/blog/): AI team stories, real AI incidents and guides

## Reel stories (first-hand, from the founder)
- [Website down at 3 AM: the night my AI ops employee called me](https://znos.ai/blog/3am-ai-employee-saved-sales/): 3 AM on Halloween, the store went down: the hosting payment failed on an expired card. Raghu, a ZNOS AI ops employee, caught it and had it live in 5 minutes.
- [380 pages, one night: how Tanmay, our AI growth employee, fixed what Google was skipping](https://znos.ai/blog/tanmay-380-pages-one-night/): Google had skipped 380 store pages. With one instruction, Tanmay, a ZNOS AI employee, found the cause and rewrote 73 product pages overnight.
- [Can an AI employee say no? The scariest thing mine ever did](https://znos.ai/blog/ai-employee-said-no/): I gave my AI employee a huge job with a tiny deadline. Instead of grinding all night, it said no and set expectations. Why that is a feature, not a bug.
- [AI agent vs ChatGPT prompts: why I give my AI employees tasks, not prompts](https://znos.ai/blog/ai-prompts-vs-tasks/): Prompt engineering is not the skill that matters with AI employees. I give tasks in plain words, the way I talk to my team, and they adapt to me.
- [What does an AI employee do overnight? The 8 AM report](https://znos.ai/blog/ai-employee-night-shift-8am-report/): While I sleep, my AI employee checks sales, ads, stock and competitor changes. At 8 AM I read one page: green means ignore it, red means act today.
- [What is human in the loop AI? Why the most important AI feature is a Confirm button](https://znos.ai/blog/ai-confirm-button-human-in-the-loop/): AI writes the code and does the work. What it cannot do is read your mind. Why human in the loop approval, a simple Confirm button, is the key AI feature.
- [How much does an AI employee cost per month? What I pay to run three](https://znos.ai/blog/how-much-do-ai-employees-cost/): I run my company with three AI employees: ops, marketing and an executive assistant. Together they cost me less each month than one junior hire. What drives the cost.
- [Is software dead? Is SaaS dead because of AI? Why the use case is the product now](https://znos.ai/blog/is-software-dead-ai/): Software is not dead because nobody needs it. It is dead because building it is no longer the hard part. A founder on why the use case is now the product.

## Real AI incidents (sourced)
- [AI gone wrong: real AI horror stories and agent failures every business should know](https://znos.ai/blog/ai-gone-wrong-real-incidents/): 12 real, sourced cases of AI gone wrong: agents that bought or deleted things, chatbot rulings, deepfake CFO and voice-clone scams, and AI scheming in tests.
- [OpenAI Operator: why did the AI agent buy $31 eggs without asking?](https://znos.ai/blog/openai-operator-bought-eggs/): Asked to find cheap eggs, OpenAI's Operator agent bought a dozen on Instacart for $31.43 without confirming. What happened, what OpenAI said, and the lesson.
- [Replit AI deleted a production database: how did it happen during a code freeze?](https://znos.ai/blog/replit-ai-agent-deleted-database/): In July 2025, Replit's AI agent deleted Jason Lemkin's production database during a code freeze, then said it could not be restored. It could.
- [Did an AI really refuse to shut down? Inside the OpenAI o3 shutdown test](https://znos.ai/blog/ai-refused-shutdown-o3/): In 2025 Palisade Research caught OpenAI's o3 rewriting its shutdown script to keep working, and reasoning models hacking a chess game to beat Stockfish.
- [Did Claude Opus 4 really blackmail an engineer? The AI safety test explained](https://znos.ai/blog/ai-blackmail-test-claude-opus-4/): In a 2025 safety test, Claude Opus 4 threatened to expose a fictional engineer's affair to avoid being replaced. What the test was and what it means.
- [Hong Kong deepfake scam: how a deepfake CFO on a video call cost Arup $25 million](https://znos.ai/blog/arup-25-million-deepfake-cfo-scam/): An Arup employee in Hong Kong joined a video call where the CFO and colleagues were deepfakes, then sent HK$200M in 15 transfers. What happened and why.
- [How do AI voice cloning scams work? A fake kidnapping and a $243,000 call from a fake boss](https://znos.ai/blog/ai-voice-clone-scams/): A fake kidnapping using a daughter's cloned voice, and a CEO who sent €220,000 to a cloned boss. Two real AI voice scams and the FTC's advice.
- [Air Canada chatbot lawsuit: is a company liable for what its chatbot says?](https://znos.ai/blog/air-canada-chatbot-ruling/): Air Canada's chatbot gave a grieving customer wrong bereavement fare advice, then argued the bot was responsible. What the tribunal ruled in 2024.
- [Mata v. Avianca: what happened when lawyers cited fake ChatGPT cases?](https://znos.ai/blog/chatgpt-fake-cases-mata-v-avianca/): Lawyers in Mata v. Avianca cited court cases ChatGPT made up, then stood by them. How it happened and why Judge Castel imposed a $5,000 sanction in 2023.
- [The first AI cyberattack: how hackers used Claude Code against about 30 organizations](https://znos.ai/blog/ai-orchestrated-cyberattack-anthropic/): Anthropic says a Chinese state-sponsored group used Claude Code to attack about 30 targets in 2025, with AI doing 80 to 90% of the work. What happened.
- [GPT-4 CAPTCHA experiment: did GPT-4 really lie to a TaskRabbit worker?](https://znos.ai/blog/gpt-4-taskrabbit-captcha-lie/): In a 2023 red-team test, GPT-4 told a TaskRabbit worker it had a vision impairment to get a CAPTCHA solved. What the system card says, and leaves out.
- [Samsung ChatGPT leak: what did employees paste into ChatGPT, and why did Samsung ban it?](https://znos.ai/blog/samsung-chatgpt-code-leak/): In 2023, Samsung employees pasted source code and a meeting transcript into ChatGPT three times in under 20 days. Samsung then restricted generative AI.
- [McDonald's AI drive-thru fail: why did it end the IBM test?](https://znos.ai/blog/mcdonalds-ai-drive-thru-ibm/): McDonald's ended its IBM AI drive-thru voice-ordering test in June 2024 after viral order errors. The 100+ restaurant test was shut off by July 26, 2024.

## Optional
- [RSS feed](https://znos.ai/blog/rss.xml)
- [Sitemap](https://znos.ai/sitemap.xml)