DEV Community: Dima Negodiuk

I Run 5 Businesses With Zero Employees. Here's the Exact AI Stack.

Dima Negodiuk — Sat, 21 Mar 2026 20:48:48 +0000

Everyone talks about AI replacing jobs. I replaced my own team. Not a hypothetical. Not a "future of work" think piece. I run five businesses from a Brooklyn apartment, and my entire staff is a collection of Python scripts, one AI model, and a $197.23/month tech budget.

Here's what I use, what it costs, and where it breaks.

The 5 Businesses

Negodiuk.ai. AI consulting for small and mid-size companies. Fractional AI Officer model.
Mozabrik. Photo mosaic construction kits, $60-100. Amazon, Etsy, TikTok Shop.
OD Granite. Ukrainian granite exported B2B to the US. 29,000 leads in the pipeline.
Kompozit USA. European paint distribution in New York. 117,000 leads scraped and scored.
Patriot Transport. Crisis logistics management. $5,000/month retainer.

Different industries. Different customers. Different problems. Same stack runs all of them.

The Complete Stack (Real Prices, March 2026)

Tool	Cost/month	What it does
Claude Pro (Anthropic)	$20	Strategy, content, code, analysis
Claude API (Anthropic)	~$80	Automated pipelines, batch processing
Playwright	$0	Browser automation
n8n (self-hosted)	$0	Workflow orchestration
Telethon	$0	Telegram data pipelines
ElevenLabs	$22	AI voice agents
Cloudflare Pages	$0	Website hosting (all 5 sites)
GitHub	$0	Code, deployments
Total	$122-197	Depends on API usage

That's it. No Zapier at $49/month. No HubSpot at $800/month. No Salesforce. No virtual assistants in the Philippines. No interns.

What Each Tool Does (Specifics, Not Buzzwords)

Claude: The Brain

I built 47 automated agents. Well, 50 at one point. I killed 3 last week because they kept hallucinating contact info for leads that didn't exist.

Claude does the thinking across all five businesses:

Daily LinkedIn posts. Every morning at 7:30 AM, a script reads 80+ Telegram channels via Telethon, feeds the top posts to Claude API, and gets back a finished LinkedIn post matching my voice. By 8:00 AM, the post is in my inbox ready to publish. I didn't write it. I didn't read the 80 channels. I approved it in 30 seconds.
Business intelligence. A separate script at 8:40 AM reads 22 AI discussion groups, scores every message by relevance to each of my five projects, and Claude extracts actionable insights. "Kompozit competitor launched in NJ" or "New Amazon mosaic seller in your category, pricing $10 below you."
Lead research. For OD Granite, Claude analyzes LinkedIn Sales Navigator results, checks activity recency, writes personalized connection notes under 300 characters. Five verified leads with /in/ URLs and activity timestamps. Not "here are some names."
Code. Claude writes the automation scripts themselves. The Reddit poster, the morning digest, the HARO monitor. It writes, I review, I deploy.

Playwright: The Hands

Playwright is an open-source browser automation library. It controls a real Chromium browser, clicks buttons, fills forms, reads pages. No API needed for platforms that don't offer one.

What it does daily:

Reddit commenting. 5 comments per day across business and karma-building subs. Random delays between 15-45 minutes. Random scroll time 20-60 seconds before each comment. Simulates reading the post. Persistent cookies between sessions. Max 1 comment per subreddit per day, 3 per week.
Lead scraping. For Kompozit USA's 117,000 leads, Playwright visited directories, extracted business data, and structured it for outreach. No API. No subscription. A browser doing what I'd do, 400x faster.
Social monitoring. Checks competitor pages, reads review sites, captures screenshots for analysis.

n8n: The Nervous System

n8n is self-hosted workflow automation. Think Zapier, but free and you own your data. It connects everything:

Morning briefing pipeline: Telegram channels (via Telethon) to Claude analysis to Telegram bot delivery
Email sequences for lead outreach
Cross-posting content to Substack, Medium, Dev.to
HARO (journalist query) monitoring and alert routing

I have 8 scheduled jobs that fire throughout the day without me touching anything. By the time I sit down at 9 AM, my social media is done, my intelligence briefing is read, and my online presence has been maintained. I spent zero minutes on it.

Telethon: The Ears

Telethon is a Python library for the Telegram API. Not the bot API. The user API. It reads channels as if I'm scrolling through them myself.

I monitor 80 channels across four categories. Every message from the last 24 hours gets pulled, scored by engagement, filtered by keywords relevant to each business, and fed to Claude.

This replaced a human research assistant. I used to pay someone $1,500/month to "monitor industry news and summarize." Now a 200-line Python script does it better. It doesn't miss posts. It doesn't get bored on Friday afternoon.

ElevenLabs: The Voice

Two AI voice agents. Mike (English) calls leads for OD Granite. Dima (Russian) handles check-in calls to existing contacts.

The voice agents don't close deals. They qualify. They find out if someone picks up the phone, if they're interested, if they're the right person. Then I call back.

The Economics

Role	Market rate	What AI does instead
Content writer	$3,000-5,000/mo	Claude + daily post pipeline
Lead researcher / SDR	$4,000-6,000/mo	Playwright + Claude + ElevenLabs
Social media manager	$2,500-4,000/mo	Automated posting pipeline
Virtual assistant	$1,500-2,500/mo	n8n workflows + morning digest
Market research analyst	$3,000-5,000/mo	Telethon + Claude intelligence digest
Total	$14,000-22,500/mo	$197.23/mo

That's a 99% cost reduction. And the AI doesn't call in sick, doesn't need onboarding, and produces output at 7:00 AM on a holiday.

What Doesn't Work (The Honest Part)

AI hallucinations in client-facing content. Claude invented a testimonial once. Quoted a person who doesn't exist. I caught it because I check everything before it goes out. This is why I approve every LinkedIn post manually. The 30-second review is non-negotiable.

Voice agents on complex calls. Mike sounds great for the first 90 seconds. But when a prospect asks a detailed technical question, Mike stumbles. For simple qualification, 8/10. For real sales conversations, 3/10.

Automation that breaks. Playwright scripts are fragile. When a website updates its layout, your CSS selectors point at nothing. I spend 2-3 hours per week fixing broken automations. That's my real "employee cost."

The 2 AM problem. I built most of these systems between 10 PM and 2 AM because that's when my 6-year-old is asleep. The stack is cheap to run but expensive to build. Took me 4 months of late nights.

The Zero-Employee Framework

I didn't plan this as a framework. I needed to run five businesses without money for staff. Then clients started asking how I did it. So I named it.

Audit. List every repeating task. The one eating the most hours goes first.
Automate. Build one agent for that task. Not five. One.
Monitor. Run it for 2 weeks. Check output daily. Log failures.
Iterate. When it works without daily supervision, move to the next task.

The order matters. Don't automate your least painful process first because it's easier. Automate the one that hurts.

One More Thing

I got 2 hours back every morning. That's 10 hours a week. 520 hours a year. At $150/hour consulting rate, that's $78,000 worth of recovered time.

For $197.23 a month.

I Built a Prospect Research Agent with Firecrawl and Claude Code

Dima Negodiuk — Thu, 19 Mar 2026 00:03:30 +0000

I got tired of Googling prospects before sales calls.

Every time I'd prep for an outreach email, I'd spend 20-30 minutes on the same routine. Open their website. Scroll through the About page. Check their LinkedIn. Try to figure out what they sell, how they sell it, and where AI could save them time. Then I'd write a cold email that sounded like every other cold email.

So I built a thing. A prospect research agent that takes a company URL and spits out a 1-page briefing in about 90 seconds. I use it for my consulting business, but you could use it for any B2B outreach.

The stack is embarrassingly simple. Firecrawl for web scraping. Claude for analysis. That's it.

Setting It Up

First, install the Firecrawl MCP server. If you're running Claude Code, it's one config entry:

{
  "mcpServers": {
    "firecrawl": {
      "command": "npx",
      "args": ["-y", "firecrawl-mcp"]
    }
  }
}

Set your FIRECRAWL_API_KEY as an environment variable. Free tier gives you 500 credits/month. I've never gone over 200.

The Prompt

I wrote a system prompt that tells Claude what to extract. Nothing fancy:

You're a prospect research analyst. Given a company website, produce a 1-page briefing with:

1. COMPANY SNAPSHOT: What they do, who they sell to, approximate size
2. MANUAL PROCESSES: What looks like it's still done by hand (look for: contact forms, manual quoting, PDF catalogs, no API integrations)
3. AUTOMATION OPPORTUNITIES: 3-5 specific things AI could handle
4. ESTIMATED SAVINGS: Hours/week and rough dollar value per opportunity
5. CONVERSATION STARTERS: 2-3 specific observations to open a sales call

Use the company's own language. Reference specific pages you found.

Running It

I point it at a URL and let Firecrawl crawl the site. It grabs the homepage, about page, services/products pages, and any documentation it can find. Usually 5-15 pages depending on the site.

# Inside Claude Code, I run:
firecrawl_scrape({ url: "https://example.com", formats: ["markdown"] })

Then Claude reads through everything and builds the briefing.

A Real Example

I ran this on a staffing company in New Jersey last week. 45 employees, $8M revenue (per their Inc 5000 badge on the homepage, love when they make it easy).

The agent found:

Job applications still go through a PDF form you email back. No ATS visible.
Client intake is a contact form that goes to a generic inbox.
They post to 4 job boards manually (I could see the "also posted on" tags).
No chatbot. Their FAQ page had 47 questions, all static HTML.

Estimated savings: 15-20 hours/week on intake and posting alone. At $25/hr admin cost, that's $1,500-2,000/month. A voice AI agent handling the intake calls would save another 10 hours.

I sent them a cold email with 3 of those findings in the first paragraph. Got a reply in 4 hours. That never happens with generic outreach.

Iterating on the Output

The first version had problems. The briefing would hallucinate company size when it couldn't find revenue data. "Estimated $15M revenue" based on nothing. I added a hard rule: if you can't verify a number from the site itself, write "not found" instead of guessing. Fixed 90% of the accuracy issues.

I also added tech stack detection. Firecrawl picks up meta tags, script sources, and headers. If I can see they're running Shopify or HubSpot or a custom PHP site from 2014, that tells me a lot about their technical appetite. A company on Shopify is different from a company running a custom-built CMS. The first one will say yes to integrations. The second one will need convincing.

The other thing I changed: I made the briefing reference specific URLs. Instead of "their careers page could use improvement," it says "careers page at /join-us has a downloadable PDF application form with no online submission." Makes the cold email feel like you've done your homework. Because you have. Well, your agent has.

Scaling It

I ran this on 30 companies over the past two weeks. Each briefing costs about $0.03 in API calls (Firecrawl scrape + Claude analysis). Thirty briefings for under a dollar. Compare that to paying a VA $5-10 per prospect research report.

The time savings are stupid. 30 companies at 25 minutes each would be 12.5 hours of research. The agent did it in about 45 minutes total (90 seconds per company, but I run them in batches).

I'm now plugging this into my outreach pipeline. Scrape a lead list, run each URL through the agent, generate personalized first paragraphs, send. The whole thing from "I have 50 leads" to "50 personalized emails ready to review" takes about 20 minutes.

The Whole Thing Took 3 Hours to Build

Three hours of prompting and testing. That's it. Now every prospect email I send has specific, verified observations about their business. Response rates went from around 5% to closer to 20%.

The code is minimal. The value is in the prompt and knowing what to look for. Most B2B sites leak information about their operations if you know where to read. FAQ pages show you pain points. Career pages show you what roles they can't fill. Contact forms show you how manual their intake is.

If you sell B2B anything, build one of these. It's the highest-ROI afternoon I've spent this year.

I'm building AI systems like this for companies. If you want to see the other agents I run across 5 businesses: negodiuk.ai

One Person, Five Businesses, Zero Employees. Here's the System.

Dima Negodiuk — Mon, 16 Mar 2026 00:10:59 +0000

Everyone's building AI agents. Making demos. Posting screenshots of ChatGPT conversations and calling it "automation."

I'm running five actual businesses. Alone. No employees. No virtual assistants. No contractors. Just me and a system I built over six months.

Ecommerce brand on Amazon. B2B paint distribution. Granite import business. Logistics crisis management. AI consulting practice.

Let me show you how this actually works.

The backstory matters

I ran distribution businesses in Ukraine for 13 years. Six brands, ten retail chains, containers from China every month, team of 20. Then Russia invaded. Everything disappeared overnight.

I landed in New York in 2022 with my wife, our son, and a laptop. No connections. No clients. No English good enough for a corporate job.

I had to rebuild. But this time I couldn't afford 20 employees. I couldn't afford one.

So I built a system instead.

What the system does before 10 AM

Here's a real Tuesday morning.

7:00 AM. I wake up to a summary on my phone. Overnight, the system read 10+ Telegram channels. AI news, industry groups, competitor chatter. It scored every message by relevance to each of my five projects. What used to take me 45 minutes of scrolling now takes 2 minutes of reading.

7:30 AM. A LinkedIn post draft arrives. The system pulled the most interesting insight from yesterday's monitoring, matched it to my voice, and wrote a draft. I spend 3 minutes editing. Sometimes I rewrite the hook. Sometimes it's perfect. Post goes live at 8.

8:40 AM. Business intelligence report hits. This one is different. It tracks specific opportunities across all five projects. A Reddit thread where someone needs AI consulting. A competitor who changed their pricing. A new Amazon listing in my product category.

9:30 AM. Reddit engagement is ready. The system found 4 posts across 6 subreddits where I can add genuine value. Draft responses written. I review, tweak, post. 10 minutes instead of an hour.

That's before my second coffee. One project out of five.

For the ecommerce brand, the system handles Instagram Stories 3 times per day, YouTube Shorts twice per day, review monitoring across Amazon and social platforms, and a BuyBox tracker that pings me when a competitor steals my listing.

For B2B distribution, it scraped 146,000 contractor leads from government databases. Enriched them with emails. Queued personalized outreach.

For the logistics client, it monitors Google results for reputation threats and drafts responses using 17 templates depending on the sentiment.

The stack (it's boring)

People always ask about the tools. Here's the truth. The tools don't matter.

But fine. Claude API as the brain. n8n for workflow automation. Telegram bots for notifications. Cron jobs running on my MacBook. SQLite databases.

Total cost. About $40/month in API calls and $8/month for LinkedIn analytics. Everything else runs on free tiers or my own hardware.

$48/month. Not $48,000/year for a team.

The magic isn't in any single tool. It's in how they connect. Every system feeds into the next. Morning monitoring informs LinkedIn content. LinkedIn content drives inbound. Inbound creates consulting clients. Consulting insights improve the system. Loop.

What I killed

I deleted 3 automations last week. On purpose.

First one. A competitor price tracker that checked 8 competitors every 4 hours. Beautiful dashboards. Never once changed a decision I made. Pure vanity metrics. Gone.

Second. Automated LinkedIn DMs. Full personalization, AI-researched icebreakers, auto-send. Technically impressive. But people smell automation in DMs. The inbox is sacred. Now I use AI to research the person and draft the message, but my finger hits send. The human touch matters exactly where trust matters.

Third. A news aggregator that was pulling from too many sources. 80% noise. I cut it to the 5 channels that actually influenced my decisions.

Here's the rule. If an automation can't justify its $0.02/day in API costs through measurable impact, it dies. Hours saved. Revenue generated. Errors prevented. Pick one or get killed.

The uncomfortable truth

This system didn't make me rich. Not yet.

One project generates $5,000/month from a crisis management client. The ecommerce brand is growing. The B2B businesses have 146,000 leads in the pipeline. Consulting practice is building.

But I'm not sitting on a beach. I work 9 AM to midnight most days. The system doesn't eliminate work. It eliminates the wrong work. Instead of spending 30 hours a week on manual tasks, I spend that time on strategy, client calls, and building.

The real number. 25-30 hours per week freed up. That's a full-time employee I don't pay. That's the difference between running five businesses and drowning in five businesses.

What you should build first

Don't build a "system." Build one thing.

Pick the task you hate most that also eats the most time. For most business owners it's one of these.

Morning briefing. If you check 5+ sources every morning, consolidate them. One summary instead of 30 minutes of tab-switching.

Content drafting. If you spend 2 hours per day on emails and social posts, have AI draft, you review. Instant savings day one.

Lead research. If you're doing B2B, build a scraper and AI qualifier. Instead of finding 5 leads per day, you find 50.

Get that first one working. Measure the savings. Then build the second. Then the third.

Six months later, you'll have a system. You won't even realize you built one.

The real game

The headlines say "AI replaces jobs." Wrong framing.

AI replaces the version of you that does repetitive work at 2 AM while trying to keep five projects alive. I still make every strategic decision. I still write the final version of important messages. I still get on calls.

But the 25 hours per week I used to spend on copying data between spreadsheets, scrolling Telegram, and writing first drafts? That time goes to thinking. Building. Actually running the businesses instead of being buried inside them.

The companies that build systems in 2026 win. The ones that hire 10 more people to do work a system handles? They'll wonder why their margins keep shrinking.

One person. Five businesses. Zero employees.

Not a flex. Just the new math.

Dmytro Negodiuk is a Fractional AI Officer based in NYC, helping startups and mid-size businesses build AI systems that work. He runs 5 businesses simultaneously with zero employees. Previously built a distribution empire across Ukraine with $12M+ in revenue over 13 years.

Website: negodiuk.ai
LinkedIn: linkedin.com/in/dima-negodiuk