How I run 7+ projects without a team

I want to show you something a bit wild.

~8ish weeks ago I set up openclaw agent on a cheap VPS. $7/month. Nothing fancy.

Today it runs 16 automated workflows, manages client projects, writes code, generates ad creatives, does media buying, processes leads, syncs data across 15+ tools, and has shipped over 50,000 lines of code on its own.

aand the funny part is that I can’t write a single line of code…

Most people I talk to are using AI to write captions or summarize articles. Cool. But that’s like buying a commercial kitchen to heat up leftovers.

Let me show you what this thing actually does.

what happens before I wake up

Every morning, 7am. A bunch of jobs just… run.

It pulls all my Meta data across every client. Writes it into my custom Google Sheets dashboard. By the time I open my laptop, numbers are there. I didn’t do anything.

Then it looks at every single ad. CTR, link clicks, checkout signals, purchases. Scores them. Writes a report with kill/scale recommendations. Last week it told me to kill 5 ads and scale 2 winners running at $11 CPA. I just… agreed. It was right.

A separate job runs pattern detection. Not just “this ad works” but stuff like “hooks framed as questions outperform statements by 40%” or “the inefficiency exposure angle converts but speed/AI angles don’t.” These patterns get saved to a knowledge file that feeds into the next batch of creative decisions. So it’s learning what works for MY business, not marketing in general.

It also syncs newsletter metrics, LinkedIn engagement, even X impressions (even though I haven’t really ramped up my X presence yet…) into my Notion content pipeline. I never manually check how a post performed. It’s just… there.

All before coffee.

the 15-minute loops

Two jobs run every 15 minutes, all day:

Lead processor. New lead hits my personal website, gets enriched, routed. If it’s hot I get a Discord ping immediately and they get automatic follow up from BOSSMODE (that’s my agent’s name).

Sales monitor. Watches adhookvault.com for purchases, checks delivery emails, flags payment issues. Last week it caught that one of our products wasn’t auto-delivering access. Would’ve been a terrible customer experience. Flagged it, fixed it same day.

But honestly the stuff that blows my mind most is the ad-hoc work.

client calls

I had a discovery call with a potential client recently. Gave my agent an email thread and a PDF of my mentor Jonathan Stark’s “Learn Your Lines” framework.

It spawned 3 research agents. In parallel. One dug into the person (background, agency, career history). One found and analyzed their outdoor brand. One mapped out their skincare brand, competitive landscape, current Meta strategy, pricing.

5 minutes. I had a complete call prep doc with discovery questions, pricing responses, competitive intel. Ready.

After the call I fed it the transcript. Got back a debrief with 10 action items and a summary written specifically for my partner who wasn’t on the call. His style, his context.

Another time I needed to look into a potential referral. Nobody could find their brand anywhere. Spawned 3 agents, they came back with “this doesn’t exist publicly yet, probably an internal name” Saved me an hour of googling and looking like an idiot asking “so what do you guys actually do?”

$0.07-$0.10 ad creatives

This one might be my favorite.

Built an AI pipeline that generates ad creatives autonomously. Research, psychology-based hooks, AI image generation, QA. Full pipeline.

Cost per creative: somewhere between $0.07-$0.10. Not a typo.

We’ve tested 50+ ads in one campaign now. 9 converted to purchases. Best performer hit $11.60 CPA. And the pipeline audits its own output daily. Reviews what worked, what flopped, updates its own knowledge, generates better stuff next round.

It’s getting better at my specific product. Not ads in general. MY product.

the posthog moment

One morning I asked it to deep dive our funnel analytics. Device, browser, country, session duration, rage clicks, everything.

Three findings that would’ve taken me weeks to notice manually:

Facebook’s in-app browser converting at 1.8%. Desktop at 4.4%. Biggest hidden conversion killer sitting right there in the data, nobody looking at it.

People who clicked our FAQ converted 6x better than people who didn’t. 6X!!! (this is a learning to move FAQs or build an educational LP)

Google organic converting at 25% (which makes sense, they’re searching for exactly what we sell). But it quantified it - put a dollar number on what SEO is worth per month. Made the business case for me without me asking.

geo expansion

Decided to test international markets. Told the agent which countries. It built 3 campaigns (UAE, Thailand, South Africa), each with proven ads, same post IDs. Budgets, targeting, age ranges - all set up.

Then it ran a legal check on our LLC structure for international sales tax at current volume. I didn’t ask for that. It just… did it. Because it knew we were expanding.

Thailand hit 4 purchases on $39 spend the first day. The data was in my sheets before I even saw the notification.

When tracking was off, it ran a full audit. Pixel, CAPI Bridge, PostHog, Stape. Found 4 issues. Wrote fix scripts. One of them - UTM passthrough - was silently breaking our attribution. We would’ve been making decisions on broken data for weeks.

the part people don’t believe

My agent keeps a daily diary. 52 files so far.

Every morning at 6am it reviews yesterday’s notes, extracts patterns, updates its own knowledge base. Sunday nights it runs a full self-audit. What should it change about how it operates?

It has playbooks it wrote for itself. A growing understanding of my business, my clients, my preferences, even my writing voice (it wrote this newsletter draft… I edited it, but still).

When I start a new session it reads yesterday’s notes. Picks up where we left off. No “remind me what we were doing” needed.

so what does this mean for you

Look, you don’t need my exact setup. The shift is simpler than you think.

Stop using AI as a tool you open when you need something. Start treating it like someone who works for you.

Here’s what I’d do starting this week:

• Give it memory. Write a one-page brief about your business - what you sell, who buys, priorities, voice. Paste it every conversation. That alone changes everything.
• Give it recurring tasks. Whatever you do every Monday, write it as a checklist. Run it through AI weekly. Same prompt. By week 4 it feels like a trained employee who knows your stuff.
• Connect it to real data. “Write me ad copy” is fine. “Look at my top 5 performing ads, tell me why they work, then write 3 new ones” is a completely different universe.
• Let it teach itself. After every task: “what would you do differently?” Save the answer. Feed it back next time.

The gap between “I use AI sometimes” and “AI runs my operations” is way smaller than you think. You just have to stop treating it like a search engine.

Talk soon,

Kris

P.S. Want me to go deep on any of these? The creative pipeline, the call prep system, the analytics setup - hit reply and tell me which one. I’ll break it down in a future issue.

Kristijan Arapov is a DTC growth consultant who has managed $100M+ in Meta ad spend. He helps consumer brands break through revenue plateaus with AI-enabled marketing systems. Work with Kristijan