Hamza Akbar · Operator · Systems · AI
I'm the COO of a 35-person agency, and I've built the systems we run on with AI. Now I want to help other teams work out what to systemize and where AI actually fits into their work.
Right in the middle of university, I started out as a graphic designer. Not long after, I shifted to Amazon.
I'm self-taught. I learned by doing the work in front of me and figuring it out as I went. Over time the technical side became my strength, and in 2021 that's what got me the COO role.
Today I run operations for a 35-person team managing 70+ Amazon brands. Most of what keeps it running is systems I built with AI. I still work the same way: build something, test it on real work, fix what doesn't hold.
Started designing in the middle of university.
Learned PPC, listings and account work on the job.
Took over the full operations stack: accounts, PPC, design, partners.
Learning, applying it inside the agency, and learning from what breaks.
I'm in the daily grind of running an agency: client calls, team problems, accounts that slip. When I build a system, it's for a problem I deal with every week. I think like an operator, the same way my clients do.
Direct work with clients and the team. I know where things actually break.
Automation, workflows and AI models, turned into things the team uses every day.
Systems that fit how an operation really runs, not how it looks on paper.
I'm not claiming to have this figured out. I learn something, apply it to a real problem inside the agency, see what breaks, and learn from that.
Go deeper into systems building and AI, and help other agencies and operators build the same kind of systems for their own work.
Talk systems with me →I'm not selling a platform. I help teams work out what's worth automating, design it around how they actually operate, and bring AI in where it earns its place.
Look at where your team makes the same decisions or does the same manual work every week, and pick what's worth building first.
Map the steps, the rules and who owns what before anything gets built, so the system fits the way your team works.
Use AI where it actually helps: reading calls, summarizing, classifying, drafting. People keep the decisions that matter.
Build on tools you already use, test it on real data, and make sure the team actually uses it. Adoption is usually the hard part.
Each one started with a problem that kept coming back every week. I work out the logic, AI writes the code, and we test it on real work before the team relies on it. These run inside my own agency.
Built with
Drag the cards, or use the arrows
Lessons from building systems inside a real business. Most of them came from getting something wrong first.
If someone makes the same call every week, that's the first thing worth turning into a system. It's usually not the most exciting problem, just the most frequent one.
Same decision weekly → systemWhen people keep breaking a process, it's usually too expensive to follow. Make the right thing easier and compliance follows. More rules rarely help.
Less friction > more rulesA system isn't done when it runs. It's done when it's been checked against real output and the team understands what it's doing, not just trusts it.
Real data before rolloutMost of what I've built runs on spreadsheets, scripts, Slack and an AI model. You don't need a complicated tech stack to remove hours of manual work.
Weekly updates had an SOP and a format. Compliance was still poor: few went out consistently and formats varied. So I stopped enforcing and removed the friction instead.
Each team member gets a pre-written, structured update every Friday midday, generated from their tasks.
When people keep breaking an SOP, it's usually too expensive to follow.
Support tickets with Amazon were closing on their own because nobody noticed the replies in time. The fix took about ten seconds per ticket to set up.
Tell me about it. I'll tell you honestly whether a system would help.
If your team is buried in repeated manual work, or you want to bring AI into how you operate but don't know where to start, send me a message.