Zaher Chehayeb —
AI Systems and Automations —
Creative Operations

AI systems thatknow when to escalate.

From brief to sign-off, I build the agents and automations that run the work in between, and escalate when needed.

Ten years across creative operations, production, and design taught me one loop. Brief, spec, proof, sign-off. I build it into systems for creative teams and business operations.

How every system runs

Ten years across creative operations, production, and design taught me one loop. Brief, spec, proof, sign-off. I build it into systems for creative teams and business operations.

Input: It starts with mess.

A creative brief written as prose. Supplier quotes and invoices in any layout. Questions in any wording. Every system I build begins where the data is untidy.

Read: The model reads.

Language becomes structure: a spec, a quantity, a deadline, an owner, a total. That is the model's whole job. It never does the arithmetic.

Check: The rules decide.

Scripts check every number, ID, date and spec against the brief. Anything that doesn't fit is escalated for human intervention, once, however many times it's asked.

Routed for human intervention

Deliver: Then it ships.

A signed-off job, a ledger row, a cited answer, a status report. Each one carries its source, so anyone can see why it says what it says.

Selected systems.

Each one built with its rules written down, and a person at the end of every fallback path.

telegram bot + dashboard · real estate

Office Management Bot

Eleven agents ask the bot first. It answers only from the coach's approved handbook and cites the section every time. Anything new reaches the coach once, however many ask, and his reply comes back as a draft section to approve.

11
agents asking one bot
1
message to the coach
28/28
acceptance checks passed

claude api · telegram · next.js · supabase · pgvector · vercel

office-management-bot / coachdemo data
Office Management Botanswers from the approved handbook
The coach's dashboard: 30 days of questions drawn as bricks, gold where the coach answered, with resolution rate and accuracy

agent · clickup + slack

Project Management Bot

Turns a brief into ClickUp tasks with one owner, a due date and dependencies. Every minute it watches for blocked work, chases the owner on Slack and escalates when nobody replies. "Thanks" is never taken as approval.

1 min
watcher interval
4
rules the model can't override
0
tasks closed unapproved

Watch it run, from brief to Slack

pm-bot / workflowwatching

next.js · marketing performance

Marketing dashboard

For an agency owner who doesn't read spreadsheets. It pulls every opportunity from GoHighLevel, cleans and attributes it, and never sends a raw record to the browser. Missing sources are counted, not dropped.

62
unit tests on the data logic
0
raw records in the browser
14 d
stale-lead flag

Open the live dashboard

marketing-dashboard / livevercel
The marketing dashboard: lead volume by source as a bar chart, with the pipeline by stage below

n8n · inbox to ledger

Invoice Processing Workflow

The model extracts vendor, number, date, total and currency from emailed documents in any layout. The workflow cleans them, files them in Drive, catches duplicates the moment they arrive, and emails a daily summary.

31
nodes in one workflow
2
duplicate checks before filing
7
real-world edge cases

n8n · gemini · gmail · google drive · sheets · javascript

invoice-workflow / n8n31 nodes
The Invoice Processing Workflow in the n8n editor, from Gmail trigger to daily summary
  • 2.280,04 € read as 2280.04 EURfiled
  • second notice already in the ledgerduplicate
  • delivery note "this is not an invoice"rejected
  • scanned pdf no text layervision model

n8n

More automations.

The Support Qualifier workflow in n8n

AI Customer Support Qualifier

Classifies every email to a beauty brand's inbox as a refund or an inquiry, logs it and replies by name. Two categories, nothing else allowed.

The Sourcing Agent workflow in n8n

AI Sourcing Agent

Files every candidate CV from ClickUp into a clean Drive database and a vector store for search. One file per candidate, matched by email.

About Me

Sign‑offs, now in code.

For more than a decade I turned creative concepts into artwork and products for the real world, working among designers, suppliers, sales and manufacturing teams, where one wrong value means a pallet of waste.

That work was always about systems: structure, specifications, checks and sign-offs. AI gave me a faster way to build them. Now I build agents, automations and data tools for the same kind of operational work, the way I prepared production files for years: every input checked, every exception routed, nothing assumed.

Zaher Chehayeb, black-and-white portrait

Let's build together.

Have a process that breaks when someone's on leave? Need an agent to run your setup for you? Tell me what eats your team's hours, and let's talk.

zaherchehayeb@gmail.com