Beginner Volume I · Foundations
Understanding Agentic AI Systems
From GenAI to Agentic AI, Explained Simply (A Beginner's Guide)
For software people new to AI. Assumes you can read code, no AI background required.
Home / About
About
Muhammad Arbab
Technology executive. 20 years building enterprise systems, 14 of them shipping AI.
Muhammad Arbab
Technology executive with 20 years in enterprise systems and 14 shipping AI in Fortune 100 organizations.
The work has been about getting AI to actually run in the messy reality of large enterprises, not the demo-grade kind. Most of it has been the unglamorous middle, not the launch demo and not the research paper, the part where a system has to pass an audit, fit a budget, and not embarrass anyone at the next board meeting.
I help teams design and ship agentic AI systems in production. The two books at Building Agentic AI are how I share that work at scale. They exist because the existing material on agentic AI tends to fall into one of two buckets: either a vendor explainer with no engineering substance, or research-flavored writing that skips over the parts that make a system actually work in production. I wanted clear, honest books that respect the reader's time and their experience.
Book 1 is the on-ramp for software people new to AI. Book 2 is the architect's field guide for engineers shipping these systems for real. They are a deliberate pair, and the articles cover the same ground in smaller pieces.
Why I'm building this
The thesis is simple: there should be more agentic AI that actually ships. Not more demos. Not more vendor decks. Working systems that pass security review, run on a budget, and survive on-call.
The gap between "we have a promising prototype" and "this is a system real users depend on" is where most agentic AI initiatives stall. It is also where the work I have done for the last decade lives. The books, the articles, and the hands-on help are all aimed at closing that gap, one team at a time.
How I work
- · Listen before pitching. Most teams already know roughly what they want to build. The work is figuring out which 10% of the scope can ship reliably this quarter and which 90% will cause pain six months in.
- · Say so when something is a bad idea, including when the bad idea is mine. A senior second opinion is worth what it costs precisely because it does not validate every plan.
- · Fixed scope and fixed price. Open-ended retainers drift, lose accountability, and stop being good for either side. Every engagement is scoped in writing before it starts.
- · Write things down. Decisions, tradeoffs, and the rationale that produced them. Memos survive the engagement, conversations do not.
- · Leave systems your team can run, test, and trust without me. Knowledge transfer is part of the deliverable, not an afterthought.
Working together
A 30-minute call is enough to see if there's a fit.
Background
- · Operating model: getting AI past the demo wall and into systems that pass security review, run on a budget, and survive on-call.
- · Writing here is the same material, simplified and made public. No vendor pitch, no hype, no listicles.
What I write about
- · Agents, the agent loop, tools, memory, and bounded autonomy.
- · Production architecture: gateways, RAG pipelines, model routing, multi-tenant platforms.
- · Evaluation, reliability, cost, and the failure modes that actually bite.
- · Security, governance, and the OWASP LLM Top 10 in real systems.
Find me
- LinkedIn: linkedin.com/in/arbab
- Twitter / X: @buildingagentic
- Articles: /blog (RSS at /rss.xml )
The books
Enterprise Volume II · Practice
Designing Enterprise Agentic AI Systems
An Architect's Field Guide for Engineers
For engineers and architects building agentic systems in production. Assumes a working understanding of GenAI and agents (the level Book 1 teaches). Also serves senior AI-engineering interview prep.