Useful autonomy
Agents should own outcomes, not merely generate responses. Every degree of autonomy must earn its place through measurable utility.
Agentic systems · Architecture · Delivery
I engineer
agency for
the enterprise.
Mohsen Amiribesheli, PhD. Director, AI Labs at HCLTech. Author and instructor, based in London. I turn autonomous-system ambition into secure, observable and accountable production systems.
Human in command
Mohsen AmiribesheliDirector, AI Labs · HCLTech
51.5072° N0.1276° W
Building systems that can
reason, act and answer for it.
Director, AI Labs · HCLTech
London, United Kingdom
My thesis
The next enterprise interface isn't another dashboard. It is an agent that can make progress inside clear boundaries, with evidence.
Control by design
Identity, permissions, policy, human approval and auditability belong in the architecture, not in a checklist added before launch.
Production truth
A compelling demonstration is only a hypothesis. Traces, evaluations, incidents and business outcomes reveal whether the system works.
The book
01 / 05
The production field guide
Agentic AI
Engineering
Engineering
Built for the gap between demo and deployment.
A practitioner's guide to designing, shipping and governing agent systems that survive contact with real enterprises.
Building an agent demonstration takes an afternoon. Getting one through security review, compliance audit and its first month in production takes far longer.
This book writes down what survived real enterprise delivery. It draws on work across pharma, financial services and manufacturing, so engineers, architects and leaders can skip the most expensive lessons.
Reference architecture
Five subsystems. One accountable agent.
The same production architecture runs through every chapter, turning abstract patterns into an operable system.
System 00Agent
Runtime
Runtime
Reasoning core
How the agent plans, decides and adapts to each job.
Tool registry
What it can touch, under which identity and with what trust.
Memory
What persists, for how long and who else can retrieve it.
Orchestration
How agents route work, hand off and recover from failure.
Constraints
Policy, approvals and evidence that make agency accountable.
What I do
From boardroom ambition to operational agency.
At HCLTech AI Labs in London, I lead advisory work on agent architecture, governance and enterprise transformation. My work sits at the seam between research and production, where regulation, legacy systems and organisational reality are part of the engineering brief. I care about what works on Monday morning.
Agent architecture
Multi-agent systems built on open protocols, designed for observability, testability, interoperability and cost control.
↗Governance & assurance
Guardrails, identity, evaluation and oversight that let risk, compliance and security teams say yes with confidence.
↗Production strategy
Clear paths from proof of value to reliable operations across regulated, complex and high-stakes environments.
↗Executive advisory
Sharper investment decisions: where agents create durable advantage, what they require and what should never be autonomous.
↗Ideas in motion
Field notes.
Writing, research and practical frameworks for people building the next generation of agent systems.
LinkedIn · Ongoing↗
Production lessons from the front line of agent engineering.
Short-form analysis, architecture decisions and field-tested patterns.
YouTube · Watch↗
Agent systems, explained without the theatre.
Technical walkthroughs and practical perspectives.
Scholar · Research↗
Research across adaptive and intelligent systems.
Peer-reviewed publications and academic work.
Have an ambitious problem?
Let's give it agency.
For advisory work, collaborations, speaking or conversations about the book, start here.
besheli@outlook.com ↗