diego bianqui

Agentic AI, from experiment into production.

I’m Diego Bianqui, an Agentic AI Engineering Manager and Technology Consultant. I help organizations adopt, engineer and operationalize AI agents: secure, governed and built to run.

Get in touch

Opens an email, or copy the address: hello@diegobianqui.com

Expertise

Engineer it. Adopt it. Secure it.

Agentic AI engineering

AI agents and LLM-based systems: orchestration, tool use and evaluation, built for production.

Context and Harness Engineering

Agent architectures and orchestration. Context engineering: what the model sees, and when. Harness engineering: the loop, tools, permissions and guardrails around the model.

AgentOps

Running agents reliably after launch: evaluation, tracing, cost and latency control, and the release practices that keep quality from drifting.

Adoption and strategy

From AI experimentation to scalable, production-ready agentic systems.

Where agents create value

Identifying the processes where agentic architectures pay off, and the ones where a simpler workflow is the better answer.

Teams and practices

The engineering practices, architecture standards and multidisciplinary teams needed to put agents into production and keep them there.

Security, governance and compliance⁠

Established security and risk frameworks applied to the design, development and deployment of AI systems.

Agent security

Threat modelling and controls mapped to the OWASP Top 10 for Agentic Applications 2026 (ASI01 to ASI10), from the OWASP GenAI Security Project’s Agentic Security Initiative.

AI governance

AI management systems aligned with ISO/IEC 42001:2023 and risk management with the NIST AI Risk Management Framework, for secure, trustworthy and compliant AI.

Agent protocols

Agents talk to people, to other agents, to tools and to markets. Each layer has its protocols.

  • Human ↔ Agent

    AG-UI · A2UI

    Agent–User Interaction Protocol · Agent-to-User Interface

    Streams the agent’s state to the user and generates dynamic interfaces.

  • Agent ↔ Agent

    A2A · ACP · ANP

    Agent2Agent · Agent Communication Protocol · Agent Network Protocol

    Cooperation, task delegation and negotiation between different agents.

  • Agent ↔ Tools

    MCP

    Model Context Protocol

    A universal, secure connection to databases, files and third-party APIs.

  • Agent ↔ Economy

    AP2 · UCP

    Agent Payments Protocol · Universal Commerce Protocol

    Autonomous payments, budget limits and commercial transactions.

Experience

15 years in the corporate top flight.

At Accenture I led the delivery of software engineering and audit services to a wide range of Fortune 500 clients in telecommunications, resources and metals, financial services and health care. I then created and led a multidisciplinary software engineering studio for Accenture’s internal organization, combining hands-on engineering with architecture, delivery leadership and organizational transformation.

  • Now specializing in agentic AI through continuous hands-on experimentation and research.
  • Certifications and training from Microsoft, NVIDIA and Anthropic.
  • A broader background in software engineering, automation, emerging technologies and blockchain.

Certifications

Verified, not just claimed.

How I work

Hands-on engineering and organizational adoption, together. The goal is always the same: agents that are useful, reliable and safe in production, and a team that can run them.

  1. Assess.

    Where agents create value, what they need to touch, and the risks that come with it.

  2. Architect.

    Agent architecture, orchestration and protocols such as MCP and A2A, with the controls the risks call for.

  3. Engineer.

    Context and harness engineering, tool use and evaluation, built with your engineers.

  4. Operationalize.

    AgentOps practices, governance aligned with ISO/IEC 42001 and the NIST AI RMF, and the team to run it.

Key concepts

What is agentic AI?

Software that uses large language models to plan and take actions toward a goal, calling tools and APIs under delegated authority, instead of only answering a prompt.

What is context engineering?

Deciding what goes into a model’s context window at each step: instructions, retrieved knowledge, memory, tool results and state. Most agent failures are context failures, not model failures.

What is harness engineering?

Designing everything around the model that turns it into a working agent: the control loop, tool definitions, permissions, guardrails, retries, human approval steps and the evaluation that proves it works.

What is AgentOps?

The operational discipline for AI agents in production: tracing every step and tool call, evaluating quality continuously, controlling cost and latency, and releasing changes safely.

Which protocols do AI agents use?

MCP connects agents to tools and data. A2A, ACP and ANP let agents work with other agents. AG-UI and A2UI connect agents to people and interfaces. AP2 and UCP cover payments and commerce.

How do you secure and govern AI agents?

With established frameworks: the OWASP Top 10 for Agentic Applications for agent-specific threats such as goal hijacking and tool misuse, ISO/IEC 42001 for an AI management system, and the NIST AI RMF for AI risk management.

Contact

Moving from AI experiments to agents in production? Let’s talk.

Email hello@diegobianqui.com

Remote, worldwide.