Napačna izbira? Nič za to! Izdelke lahko vrnete do 30 dni
Z darilnim bonom ne morete zgrešiti. Obdarovanec lahko v zameno za darilni bon izbere karkoli iz naše ponudbe.
Do 30 dni za vračilo
Every AI agent proposal makes a promise. This book shows you how to make it prove it.
Management wants an agent. A vendor promises it will transform the business. The demo looked flawless. Now you are the architect who has to decide. Prove It, Agent is the playbook for that moment: a practical method for testing AI and agent proposals against evidence before they reach production, whether they are built in Microsoft Copilot Studio, Microsoft Foundry, Claude, or custom code.
Written for solution and enterprise architects, architecture review boards, and the security, governance, and engineering leads who work with them, the book follows the order in which a real decision unfolds. It moves from triage and risk tiering, through autonomy, platform, and model choices, to integration, security, governance, and operations.
Inside, you will learn how to:
- Triage any AI request and assign a risk tier before choosing a model
- Choose between Copilot Studio, Microsoft Foundry, Claude, or a custom build
- Run model bake-offs on your own data and avoid eighteen common selection pitfalls
- Design MCP servers, tools, and identities that keep an agent's reach in check
- Contain prompt injection and engineer down an agent's blast radius
- Govern agents with the EU AI Act, NIST AI RMF, and Microsoft Agent 365
Every chapter ends with the questions to ask a proposing team and the evidence to require before approval. The appendices include a consolidated architecture review checklist and an AI decision record template you can adopt as is. Product details that change often are marked with verification flags, so the reasoning stays useful as the platforms evolve.
Pozdravljeni! Sem Libroamiko, vaš knjižni svetovalec.
Kako vam lahko pomagam?