This TechTalk series focuses exclusively on agentic AI and enterprise AI agents, covering how agents are designed, evaluated, governed, and operated across their full lifecycle. The sessions provide a structured progression, from foundational evaluation concepts for non‑deterministic AI systems, to practical design of evaluation metrics and success criteria, and finally to governance, lifecycle gates, and responsible operation in production environments. Collectively, the series equips technical and business leaders with a shared framework to assess quality, safety, reliability, alignment, and business value of AI agents, enabling confident adoption and scalable operation of agentic solutions within enterprise business processes.
This web conference is virtual event and is being offered live at a single time. If you’re unable to join, please note that the recording will be made available on the TechTalks Dynamics Community Page following the event.
This is part of our recurrent Wednesday meeting. To join, download the .ics file attached in this blog post https://aka.ms/D365TTSchedule. There you can check more upcoming TechTalks.
| Date | Start Time | TechTalk Title | Description | Presenters |
| Apr/22/2026 | 7:00 AM | Agent Evaluation Series: Part 1-Foundations & Evaluation Fundamentals for Agentic AI | As AI agents become embedded in enterprise business processes, traditional testing alone is no longer sufficient to ensure quality, safety, and business value. This session introduces the foundational concepts of agent evaluation—why evaluations are necessary for non‑deterministic AI systems, how they complement existing testing practices, and what “good” looks like for agent behavior. Attendees will learn the core evaluation dimensions (correctness, safety, reliability, alignment, and efficiency), the difference between offline and online evaluation, and the primary evaluation methods used across the agent lifecycle. This session establishes a shared language and mental model for evaluating agentic AI solutions with confidence. | Vishal Singh
Akshat Singh
Sauarbh Bharati
Ashley Desiongco
Timo Gossen |
| May/6/2026 | 7:00 AM | Agent Evaluation Series: Part 2: Designing Evaluation Sets, Metrics, and the Evaluation Blueprint | Effective agent evaluation starts with intentional design. In this session, we move from theory to practice by exploring how to design evaluation sets, metrics, and success criteria that reflect real business processes. Attendees will learn how to structure scenario‑based evaluation sets, balance representative and edge‑case coverage, and choose between synthetic and real‑world derived data. The session also introduces the Evaluation Design Document (EDD) as a practical blueprint for defining what is evaluated, why it matters, and how results inform decisions. By the end, participants will understand how to design evaluation programs that are measurable, repeatable, and aligned to business outcomes. | Amira Beldjilali
Ahmet Yildirim
Ashley Desiongco
Timo Gossen |
| May/20/2026 | 7:00 AM | Agent Evaluation Series: Part 3- Governance, Lifecycle Gates, and Operating Agents in Production | Evaluating an agent does not stop at go‑live. This session focuses on how evaluation is operationalized through governance, lifecycle stages, and continuous operation. Attendees will learn how evaluation requirements evolve from design through production, how evidence‑based gates support go/no‑go decisions, and how Responsible AI considerations are enforced throughout the lifecycle. The session also covers production experimentation patterns (such as shadow mode and A/B testing), continuous monitoring versus evaluation, and how production signals feed back into ongoing improvement. This session equips teams to operate agentic AI solutions responsibly, transparently, and at scale. | Priyanka Sinha
Ashley Desiongco
Timo Gossen |
