What exactly is an Agentic AI product? (full code and low code)

Background 

Thanks for all the feedback to my post about AI agents glossary 

I created that glossary for the  Oxford AI summit is based on the theme of Autonomous AI agents - this event enables you to get a certificate from Oxford in Autonomous AI agents. 

I am extending these ideas to a few more concepts for agents 

In this post 

I explore the idea

What exactly is an agentic AI product? (full code and low code). It seems - like the image of classroom - everyone raises their hand when you ask them if they are an ‘agentic’ product.  So, how do you understand what is truly an agentic product - full code and low code

Claudia Saleh and I are working on the idea of a session in the summit on low code agentic AI products. Here are some ideas which we have been thinking of 

I used the glossary to discern these characteristics 

Overall characteristics of an Agentic product

Autonomy: The product should be able to perceive, decide, and act independently based on goals.

Goal-Oriented Behavior: The product accepts and optimizes for explicit or evolving goals, using tools like planners or reward functions.

Tool Use / Toolformer Ability: The agent can decide when and how to use APIs, external tools, or plugins to achieve tasks.

Memory and Context Awareness:: Maintains persistent memory across interactions; recalls previous actions, user preferences, and plans.

Planning and Reflection: Decomposes goals into sub-tasks, evaluates outcomes, and revises its strategy over time.

Perception and Environment Awareness: Understands and interacts with its digital or physical environment (e.g., files, sensors, web).

Action Execution : Can execute actions directly (e.g., send emails, move a robot arm, update databases).

Multi-Agent Collaboration: Can communicate and collaborate with other agents to perform complex or distributed tasks.

Self-Improvement and Learning: Learns from experience to improve performance over time without reprogramming.

Personalization and Adaptability : Adapts to individual user needs, roles, and domains using profiles or learned preferences.

Interoperability with External Systems : Works within MLOps/DevOps pipelines, integrates with APIs, databases, enterprise tools.

Safety, Control, and Governance : Built-in ethical alignment, safety protocols, and audit trails for decisions and actions. 

Now .. what are the characteristics of a low code agentic system? 

Characteristics of a a Low code Agentic product

Drag-and-Drop Goal Specification: Users define goals and workflows through a visual interface (e.g., flowcharts, prompts, rules), not code.

Prebuilt Tool Integrations: Comes with ready-to-use connectors for APIs, databases, CRMs, documents, web browsing, etc.

Visual Memory and Context Management: Provides dashboards or timelines to visualize agent memory, past tasks, and user interactions.

No-Code Prompt Engineering: Allows prompt tuning and chain-of-thought design via templates, toggles, and dropdowns instead of manual text entry.

Composable Agent Architectures: Lets users combine multiple agents visually (e.g., planner + executor + retriever) to build workflows.

Domain-Specific Personas: Users select or create agents for specific roles (e.g., marketing analyst, legal assistant) with predefined capabilities and constraints.

Built-in Guardrails and Governance : Users can set guardrails visually—like banned actions, approval loops, or restricted tools.

Real-Time Simulation and Debugging: Sandbox mode for testing agent decisions before deployment with visual logs and outcomes.

AgentOps Layer (Deployment & Monitoring):  GUI-based dashboards for launching agents, setting triggers (e.g., time-based, event-based), and monitoring performance.

Template and Component Marketplace: Library of reusable components (agents, tools, prompt flows) that users can plug and play.

Learning and Personalization Modules : Prebuilt UI to feed feedback to agents, adapt behavior, and personalize across sessions.

Natural Language Interfaces: Users can describe tasks or ask questions in plain language, and the system converts them into agent actions or workflows.

Low code AI products we are exploring

• Langflow
• Flowise, 
• Microsoft Power Platform (including the 365 platform)
• Zapier
• make
• Crewai
• OpenAI

Claudia has convinced me of an ambitious plan of conducting a session on “Design thinking and Product management for low code AI agents” Hence, this exploration

Welcome thoughts 

If you want to meet us see the  Oxford AI summit is based on the theme of Autonomous AI agents - this event enables you to get a certificate from Oxford in Autonomous AI agents.