Getting Support

Choosing the Right AI Solution

A Simple Guide for Non-Technical Users

Companion, Assistant & Agent

When exploring AI solutions, it’s important to understand the different roles AI can play. Let’s first outline key distinctions between an AI companion, an AI assistant, and an AI agent:

1. AI Companion: Any generic language models (aka the “free” version of copilot, but also other commercial solutions like ChatGPT or Gemini..)  that can answer questions based on the information that was used to train the model and information that can be sourced publicly in the web

2. AI Assistant: An improved companion that can respond not only from the web but also based on personal information available to a user through his own email, teams, SharePoint and OneDrive document. This is precisely what “Copilot for Business” is. Assistant operates in line with data confidentiality concerns and are tied to IOM subscription to Microsoft.

3. AI Agent: An AI agent is a more advanced and specialized tool that can perform complex tasks and workflows. It can interact with other systems, handle data processing, and execute specific functions based on specific instructions and learning patterns.   Agents are usually created and developed based on specific requirements.

When to use what?

When it comes to implementing AI solutions, it can be overwhelming to decide which path to take. This guide will help you understand the key questions to ask and the best options available, using a simple flowchart to illustrate the decision-making process.  

Before diving into the specifics, it’s important to ask yourself a few fundamental questions about your needs and goals. These questions will guide you to the most suitable AI solution for your situation.

1. Do You Need a Highly Customized Solution?  Imagine you run a case management unit and need an AI that can handle very specific tasks, like understanding unique migrant queries and providing tailored responses.

2. Will the AI Agent Need to Handle Complex Workflows?  Consider a scenario where your AI needs to manage complex workflows, such as creating standard documents or interacting with various systems to read and write data.

3. Is Scalability a Concern for Organization-Wide Use?  Suppose your AI solution needs to scale up to support IOM beyond your mission or unit, handling a large number of interactions simultaneously.

4. Do You Need to Access User Interaction Data for Continuous Improvement?  Think about whether you need to monitor how users interact with your AI to make ongoing improvements based on their feedback.

5. Are Advanced Security and Compliance Features Necessary?  Consider if your application deals with sensitive data or processes that require advanced security and compliance measures.

Getting Support

To summarize, when creating an AI solutions, you can therefore consider the following approaches:

  2. Use Advanced Prompt Engineering directly with your own in Copilot: Ideal for straightforward tasks that don’t require extensive customization or complex workflows. If you need support on prompt engineering, either post your question here: xxx or reach out to support-ict@iom.int

  4. Use Copilot Studio, a low-code platform to create simple agent: Suitable for more customized solutions that need some level of complexity and scalability but don’t require the highest level of security and compliance. If you need support on prompt engineering, either post your question here: xxx or reach out to xxx

  6. Use Azure AI, an advanced system for robust AI implementation: Appropriate for highly customized, complex solutions that need to scale organization-wide and require advanced security and compliance features. If you need support on prompt engineering, either post your question here: xxx or reach out to xxx

Azure-AI based solutions will constitute our highest standard level for AI implementation, supported centrally and aiming for the continuous improvement through user feedback loop, through a consistent user experience all available through http://ai.iom.int.

The ICT division plans to use multiple AI tools together, each specializing in different tasks, to make the system more efficient, aka a multi-agent framework based on Azure AI & a library called Autogen. In simple terms, a multi-agent framework involves using multiple AI agents that work together to perform different tasks. Each agent is specialized in a specific area, and can be orchestrated to collaborate to achieve a common goal. This will come with the following advantage: Different agents will be able to handle different tasks simultaneously, making the overall system more efficient. In addition, each agent can be tailored to excel in a specific function and continuously monitored and evaluated in thar process, improving the quality of the output. Last, such system can easily scale by adding more agents as needed, without overloading a single agent. This allows for a modular approach where agents can be updated or replaced independently, ensuring the system remains adaptable to new requirements.

Join the discussion

The ICT Division two dedicated “Center of Excellence” groups to promote a collaborative approach in the creation of IOM AI Capacity. If you are interested to contribute, please join: 

  • The AI Champ Group to ideate around potential AI Agents in Azure AI.

  • The AI Dev Group to develop AI Agents within the Azure AI Framework.