This article only applies to Bot-type AI Agents using GAIA (and not third-party AI).

What does Agentic AI allow?

Alcmeon AI integrates the latest available technologies: agentic AI is an AI capable of connecting to other systems (APIs), choosing the one to use to meet the customer's needs, asking for missing information before retrieving the data and generating a personalized response.

An agentic AI is even more powerful than that, since it not only responds, but also acts! The possible use cases are numerous: where is my order, product returns, loyalty, shopping assistant, booking a time slot in a calendar, etc.

What is an AI Skill?

An AI Skill is the ability for an AI Agent to obtain information from your CRM (or any other of your systems), or conversely to "push" data to update your information systems.

For example, if you want your AI agent to be able to suggest time slots to the client in a calendar and, once a slot is chosen, reserve it and send them an invitation, then you can configure the following AI Skills in Alcmeon:

1. A first skill involves retrieving slots from a calendar
   1. By default, this skill can search for the next 3 available slots
   2. If the client specifies a particular day or time slots, then the service provider can take that into account.
2. A second skill handles the booking of the chosen time slot.
   1. This skill will need to be based on the choice of time slot with the previous skill.
   2. She will also need to ask the client for their email address if she doesn't already have it.

Other examples:

Where is my order?	Shipments abroad	Item returns	Loyalty

How do I add a Skill to my AI Agent?

From the “Settings / Alcmeon AI” section, click on the “AI Agent Skills” tab.

This page allows you to create new Skills and lists those already created with the following information:

• Name : the description of the skill
• Type : a skill can be an API call or the use of an MCP
• AI Agents Using It : Lists the AI Agents who have been given this skill
• Updated on : last update of this skill
• Icon for deleting a skill

Note : In Alcmeon, the same AI Skill can be used by one or more AI Agents.

API Call type skills

Here are the fields available to create an AI skill that calls an API.

Description

This is where you should describe the scenarios in which your AI agent should use this API. Be concise and clear to avoid any misunderstandings.

Example for a skill that retrieves time slots from a calendar:

Allows you to find available meeting slots with the person you’re speaking to.
By default, searche for the next 3 days from now.
If you have already provided available slots and the client specifies new criteria, then make another API call to find new results, as availability can change quickly.
Always present the options in a user-friendly way: “Tuesday 15th at 10:00” instead of the technical format.

Customer data required

Add the necessary data to call the API defined above, which the AI agent will request from the client if it is missing from their request. For each piece of data, you will specify:

• Label : the name of the data, which you can then use in any other field of this Skill in the form {{label}}. For example, for data labeled “startDateTime”, it can be used in the API call in the form “{{startDateTime}}”.
• Type : data type. You can choose from String (character string), Integer (integer), Boolean (true or false) or Float (real number).
• Mandatory : specifies whether this data must be retrieved by the AI, either from the client or by using another API.
• Default : default value if the AI has not retrieved any data.
• Description : You must describe the nature of this data, what it is used for functionally, and its format for example; it is about indicating to the AI any useful information so that it knows how to ask the client for this information if it does not already have it.

Example description for our slot search capability, for the data point "slot search start date":

The time from which available slots should be checked.
Must be expressed in the UTC time zone as an ISO 8601 date string.
You can specify a date without a time, in which case the default value is the start of the day, or specify the time: 2024-08-13 (corresponding to 00:00:00, i.e. the very start of the date); you can also specify the time manually, for example 2024-08-13T09:00:00Z.

You can add as much data as needed to your skill.

API call

In this section, enter the technical parameters for calling your API.

• API URL : Enter the URL of the API endpoint to query. You can use the {{label}} defined in "Required Client Data" as a variable in this field. Example: https://api.aaa.com/{{userid}}
• API Method : Supports all common HTTP methods, namely GET, POST, PUT, DELETE, PACTH, OPTIONS, HEAD
• HTTP Headers : Add as many HTTP headers as needed in the form of a label/value pair. For example, for Bearer token authentication: label = “Authorization” and value = “Bearer mytoken65sdf65sf1sfd165d65sfd6”
• HTTP Body : Optionally configure the HTTP body of your request. You can use the {{label}} defined in "Required Client Data" as a variable in this field.

Data retrieved from the API

The raw data retrieved via the API must be formatted to be understood by the AI agent. Use {{$.myJsonPath}} in your formatted sentence to insert the retrieved data. Example: "Parcel number {{$.parcel.number}} has the following status: {{$.parcel.status}}".

Example for our slot search capability:

{{#$.data}}Here are up to 3 available slots: {{$.data.*[:3].start}}. Please note that the dates are in GMT; you will need to convert them to the time zone for the ‘Belgium’ region (UTC+1)
{{/$.data}}
{{^$.data}}
No availability. You must apologise to the customer and explain that there are no available slots (specifying the dates you searched for), and that the customer can suggest other dates to search for slots again.
{{/$.data}}

This is written using the Mustache script language. Please check below for more information.

Skills related to “Using the MCP”

The Model Context Protocol (MCP) is an open standard protocol designed to connect artificial intelligence (AI) models (including large language models, large reasoning models, and chatbots) to external tools, services, and data sources[1]. It was introduced by the American company Anthropic in late 2024, with the aim of providing a universal interface—compared by some to a “USB-C port” for AI—for integrating contextual data and external functionality into models, regardless of the vendor.

In Alcmeon, you can define a Skill by simply specifying an MCP (Method of Communication Process). This MCP then contains all the information necessary for an AI Agent to use it correctly. It's the equivalent of several API call skills combined into a perfectly specified package.

All you have to do is enter the URL of the MCP you want to use.

Mustache scripting language

All text fields accept the syntax of the “Mustache” scripting language: see the specifications or use their playground to test.

Measuring the use of AI skills

Each use of an AI Skill triggers in Alcmeon a “bot event” which then counts 2 metrics over the period of time you want: the number of uses, and the number of unique customers.

Each bot event is titled as follows: "AI - Skill used - $skillName". For example, for a skill called "CalendarSlots", the associated bot event will be named "AI - Skill used - CalendarSlots".

Go to the “Analytics / Volumes / Bots” section to find the “Bot Events” dashboard. You can then filter at the top of the page to see the event(s) you wish to track.