Agentic AI

Lea knows your learning. And its limits.

Lea is the agentic AI built into Masar. An orchestrator reads each request, plans the steps and hands them to specialist agents that work from your courses, people and reports, reaching only what the person asking may reach.

LeaOnline

What should I focus on this week?

Two things are due this week: the “Coaching for performance” lesson in Leading High-Performing Teams (you are 71% through) and the live workshop on Tuesday. After that, you are one course away from your New Manager Pathway certificate.

How Lea works

One orchestrator, a team of specialist agents

Lea is not one model trying to do everything. The orchestrator understands the request, plans the steps and hands each one to the agent built for it.

  1. 01

    You ask

    In your own words and your own language, from any page.

  2. 02

    The orchestrator plans

    It works out what you need and which agents should handle it.

  3. 03

    Specialist agents act

    Each agent knows its area: learning, skills, team, reports, courses.

  4. 04

    Platform tools do the work

    Agents call the platform’s own tools, never the database directly.

  5. 05

    Checked at every step

    Every call runs with the permissions of the person who asked.

One assistant, a different job for each person

A tutor who has read the course

Lea answers from the course a learner is taking, explains a point another way and checks understanding. When someone wants to grow a skill, it builds a plan from real courses and follows the steps with them.

  • Answers from the course content
  • Plans toward a skill, step by step
  • What is due, what is next, how am I doing
Lea answering a learner

Straight answers about the team

Managers ask about their people in plain words: who is overdue, how one person is doing, who is ready for more. Lea reads the same records the manager can open, nothing beyond them.

  • Progress and overdue training for the team
  • One person in detail
  • Review the learning plan a team member built
A manager’s team page

A course designer and a report analyst

Give Lea a short brief and its course design agents draft the course: outline, lessons, tests before and after, a survey and a certificate, ready to refine in Studio. Ask for a figure and the reports agent reads the report source for you.

  • A full course from a brief
  • Figures from the report sources by asking
  • Drafting help in every text editor
The AI course designer

The agents

Ten specialists, each with one job

Every agent declares what it answers, who may use it and which tools it calls. You can switch each one on, off or to read only.

See the AI course designer

Beyond the chat

Agents that act before anyone asks

Lea’s automations run on a schedule, as a system account with the same rules, and reach people through their notifications.

Nudges before deadlines

Reminders for training that is due soon or already overdue.

A weekly brief for managers

Who in the team needs attention, delivered every week.

Bringing people back

A personal nudge for learners who have gone quiet.

Open to your own agents

Lea’s tools are available over the Model Context Protocol (MCP), with each person’s permissions.

You stay in charge

An assistant inside a people system has to be careful by design. These are the rules Lea runs on.

It acts as the person asking

Every lookup runs with that person’s permissions. A learner sees their own records; a manager sees their own team.

You choose the AI provider

Connect OpenAI, Azure OpenAI, Google Gemini or Anthropic Claude with your own key, or a compatible model running inside your own network.

You choose what it may do

Each ability can be switched on or off from the console.

It answers from your data

Figures come from the platform’s own records, not from the model’s memory.

Your language

Ask in the language you work in and get the answer in the same one.

People decide

Agents propose, people confirm. Enrolments, bookings and messages run only on a tap, plans need a manager’s approval, drafted courses an editor, and every action is recorded.

About Lea

Can Lea show someone data they should not see?

No. Lea does not have access of its own. Each question is answered by reading the platform as the person who asked, with exactly their permissions.

Which AI model does it use?

The one you connect. Masar works with OpenAI, Azure OpenAI, Google Gemini and Anthropic Claude, using your own account with that provider.

Does a drafted course go live on its own?

No. The designer produces a draft. An editor reviews it, changes what needs changing in Studio and publishes it.

Is Lea a chatbot or an agent?

An agentic system. An orchestrator plans each request and specialist agents carry it out through the platform’s tools: they look things up, draft plans and courses, and, when the person taps to confirm, enrol, book or send.

Does Lea work without a generative AI model?

Yes. Without a model, Lea still answers the common questions from the platform’s data with set phrasing. Connecting a model adds free-text understanding, planning across agents and drafting.

Can we switch the assistant off?

Yes. The assistant, and each of its abilities, can be switched on or off for your site.

See your path on Masar

A live walkthrough around your own use cases, in your language.