AI Avatar Coach for managers and key people in companies
This is the Evolmeo coaching platform. The web platform combines text and voice conversation, a video avatar, a knowledge base and goal-tracking tools, so users can open a work or personal topic at the moment it matters to them.
The result is an AI coach available in the browser with nothing to install. It leads the conversation using an experienced coach's methodology, builds on previous sessions and helps users turn reflection into a concrete next step.
Who the platform was built for
Managers and other key people often need to step back from a situation, think a decision through or prepare for a sensitive conversation. An experienced coach doesn't hand them a ready answer. Through questions and dialogue, the coach helps them see options that are easy to miss under everyday pressure.
Good coaches have limited capacity and their time is booked in advance. AI Avatar Coach complements personal coaching with support available between sessions, in the evening, at the weekend and whenever a user needs to work on a topic without delay.
Expert guarantor
Expert oversight of the design and testing was provided by a coach with more than fifteen years of experience with senior management.
When you need a different perspective right now
Decision-making at the top of a company can be lonely. Some topics can't safely be raised in front of the team, and an external coach may not have a free slot. Yet ongoing work on goals, relationships and leadership style is what decides whether a good intention turns into practice.
Topics to work on with the avatar
- Leading people, delegating, giving feedback and preparing for difficult conversations.
- Career decisions, a change of role, personal goals and work habits.
- Team relationships, collaboration, boundaries and social situations at work.
- Preparing the next step in business or professional development.
Limits of the tool
AI Avatar Coach is a tool for development and reflection. It does not replace psychotherapy, crisis intervention or healthcare. If a conversation goes beyond its safely defined scope, the system directs the user to a person or to appropriate professional help.
From methodology to technology
Development didn't start with choosing an avatar. First, together with the expert guarantor, we broke down how he leads a conversation, which questions he uses and how he recognises when to deepen a topic, close it or hand it over to a person. Only then did we turn these practices into scenarios and a knowledge base.
- 1
Mapping the coaching approach
We described the structure of the conversation, from agreeing on a goal and working with context to the summary and a concrete next step.
- 2
Knowledge base
We organised the methodology, exercises, study materials and typical situations so that the AI draws on verified content instead of generic advice from the internet.
- 3
The avatar's tone and personality
We tested how questions are asked, the length of answers, the pace of the voice, the use of silence and the limits of what the avatar should handle on its own.
- 4
Ongoing testing with the coach
The expert guarantor assessed conversations from the perspective of coaching practice and adjusted the logic until the answers felt natural and followed the intended approach.
Human oversight remains part of the product
The knowledge base is continuously maintained, versioned and extended. The team monitors answer quality, user feedback and the situations in which the avatar hands the conversation over. Product development therefore rests not only on the model, but also on content work and regular expert review.
Conversation with real-time response
The application runs in the browser on desktop and mobile. Users can type, speak or switch on the video avatar. Speech is converted to text, the language model works with the relevant part of the knowledge base, and the answer is turned back into the avatar's voice and expression.
The answer is streamed as it is generated. As soon as the model produces the first meaningful part, the system sends it to the voice layer and the avatar render. This shortens the perceived response time, so the conversation doesn't feel like long waits alternating with finished blocks of text.
Choice of personality, shared methodology
Users can choose a male or female avatar, from three men and three women. Each has their own voice, pace and visual expression, but all of them work with the same methodology and safety rules. The choice of personality changes the delivery, not the quality or the limits of the answer.
- Text conversation
- Voice conversation with intonation that adapts to the context
- Real-time video avatar
- Coaching principles, as if led by a live coach
- Memory of previous sessions in the context of every new session
- Tracking the client's progress over time
Platform architecture
- React and Next.js frontend with an interface for mobile and desktop and accessibility support.
- Speech-to-text, a language model with a knowledge base, voice synthesis and real-time avatar rendering.
- An orchestration layer for routing requests, checking permissions and managing the flow of a session.
- A relational database for accounts, sessions, consents and payments, and a separate knowledge layer for finding relevant content.
- Memory of previous sessions, so each new session builds on goals and agreed steps.
- Online payments as a single session, credits or a subscription.
import { useCoach } from './coach';import { Avatar } from './Avatar'; export function CoachSession({ client }) { const { goals, history } = useCoach(client.id); // context from previous sessions const context = history.slice(-3); return ( <Avatar mode="video" voice="cs-CZ" context={context} goals={goals} /> );}From a goal to the next step
Each session has a clear start and end. Users sign in securely, choose a topic and a way of talking, and receive a summary at the end. When they next sign in, they see their goals, previous recommendations and the status of steps in progress.
- 1
Secure sign-in
The account supports multi-factor authentication, passkeys, device management and sign-in history.
- 2
Agreeing on a goal
The user chooses a topic and the expected outcome and, if they wish, notes how they currently feel.
- 3
Choosing how to talk
The conversation can take place by text, by voice or through the video avatar in push-to-talk or hands-free mode.
- 4
Live work on the topic
The avatar listens, asks follow-up questions, summarises and guides the user to a decision, a practice exercise or an actionable plan.
- 5
Summary and follow-up
After the session, the main findings, next steps and tasks are saved. The user can review the plan or have a summary sent by email.
Development is more than chat history
- Goals and a personal development plan with topics and milestones.
- Session history with summaries, tasks and follow-up recommendations.
- Role-play scenarios to prepare for conversations and work situations.
- A library of exercises and materials based on the user's goals.
- Progress overview for individuals and the option to roll the programme out to a team while preserving access rights and privacy.
A sensitive conversation needs clear rules
The platform works with voice, personal data and topics that may be sensitive. Security is therefore not a separate layer added at the end of development. It is part of sign-in, of working with the knowledge base, of storing history and of every decision about how the avatar should respond.
- Multi-factor authentication, passkeys, encrypted transport and separate access roles.
- A credential vault for sensitive sign-in details and an audit log of access and changes.
- Consent management, data export and account deletion directly from the client area.
- Anonymisation of materials before they enter the knowledge base and checks against reproducing identifying passages.
- Transparent AI labelling, clearly described limits and escalation in a crisis or risky situation.
- Control of data residency and processing rules according to the market where the platform is deployed.
GDPR and AI Act
The application complies with GDPR and the AI Act.
AI Avatar Coach combines the availability of a digital service with a methodology built on real coaching practice. Users have space to think a decision through, prepare the next step and build on previous work.
At the same time, the coach or platform operator manages content, answer quality, reporting and safety rules from one place.
For companies, the platform widens access to development support without having to schedule every short consultation. For individuals, it offers a safe and continuous space to work on goals. Personal coaching remains important for topics that require experience, responsibility and human judgement.
Would you like to build your own AI avatar on your methodology and content?
Book a demo at evenir.cz or write to info@evenir.cz.
