How we use AI – Yellow Dot

HOW WE USE AI.

Trust and understanding are core aspects of coaching. By being transparent about how our service works we hope that this helps users make the most of their experience.

We’ve taken our personal coaching knowledge and experience – as both a coach and a client – and brought this to life through Artificial Intelligence (AI). Our digital coach, “Mai”, is the counterpart to your coaching sessions and allows us to make digital coaching available to everyone.

How it works

Mai is made-up of multiple AI models doing different jobs. These models combine to function as a digital coach and facilitate the coaching experience. We use AI in three core ways:

Operating as the coach Mai and specifically generating their responses
Enabling Mai to hear and speak to you, in your preferred language
Safeguarding your interactions

The primary function of our models is to analyse your interactions to generate personalised responses. These are typically in the form of questions or insights that guide the exploration of your development topics. The formulation of these responses is based on a combination of inputs and outputs from multiple Large Language Models (LLMs), Machine Learning (ML) models, and other code.

Listening and creating a space to think are fundamental to the reflective experience that coaching is designed to facilitate. We use state-of-the-art AI models to transcribe when you speak and to give Mai a voice, as well as translate into your chosen language.

what do we mean by…

ARTIFICIAL INTELLIGENCE (AI) is a type of technology that allows computers to perform tasks that normally require human intelligence. These tasks can include understanding speech, recognising images, making decisions, or solving problems. AI is designed to mimic how humans think and act, but it does so using computer algorithms and data. AI works by using data and patterns to make decisions. However, while AI can perform impressive tasks, it does not “think” or “feel” like a human – it is just following patterns and rules based on the data it’s been given.

LARGE LANGUAGE MODEL (LLM) is a type of computer programme designed to understand and generate human-like text. It works by learning from huge amounts of text data, such as books, websites, and articles which it uses to predict what words should come next in a sentence or to answer questions. LLMs don’t think but use patterns they’ve learned from data to give responses that sound natural. It’s not perfect and may sometimes give incorrect or strange answers.

MACHINE LEARNING (ML) is a type of technology that allows computers to learn from experience, just like humans do. Instead of being programmed with specific instructions for every task, ML programmes learn by being exposed to lots of data or examples. The more data they have, the better they get at learning.

Personalising the service to you

Coaching is a highly personal experience. No two individuals have the same goals or needs, and we all develop differently. Your needs will change as the coaching conversation progresses, and as your circumstances change.

To tailor our service, your preferences, current conversation, and previous conversations are provided to our models. This means Mai is able to personalise your conversation and recall your history, so your coaching is a continuous journey.

Making it safe

All of our models, and the interactions between them, are designed, tested and monitored by our coaching experts to ensure the highest quality.

We use AI models to implement safeguarding controls to ensure the quality of the coaching provided. These models check, and where required intervene in, the interactions to ensure the discussion is appropriate and performing to the standards we, and you, expect.

As part of our on-going supervision, we have qualified coaches and other domain-relevant individuals review anonymous interactions. This ensures we identify any issues in the interactions or with Mai.

Trust and making coaching a safe space is very important to us. We take your data privacy and security extremely seriously. For more information, please see our Privacy Notice.

Training Mai to meet future expectations

Our goal for Mai is that they can meet your needs and expectations as you develop over time. Effectively, Mai needs to grow as a coach as you grow as an individual or organisation. Similar to people, to obtain the highest performance levels we need to continuously train and up-skill our models, not only on the theory and knowledge of coaching but in its real-world practice. Mastery of coaching is primarily determined by the number of hours a coach has coached. Accreditation bodies such as the International Coaching Federation (ICF) and European Mentoring & Coaching Council (EMCC) require 2,500 hours of coaching experience. We are building the equivalent experience for Mai.

Some of our models are trained with publicly available information and some are trained by a catalogue of relevant materials. This training includes using anonymised interactions from users and other coaching sessions. Please see our Privacy Notice for more information.

You remain in control

The outcomes from coaching, whether decisions or actions, rest with you rather than the coach or AI. Mai does not provide advice nor direction. As with any machine, or person, Mai can sometimes make mistakes. If you’re not happy with a question or response from Mai, or simply don’t understand what they mean, just tell them and they will adjust accordingly.