In the past two years, AI has undergone a revolutionary transformation.
Let's explore how modern AI technology can transform the practice of coaching and what it means for the future.
Newer developments in AI continue to push the envelope even further. Besides text-based LLMs, there are also multi-modal Gen AIs that can work with and generate images, video, and audio. Newer technologies and advances such as the ability to hold up to a million words of context in short-term memory during a conversation, enable even more use cases.
Today’s state of the art in AI consists of what are called “Frontier Foundation Models,” trained on several trillions of words of human text, which approaches almost everything humans have written since the beginning of time. They are fine-tuned with human feedback to be able to hold conversations and perform various tasks. These foundation models have access to almost all of human knowledge, including images, audio, and video.
To specialize in specific use cases such as assisting in legal, medical, financial, or programming tasks, or coaching for that matter, the foundation models need to be further refined and bolstered. This can be done through fine-tuning with domain-specific data, and techniques such as Reinforcement Learning through Human Feedback (RLHF). The resulting systems can be further enhanced with techniques such as Retrieval Augmented Generation (RAG) and in-context learning.
Gen AI is not without its challenges. Without adequate guardrails, foundation models sometimes suffer from “hallucinations” where the AI makes up things that are not based in fact, and they can, in some situations, exhibit various kinds of bias learned from skews present in the datasets.
One of the purposes of fine-tuning and specialization is to curb hallucination and to reduce or filter out bias. Techniques such as retrieval augmented generation, or RAG, and in-context learning are used along with domain-specific content, guardrails, and extensive quality assurance to protect against hallucinations, bias, and other ways AI might fail.
What to look for in AI-enabled Coaching Technology
Note that an AI-powered coaching platform is not a thin wrapper around a general-purpose AI model like ChatGPT, or a simple extension of work-oriented “co-pilots,” although both of these can provide some out-of-the-box benefits. There are a number of key differences between “co-pilots,” ChatGPT-like models, and a purpose-built AI-powered coaching platform.
While AI-enabled coaching platforms take advantage of Gen AI capabilities in language understanding, reasoning, and access to vast amounts of knowledge, these platforms also have to incorporate domain-specific knowledge, best practices, tools, systems, and software to recreate and even enhance the coaching experience. Some of these additions include:
- Domain-specific knowledge about coaching methods and best practices supplemented with expert-level knowledge/content on topics such as: psychology at work, leadership competencies, managerial skills and proficiencies, organizational behavior, the art and science of feedback, communication, decision-making frameworks, organizational development & effectiveness, team effectiveness, employee engagement, mindfulness and well-being, analysis and interpretation of survey and 360 feedback, action planning for employee survey results, and much more.
- Protection against hallucinations and bias, and guardrails for safe and productive conversations done using techniques such as fine-tuning, RLHF, RAG, and in-context learning, as well as extensive domain-specific QA, testing, and content guardrails.
- Easy ingestion and incorporation of company-specific content and frameworks, from mission, vision, and values to company policies, strategy, culture, goal-setting methodologies, career planning and career tracks, and management philosophies.
- Proactive Agentic AI systems for providing nudges and follow ups, executing on growth and development plans, and delivering micro-learnings.
- Built-in tools for self-assessments, collecting and analyzing feedback, self-reflection, sound-boarding, and socratic learning.
- Integration into flow-of-work enterprise tools such as scheduling, messaging, conferencing, and other work systems for immediate 24/7 access and usage in the flow of work.
- Integration into enterprise learning and development systems (LMS/LXP), HRIS systems, 360 and Survey systems, OKR & goal-setting systems, etc.
- Data analytics for reporting on usage and effectiveness, and aggregated insights from conversations such as key topics and challenges faced by managers, as well as continuously learning from and refining the product. Gen AI needs to be supplemented with conventional analytical systems to produce these insights.
- Considerations for global workforces, localization, and language and cultural differences
- Enterprise-grade security, reliability, confidentiality, and privacy, including logging and auditing.
Learn more about the AI-driven revolution in manager coaching. Read a blog post from Wisq CEO Jim Barnett about Wisq's vision to improve lives, careers, and cultures, beginning at the manager level.