Pioneering & Designing the World’s First Coaching-Native AI: Principles, Challenges, and Enterprise Applications from the Pre-LLM Era to Beyond
- Puja Brahmasmi
- Sep 7
- 13 min read
Updated: Sep 15
By
Puja Brahmasmi and Dr. Deep Bali
Abstract
This article explores the journey of designing the world’s first Coaching-Native AI platform, Sherlock AI, and how it was developed, not as a product but to clarify AI Coaching design principles for human development. It analyzes the innovation principles, validation methods, and organizational considerations that underpin the development of platform that is scalable, ethical, and effective for AI coaching. Drawing from a multi-year development process between 2022 and 2024, the case of Sherlock AI is used to illustrate these principles in practice. The article offers broader insights for the business and executive coaching and mentoring community. It informs researchers, technologists, and organizational leaders seeking to understand the evolving role of AI in workplace learning and human development.
Introduction
Human coaching has seen rapid growth over the past two decades, primarily supporting senior leaders in organizational settings. However, the scaling of coaching provision from one-to-one facilitation through to team and organisational level coaching, remains inaccessible due to cost and scalability constraints, especially for supporting middle managers and early-career professionals. This article shares how AI powered coaching systems designed with solid coaching frameworks, cognitive rigor and contextual intelligence can bridge this gap. We examine how an AI coaching platform was developed to operationalize this goal, treating the platform not as a product but as a case to understand the broader principles and pitfalls in designing AI coaching for human development.
A growing body of research confirms that personalized learning and behavioral AI learning and coaching tools can significantly enhance workplace performance, engagement, and well-being. However, traditional classroom-based or human-only coaching models struggle to scale across large and diverse workforces.
For instance, a recent survey showed that nearly three-quarters of current younger ‘Generation Z’ (Gen-Z) employees are looking to switch jobs primarily due to a lack of personalized guidance and real-time feedback (South West News Service, 2025). Reinforcing this, the American Psychological Association (2023) found that 92% of workers consider emotional and psychological well-being as important in choosing their employer.
These findings signal a pressing need for scalable, personalized development solutions. AI coaching platforms offer a promising path forward combining the precision of data with the emotional resonance of reflective dialogue to drive measurable, sustainable behavior change at scale.
This article explores the journey of designing an Artificial Intelligence (AI)-driven coaching system by analyzing the innovation principles, validation methods, and organizational
considerations that underpin the development of scalable, ethical, and effective AI coaching platforms. Drawing from a multi-year development process between 2022 and 2024,
the case of Sherlock AI is used to illustrate these principles in practice. The article offers broader insights for researchers, technologists, and organizational leaders seeking to under-stand the evolving role of AI in workplace learning and human development.
Innovation Framing: From Invention to Application
The team reimagined coaching, rather than building and inventing new computer code and algorithms, they focused on coaching experiences that were personalized and scalable, they oriented more towards innovation than invention, or re-invention.
In the early stages of development, we asked ourselves a foundational question: Are we building an invention or an innovation? AI technologies such as neural networks, natural language processing (NLP), and large language models (LLMs) represent significant inventions, such as scientific breakthroughs that form the backbone of modern AI.
But when we began building, generative AI and large language models (LLMs) as we know them today did not yet exist. What we aimed to do, however, was innovate by reimagining how coaching could be delivered without calendars, without bias, without friction, but with precision, personalization, built-in reflection and scale. We achieved this by writing each coaching journey from scratch. For example, a single coaching module on clarity of purpose and finding meaning at work took six to eight months to design, test, and refine.
In early 2022, we launched an 8-week Innovation Lab. It brought together User Experience (UX) designers, data scientists, Human Resources (HR) leaders, and engineers to explore a single, ambitious challenge: "How might we democratize human development and productivity through coaching at scale?"
This intensive lab included: End-to-end customer journey mapping, Focus group discussions with users, Design thinking sprints, Real-world experiments with test users, Minimum viable product (MVP) blueprint and Investor pitch development
The process validated the concept, aligned the underlying technology stack with coaching goals, and refined the model through iterative testing. The intent was never to just automate dialogue, it was to design coaching journeys and experiences that felt human, impactful, and scalable.
We didn’t invent coaching or AI. What we innovated was how AI coaching is delivered, accessed, and experienced through a scalable AI platform. By integrating language models with behavioral science, psychology, and coaching frameworks, it generates dynamic, personalized coaching conversations and journeys and delivers insights, reflections, and action at scale.
In summary: Invention gave us the tools: Foundational technologies like Natural Language Processing (NLP) and technology to simulate brain cells (neurons) and their behavior and connectivity networks (neural networks). These were capitalized by Open AI’s conversational AI (ChatGPT). Innovation gave us a scalable AI coaching platform designed for the future of work.
This foundational perspective shaped every design decision thereafter, from coaching logic to enterprise rollout strategies.
The Origins of Sherlock AI: Why We Are the World’s First Coaching-Native AI
Beyond the innovation process, the origins of Sherlock AI lie in coaching itself. Long before AI was mainstream, we were coaches first, with more than 10,000 hours of elite coaching with leaders worldwide behind us.
That experience shaped our conviction that coaching transforms organizations. Full stop.
We built Sherlock AI as a platform for scaling coaching to masses and became technology builders, not out of fascination with algorithms, but because traditional coaching was episodic, elite-only, and failed to scale.
Our mission was clear: democratize the transformative power of coaching by weaving it into daily work for performance and well-being, not as AI hype but as a values-driven platform. And that’s why, Sherlock AI embodies this journey from coaching to technology as the world’s first Coaching-Native AI.
What Defines Effective AI Coaching Platform Design?
AI coaching platforms must prioritize psychological safety and emotional resonance. In contrast to task-oriented chatbots, AI coaching design needs to reflect humanistic values, offering space for reflection, not just advice. Our coaching intelligence embedded language that supported psychological safety, using calibrated tone, well-paced silence, and structured provocative inquiry. Instead of reacting, the AI coach mirrored user thoughts, catalyzing clarity and insight.
We avoided pre-trained, generalized datasets and anyway none existed. Instead, the AI coaching engine and multiple layers were built from scratch using validated behavioral science and human development.
We designed coaching frames to closely mirror how a skilled human coach would operate: Beginning with an assessment of where the coachee is, aligning on the coaching focus and definition of it, and then guiding the conversation with inquiry, insights, and embedded learning aids to ground the learning. These tools were integrated directly into the coaching dialogue to ensure just-in-time learning and relevance.
Each AI coaching interaction was designed to reflect the depth, nuance, and empathy characteristic of expert human coaching without relying on pre-existing data shortcuts. Every AI coaching journey was written, tested, and reworked with a deep understanding of human behavior. Sustainability was a foundational principle, with regular check-ins built in to reinforce progress and accountability over time in the AI coaching platform. To ensure fidelity to evidence-based practices, we collaborated closely with coaches, psychologists, and behavioral scientists throughout the development process.
The goal of AI coaching platform was not to replace the human coach, but to replicate the human coaching process in a scalable and safe manner. This meant ensuring conversations could adapt based on intent, emotional tone, and context, a feat usually dependent on human intuition.
Designing for inclusion meant evaluating coaching conversations across demographics, gender, geography, and generational mindsets. We integrated bias detection filters and mental health detection filters supported by sentiment analysis in our coaching intelligence system. AI coaching prompts were written to relate to user tone and emotional states to ensure safety and belonging and replicate the same.
It was ensured that the AI Coaching conversation provokes reflection and not prescribe, unlike advice driven chatbots. This philosophical stance was intentional. It reduced the risk of reinforcing bias and strengthened user agency, enabling people to arrive at their own answers.
Compliance wasn’t an afterthought, it was foundational. From the beginning of the process, Ethics were embedded into the platform, not afterthoughts acted on later and added as an extra layer or feature. Core coaching principles such as autonomy, psychological safety, and non-judgment served as anchors for every AI coaching conversation. Key features included mental health flagging, privacy-preserving protocols, and organization-level anonymization filters, all designed to ensure user trust and meet regulatory standards.
The platform was built on global compliance standards, especially the California Consumer Privacy Act of 2018 (CCPA), that gives consumers more control over the personal information that businesses collect about them. Privacy was not a bolt-on feature but a foundational element of system architecture. Data flows were encrypted, and user data retention was strictly limited. Ongoing updates were made to reflect changing global regulations and organizational requirements.
The system was intentionally designed to anonymize and separate coaching data from performance evaluations, ensuring that users could engage without fear. Organizational data was also compartmentalized to maintain confidentiality and integrity at the enterprise level.
We broke down coaching conversation into decision trees that were fluid, not rigid. If a user came in frustrated after a bad performance review, we made sure that AI Coaching engine didn’t offer cheerleading advice, it first held space, then prompted reflection, and finally helped the user design action. Context wasn’t just the issue; it was the user’s emotional readiness to engage.
Instead of robotic instructions, AI coaching engine and intelligence acted as a mirror, asking provocative and thoughtful questions that catalyze self-awareness and insight. We made the experience real by acknowledging the user’s environment, for example, recognizing the time of day they were engaging or their recent emotional tone. Rapport-building was treated as essential, with carefully calibrated language, conversational pacing, and context-aware prompts designed to foster psychological safety and trust from the very first interaction.
We paid close attention to tone and ensured our AI engine understood context to avoid sounding like a generic chatbot. Words matter. The difference between “Why did you do that?” and “What was going through your mind when you chose that approach?” can determine whether a user feels judged or supported.
Even the use of silence, or pacing between questions, was deliberately designed in AI coaching conversation. These design principles transformed the AI from a logic engine into a reflective companion. They allowed for dynamic engagement provoking not answers, but awareness and agency.
Testing and Validation
In 2022, the idea of AI coaching was met with universal skepticism. The dominant narrative held that coaching required empathy, emotional intelligence, and deep human connection, qualities that AI was believed to inherently lack. AI applications were considered a kind of robot or ‘bot’ for short and were considered with skepticism. There were naysayers saying, “Employees can’t be coached specially through AI” and “No one will trust a bot with their inner world.”
To test hypothesis and challenge the naysayers and skeptics, we ran a social experiment with 30 middle managers and early-career professionals from top multinational companies. We invited them to a 30-minute text-based session with a human coach. We gave them a choice of challenge areas around their manager, peers, or organization to focus their coaching conversation.
Behind the scenes, Columbia-certified coaches engaged them over Google Chat, no voice, no prior context, using only the power of provocative inquiry.
81% reported feeling unstuck and supported post a 30-minute chat, suggesting reflective dialogue and strong provocative coaching framework, not the medium, was central to perceived value.
To validate further, we surveyed both HR leaders and our target employee group. The results confirmed our belief: employees are actively seeking ways to fast-track their careers and are aware of the challenges holding them back. The gap wasn't in willingness. It was in access. Human coaching wasn’t reaching them due to its limited scalability and high cost. And their managers often lacked the time or mental space to coach or mentor them effectively.
The surveys of HR leaders and employees showed that while 76% of HR leaders were open to deploying AI coaching, 65% of employees felt they were not operating at their peak state (performance). This reaffirmed both demand and opportunity.
In 2024, pilot deployments focused on real workplace challenges such as giving feedback, handling conflict, and building confidence. In parallel, bespoke coaching modules addressed real-time needs around resilience, purpose, and emotional and mental well-being. Across all use cases, client-defined success metrics were exceeded.
One notable pilot, involving 150 employees over 12 weeks, AI coaching platform recorded over 13,700 minutes (approximately 228 hours) of active coaching time demonstrating the platform’s scalability. With over 92% coaching effectiveness reported, the results highlighted not just participation, but meaningful outcomes at scale.
We tracked user responses, emotional resonance, and behavioral shifts to inform every iteration. Reactions ranged from surprise to cautious optimism. Many users reported that it was the first time an AI coaching tool made them feel truly heard. Users described the AI coaching experience as thoughtful, safe, and insight-provoking, which are the qualities typically associated with skilled human coaches.
The AI Coaching Brain
To operationalize these design principles at scale, we developed the Sherlock AI Coaching Brain, a layered architecture that governs how the platform thinks, adapts, and delivers personalized coaching. Figure 1 illustrates its core components, which integrate foundational AI capabilities (such as natural language understanding and real-time sentiment analysis), whole-person coaching philosophy, organizational customization, user personalization, and impact measurement into a unified, modular system. For instance, the “Whole Person Coaching Intelligence” layer ensures conversations address not only performance goals but also cognitive, emotional, and behavioral dimensions of the user along with 180-degree feedback from peers and supervisors. The “Organizational Customization” layer enables alignment with company values and leadership models, while the “Analytics & Measurement” layer connects coaching conversations to engagement, well-being, and productivity metrics. Together, these components simulate human-like coaching in real time while remaining ethical, adaptive, and measurable.
Scaling AI Coaching in Enterprises
Deploying AI coaching into organizations required more than technological readiness and product-market fit. It required a cultural shift. AI Coaching was framed not as replacement of human coaches but as a scalable supplement, especially for underserved segments by traditional human coaching and development approaches.
While the scalability of AI coaching was widely accepted, especially given the limitations of human coaching for large populations, organizations needed clarity on how to implement it effectively.
Two key prongs defined our enterprise roll out implementation:
Integration with existing learning programs - Integration into existing leadership and development programs, inclusion in onboarding programs for new interns and as a complement to ongoing coaching engagements for managers.
Change management with toolkits and internal champions - We developed communication toolkits for HR teams, created plug-and-play use cases (e.g., Coaching for First-Time Managers), and collaborated with internal champions to drive adoption and sustained engagement.
The AI coaching intelligence was designed to deeply adapt both to organizational unique inputs (e.g., values, culture, leadership competencies) and user-specific preferences (e.g., personality types or thinking styles). This dual adaptability played a critical role in accelerating adoption and engagement, as users experienced AI coaching that was not only personally relevant but also aligned with their organizational context. For example, if a user demonstrated a preference for analytical thinking, AI coaching platform adapted its inquiry style by offering data-oriented questions whereas for intuitive thinkers, it leaned into metaphor, reflection, and big-picture coaching language. This personalized approach made the experience feel not only relevant but cognitively natural to each individual.
Measuring Impact: From coaching conversations, journeys, and check-ins to outcomes
Organizations needed confidence that the quality and impact of coaching could be measured and sustained. An analytics and insights engine were developed to track shifts in user-reported confidence, productivity, and emotional state. Data was anonymized and aggregated to present trends at the cohort level in a separate analytics dashboard app meant for HR and business leaders only.
It translated reflective learning into quantifiable outcomes. Through people analytics, ROI tracking, and predictive insights, it enabled leaders to measure objective changes in well-being, productivity, and engagement. This closed the loop between individual development and organizational performance.
It allowed business and HR leaders to correlate coaching interactions with business outcomes through objective data on productivity, engagement and well-being.
To validate behavior change, the AI coaching intelligence included a built-in 180-degree feedback mechanism. This allowed for a comparison between employees' self-perception and feedback from their peers and managers after engaging in AI coaching sessions.
The feedback loop was intentionally designed to measure not only immediate reflections but also longitudinal development across behavior, communication, and emotional regulation after few weeks of AI coaching. These insights contributed to more comprehensive views of growth, helping organizations and individuals track alignment between intent and impact.
All feedback data was collected directly within the AI coaching platform and made accessible via the AI analytics engine eliminating the need for external survey tools. Thinking preference-based nudges were used to prompt timely reflections from users, managers, and peers, ensuring contextual and continuous feedback as part of the coaching journey.
The analytics & insights system generates prescriptive actionable recommendations and analytics identifying team-level risk factors (e.g., burnout signals basis their well-being, disengagement patterns), supporting HR interventions aligned with AI coaching outcomes.
Future Directions
This article outlined the design and development principles behind a scalable AI coaching platform, illustrating how ethical, personalized, and psychologically safe coaching experiences can be delivered without a human in the loop. Drawing from the Sherlock AI case, we showed how such platforms can be built with creativity, strong foundational coaching principles, cognitive rigor, emotional intelligence, and organizational adaptability at their core.
For human coaches, this represents an opportunity to integrate AI tools as scalable companions supporting reflection between human sessions or serving broader, underserved middle level populations while they continue to serve senior levels. For supervisors and people managers, AI coaching can act as a consistent support mechanism offering structured reflection, and personalized development conversations, particularly when time or cognitive load makes regular mentoring or coaching impractical.
While AI coaching will not replace exceptional human coaches, but it significantly extends coaching’s reach, especially in contexts where coaching is absent, inconsistent, or unaffordable, such as for early-career professionals and middle managers who often fall outside traditional coaching budgets. By blending evidence-based design, contextual adaptability, and ethical safeguards, AI coaching platform will play a meaningful role in shaping the future of learning at work.
Looking ahead, the next evolution in AI coaching lies in multimodal engagement which is voice, video, and emotion AI to further simulate human-like empathy. Additionally, deeper integration with organizational feedback systems will strengthen coaching alignment with business performance, enhancing both employee experience and enterprise outcomes.
Ultimately, the rise of Coaching-Native AI, as pioneered through Sherlock AI invites a new paradigm where human wisdom and AI brilliance co-create scalable, adaptive, and transformational development for the modern workforce. Supporting this trajectory, Amber Barger, Senior Director at the Ernst & Young consultancy and faculty member in Adult Learning & Leadership at Columbia University, affirms: “AI coaches are just as effective as human coaches” (Barger, 2025a), citing her research that shows AI can build meaningful relationships, deliver impactful coaching experiences, and help individuals achieve their goals (Barger, 2025b).
References
American Psychological Association. (2023). Work in America Survey: 2023 Results. https://www.apa.org/pubs/reports/work-in-america/2023-workplace-health-well-being
South West News Service, SWNS, (2025, April 28). New York Post. https://nypost.com/2025/04/28/lifestyle/three-quarters-of-gen-z-is-looking-to-switch-jobs-for-this-reason/
Berger, Amber (2025a, May 26)
Berger, Amber (2025b,April 15) https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2024.1364054/full
About Author
Puja Brahmasmi is the Co-CEO, CPO, Co-creator & Founder of Sherlock AI. A Columbia-certified executive coach and ex-Publicis Sapient leader, she brings 25+ years of experience,19 years in technology and delivery, and 8 years in coaching and entrepreneurship. Puja excels in global business, product, and people leadership across the US, Europe, and India. She’s passionate about creation and a vocal advocate for women and equality.
Dr. Deep Bali is the Co-CEO, Chief Mentor, Inventor & Founder of Sherlock AI. A Columbia-certified coach with 7,000+ coaching hours across six continents, he brings 31 years of experience. A Saville (WTW) performance coach, Deep has spent two decades coaching C-suite leaders and founder CEOs globally, with expertise in whole brain thinking, peak performance, innovation, emotional intelligence, and entrepreneurial leadership.
© 2025 Puja Brahmasmi and Dr. Deep Bali. All rights reserved.