Investment professionals operate in environments characterised by intense performance pressure, continuous market volatility and heightened expectations from clients and regulators. Recent empirical research focusing on financial professionals indicates that excessive workload, tight deadlines, limited managerial support and career uncertainty contribute significantly to work-related stress within the financial services sector (Kutebayev et al., 2024). Research in behavioural finance has highlighted how physiological stress markers and affective states, such as fear and anger, can significantly impair information-search behaviour and exacerbate investment biases (Nofsinger et al., 2021; Wynes, 2021). Such stressors place substantial cognitive and emotional demands on professionals and, if left unmanaged, can impair judgement, decision-making quality and sustained performance.

More broadly, contemporary organisational research consistently demonstrates that prolonged workplace stress is associated with negative outcomes for both individuals and organisations, including burnout, reduced mental health, lower engagement and diminished productivity (Shiri et al., 2023). While moderate levels of stress may have a motivating effect, sustained exposure to high demands without adequate resources or support increases the risk of emotional exhaustion and psychological strain (Taris, 2024). In high-pressure professional contexts such as asset management, these risks are further amplified by performance-based cultures and constant exposure to uncertainty.

Evidence also suggests that workplace stress is often compounded by financial and job-related concerns, with recent studies demonstrating a clear link between financial well-being and burnout (Botha et al., 2024). This is particularly relevant in asset management environments, where performance volatility, remuneration structures, and job insecurity can interact with psychological stressors to undermine overall well-being.

Despite growing awareness of workplace mental health, many organisations continue to rely on traditional support mechanisms that are underutilised because of stigma, confidentiality concerns, time constraints or a lack of perceived relevance (Shiri et al., 2023). As a result, employees may delay seeking support until stress has already escalated. Recent intervention research therefore emphasises the need for proactive, accessible, and integrated approaches to well-being that can be embedded into daily work practices and provide timely support before stress becomes chronic (Taris, 2024). Organisational coaching offers a potential solution.

Coaching is commonly defined as a collaborative, goal-directed developmental process that integrates structured, solution-focused techniques with cognitive-behavioural methods (Grant et al., 2009). Typically applied in non-clinical settings, coaching aims to enhance individual performance, facilitate personal growth, support goal attainment and promote overall well-being (Grant et al., 2009). Extending this definition, Bachkirova et al. (2018) describe coaching as a one-on-one interaction designed to support enduring personal change.

The demand for organisational coaching aimed at enhancing professional performance and personal well-being has increased significantly in recent years, backed by significant evidence of its efficacy (Athanasopoulou & Dopson, 2018). Grant et al. (2009) highlight that coaching promotes personal growth, thereby strengthening individuals, like investment professionals’ capacity to meet both organisational demands and individual aspirations. Coaching could therefore provide a structured framework for investment professionals to manage stress, maintain focus and enhance overall performance.

Central to coaching is goal-setting theory. Goal-setting theory emerged from extensive empirical research spanning nearly four decades (Locke & Latham, 2002). Within this framework, a goal is characterised as the intended outcome or target of an action (Locke & Latham, 2006). Understanding the relationship between goals, performance and well-being is important. Successfully pursuing and attaining goals enhances both professional outcomes and personal psychological health. According to Tóth et al. (2018), the achievement of a valued goal significantly contributes to subjective well-being. They argue that goal attainment is not merely about completing tasks, but about realising meaningful personal aspirations. Within the asset management industry, where investment professionals operate under sustained high-pressure conditions, it is increasingly critical for portfolio managers to pursue structured personal and professional goals, including well-being. The ability to set and achieve clear goals is essential for maintaining resilience and achieving well-being outcomes (Tuckett & Taffler, 2012). Goal-setting theory as operationalised by coaching therefore provides the foundational mechanisms for directing effort, sustaining motivation and fostering persistence (Senko & Hulleman, 2013; Spence, 2007). Coaching plays a crucial role in helping individuals clarify, pursue and attain meaningful goals that support both well-being and performance. For example, Grant et al. (2009) found that coaching interventions enhance resilience and well-being, while Spence (2007) argues that the effectiveness of coaching is closely tied to successful goal achievement.

Despite its known benefits, traditional face-to-face coaching has limitations. Terblanche (2024) notes that while coaching has demonstrated its effectiveness, it often remains inaccessible and prohibitively expensive. The advent of AI coaching could possibly address these challenges.

Artificial intelligence has experienced a resurgence in recent years, catalysed by the announcement of the Fourth Industrial Revolution in 2016 (Schwab, 2024) and the subsequent arrival of generative AI. Ng et al. (2021) define AI as a wide range of technologies, including computer vision, natural language processing, robotics, robotic process automation and virtual agents, which are designed to simulate human cognitive functions. Artificial intelligence has also made its way into the coaching profession (Terblanche, 2020). Terblanche (2024) describes AI coaching as the imitation of human coaching expertise through simulated knowledge systems that mirror how human coaches engage, guide and support clients in their development journeys. Artificial intelligence has the potential to replicate certain elements of human expertise at significantly lower cost and holds the potential to transform the coaching industry (Acemoglu & Restrepo, 2018). Artificial intelligence coaching also offers scalability and substantial cost advantages. Artificial intelligence coaches are considerably less expensive than human coaches and are available around the clock (Terblanche, 2024). With AI coaching emulators in place, organisations will have few reasons not to provide access to a ‘coach’ for all employees (Terblanche, 2024). The exploration of AI coaching technologies highlights their growing potential to support personal development and wellness outcomes. Recent studies by Santhosh et al. (2023) and Terblanche et al. (2022a) have demonstrated that AI chatbots can deliver personalised, scalable and continuous support that encourages proactive health management and goal attainment. While AI coaching is increasingly applied across sectors such as healthcare and banking, as indicated by Chong et al. (2021), there remains a notable research gap in its application within the asset management industry.

The integration of AI into coaching practices introduces a range of ethical and human considerations that must be critically examined to ensure responsible and effective implementation (Passmore et al., 2025). One of the key concerns relates to data privacy and confidentiality, as AI coaching interactions require the collection and processing of highly sensitive personal and psychological information (Passmore et al., 2025). Diller (2024) highlights that the extent to which users can trust automated systems to safeguard this information is central to the ethical adoption of AI in coaching. This issue of trust was confirmed in a study by Van der Merwe and Terblanche (2025) who showed that users’ fear of the potential harmful impact of AI deters them from engaging with an AI coaching chatbot. This all stands in contrast to the traditional norms of confidentiality upheld in human coaching relationships (Diller, 2024).

A further ethical issue involves the potential loss of human empathy, emotional connection and psychological safety (Terblanche et al., 2022b). These qualities are central to effective coaching and are closely tied to the development of rapport and trust within the coaching relationship (Terblanche et al., 2022a). Considering these limitations, scholars increasingly recommend hybrid coaching models. In these models, AI handles between-session support such as reminders, progress tracking and task support (Loughnane et al., 2025). In contrast, human coaches focus on meaning-making, values-based reflection and deeper emotional processing (Stelter, 2007).

In summary, this literature overview showed that investment professionals operate in high-pressure environments where resilience, sustained performance and clearly defined goals are critical (Ramdani et al., 2022). Organisational coaching could help but has limitations such as high cost and 24/7 availability. Artificial intelligence coaching is an alternative and has demonstrated positive outcomes in various settings, including first-time graduate employees (Terblanche & Tau, 2025) and university students (Terblanche et al., 2023). It has been shown to support self-regulation, promote behavioural change and facilitate the achievement of personal and professional goals (Santhosh et al., 2023; Terblanche et al., 2022a). Despite these developments, there is a notable lack of scholarly research on its application within the asset management industry. Little is known about whether AI coaching can support behavioural outcomes or improve goal attainment aimed at well-being specifically in this context, which is the focus of this study.

Purposive sampling was used to invite people working in the South African asset management industry to participate in the study. A total of 10 participants were recruited as shown in Table 1.

After participants completed the AI chatbot intervention, qualitative data were collected through semi-structured interviews. The interviews were guided by an open-ended question schedule that was designed to elicit insights into how investment professionals experienced and engaged with AI chatbot coaching, particularly regarding well-being and goal attainment. The interview guide ensured consistency across participants and provided flexibility to probe deeper into each participant’s responses. This approach allowed for the collection of detailed and meaningful data that aligned with the purpose of the study. There were nine main interview questions in addition to probes to gain further insights where needed. Questions covered aspects such as how participants experienced the chatbot in general, to what extent it helped them with their well-being, what aspects they liked and disliked, and how AI coaching support compared to other well-being support.

Each interview was conducted virtually, using secure video conferencing platforms, such as Microsoft Teams or Zoom, to accommodate participants across various locations. Interviews lasted between 45 min and 60 min, depending on the depth of discussion and participants’ availability. With informed consent, all interviews were audio recorded and transcribed verbatim. Participants were given the opportunity to review their transcripts to ensure accuracy before the data were analysed, which enhanced the trustworthiness of the results. Ethical protocols were maintained throughout the process, including the protection of participants’ confidentiality, secure data storage and the right to withdraw from the study at any time.

Data were analysed using Braun and Clarke’s six-phase thematic analysis framework (Braun & Clarke, 2006, 2024), which was well suited to exploring participants’ experiences of engaging with an AI chatbot coach. The analysis followed an iterative process of familiarisation, coding, categorisation and theme development.

In Phase 1, transcripts were read and re-read multiple times to ensure deep familiarisation with the data. Early analytic notes were recorded to capture initial observations, such as participants reporting increased convenience and reduced feelings of isolation when using the chatbot. Phase 2 involved line-by-line coding across all transcripts, resulting in an initial set of 80 codes. Performing comparison and refinement, overlapping and conceptually similar codes were consolidated into 35 distinct codes that captured shared patterns across participants.

In Phase 3, related codes were grouped into categories reflecting key aspects of participants’ experiences. For example, codes relating to time pressures and intrusive prompts were grouped under ‘Usability Issues’, representing practical barriers to engagement. Phase 4 involved reviewing and integrating related categories into broader themes. Categories such as accessibility, navigation, timing and privacy were combined into the theme ‘Ease of use and accessibility’, reflecting participants’ overall interaction with the chatbot.

In Phase 5, themes were refined, clearly defined and named to ensure internal coherence and distinction. Phase 6 involved integrating the themes into a coherent narrative aligned with the study’s research objective.

To promote research quality, the study followed established qualitative research criteria of credibility, transferability, dependability, confirmability and reliability through rigorous and transparent data collection and analysis procedures (Ahmed, 2024). Consistency was supported through the use of a common interview guide, verbatim transcription and a systematic analytical process (Braun & Clarke, 2006, 2024). An audit trail documenting coding and theme development decisions enhanced transparency and confirmability. Credibility was further strengthened through participants’ validation, whereby participants reviewed transcripts for accuracy prior to analysis (Mero-Jaffe, 2011). Reflexive practice was also employed throughout the study to manage potential researcher bias, including the maintenance of reflective notes documenting assumptions, emotional responses and interpretive decisions during data collection and analysis (Karcher et al., 2024).

Ethical clearance to conduct this study was obtained from all participants through informed consent and the Stellenbosch Business School Research Ethics Committee (Ref. No. 33585).

The first theme, Ease of use and accessibility, captures how participants experienced the AI chatbot’s functionality, structure and practical integration into their working lives. This theme comprises three categories: Accessibility and usability, Usability issues and Privacy and navigation. Together, these categories show how design features both enabled and constrained engagement with AI coaching.

Participants generally described the chatbot as intuitive, accessible and easy to use. Its continuous availability and structured design enabled engagement at times that would not be possible with human coaching. As S1 explained:

Similarly, N1 emphasised that the logical flow of the system reduced cognitive load:

Participants also noted the chatbot’s professional tone as supportive rather than mechanical. C1 described this balance as enabling deeper reflection:

However, participants also identified usability challenges. While the structured format was often helpful, it sometimes became restrictive when participants were under time pressure. J1 described this tension:

Others experienced frustration when the system repeated questions. LU1 explained:

Privacy perceptions were mixed. For some, anonymity enhanced openness; for others, uncertainty about data storage limited disclosure. F1 noted:

In contrast, LU1 viewed the lack of human judgement as freeing:

Navigation limitations also affected engagement. C1 described feeling constrained when the chatbot struggled to shift topics:

Overall, this theme illustrates how ease of use, flexibility and perceived safety shaped engagement. When accessibility aligned with user expectations, the chatbot supported meaningful reflection; when repetition, time pressure or uncertainty intervened, engagement diminished.

The second theme, Reflection, self-awareness and well-being, captures how participants experienced the chatbot as a reflective and supportive presence that encouraged introspection, emotional regulation, and well-being. Four categories underpin this theme: Facilitating reflection and awareness, Emotional and conflict support, Well-being and stress management and Health aid.

Participants consistently described how the chatbot’s reflective prompts encouraged them to pause and examine their thoughts and reactions. L1 explained:

T1 similarly described gaining insight through writing:

For several participants, the chatbot functioned as a mirror, increasing awareness of emotional triggers. N1 reflected:

The chatbot also provided emotional and conflict support. Participants described using it as a private outlet during moments of stress or isolation. T1 noted:

In conflict situations, reflective prompts encouraged reframing rather than rumination:

Participants also highlighted the chatbot’s role in emotional regulation. T1 explained:

In terms of well-being and stress management, participants described reduced overthinking and moments of clarity. As T1 put it:

Several participants experienced the chatbot as a form of health aid that extended beyond work-related reflection. L1 described it as a non-judgemental outlet:

LU1 highlighted its role in supporting self-care habits:

Collectively, this theme shows how the chatbot supported reflection, emotional processing and well-being. While it did not replace human empathy, it created structured moments of awareness and regulation within demanding work contexts.

The third theme, Accountability and goal attainment, reflects how participants experienced the chatbot as a tool that supported consistency, focus and progress towards personal and professional goals. This theme comprises two categories: Goal support and Application focus.

Participants described how reminders and check-ins promoted accountability. M1 explained:

PA1 similarly observed:

Participants also valued the chatbot’s support with prioritisation. J1 explained:

Acknowledgement of progress further reinforced motivation. PA1 reflected:

Beyond task management, participants described how accountability was linked to well-being. LU1 explained:

Overall, this theme demonstrates how the chatbot supported sustained focus and progress while also encouraging a more sustainable approach to performance.

The final theme, Tensions between AI and human coaching, captures how participants compared AI coaching with traditional human coaching and considered its future role. Three categories informed this theme: Comparison with human coaching, Tensions between AI and human coaching and Future adoption.

Participants consistently noted that while the chatbot offered structure and accessibility, it lacked emotional depth. PA1 stated:

LU1 echoed this:

At the same time, participants acknowledged trade-offs between emotional connection and accessibility. J1 explained:

Rather than viewing AI as a replacement, participants described it as complementary. LU1 described it as:

Others identified broader applications. N1 suggested:

T1 expressed willingness to continue using it:

Overall, participants recognised the emotional limitations of AI coaching but valued its accessibility, structure, and continuity. The findings point towards a hybrid model in which AI complements, rather than replaces, human coaching.

This study contributes to the growing body of literature on goal-setting, well-being, behavioural change and coaching by extending these concepts into the emerging domain of AI development. Research has established the effectiveness of human coaching in promoting reflection, goal commitment and well-being. A theoretical contribution lies in the evidence that reflection and self-awareness can be activated through structured, automated questioning. This challenges traditional assumptions that effective reflection depends on interpersonal interaction or emotional exchange (De Haan et al., 2016). The findings of this study demonstrate that objectivity and psychological safety within an AI context can enhance openness and authenticity. This suggests an alternative pathway for self-directed development. Furthermore, this study adds to the literature on well-being and behavioural change by introducing the concept of on-demand, self-regulated coaching, where support is accessed flexibly and integrated into daily work routines. This extends existing models of coaching, which have primarily emphasised scheduled, relational interactions.

This study offers practical insights for organisations, coaches, and technology developers seeking to integrate AI tools into professional well-being and development strategies. In high-stress asset-management environments, the findings indicate that AI chatbot coaching can complement existing well-being and performance initiatives by providing flexible, on-demand support that encourages reflection, accountability and stress regulation. Positioned as an accessible first layer of support, AI coaching can offer continuous guidance alongside or between human coaching engagements.

For asset managers, AI coaching provides a private, low-pressure space to process stress, regulate emotions, and maintain behavioural consistency during periods of market volatility. Reflective prompts and accountability features help manage cognitive overload and sustain focus, supporting healthier work habits and potentially reducing burnout risk. Human resource and development teams may benefit from integrating AI coaching into broader well-being programmes, particularly where time constraints limit access to traditional coaching. Framing AI coaching as a self-directed aid rather than a replacement may enhance engagement and normalise reflective practices.

For AI developers and coaching professionals, the findings highlight the importance of adaptive dialogue, contextual sensitivity, and reduced repetition to sustain user engagement. Finally, for human coaches, the study demonstrates how AI tools can act as effective between-session support, enabling sessions to focus more deeply on meaning, values, and long-term development. Together, AI and human coaching may extend access to support while enhancing overall coaching effectiveness.

This study has several limitations that need to be taken into account when interpreting the findings. Firstly, the study was qualitative in nature and based on a small sample of investment professionals within the South African asset management industry. While this design allowed for rich, in-depth exploration of experiences, it limits the ability to generalise findings to the broader population of financial professionals or other industries. Secondly, participant self-selection may have introduced bias. Those who volunteered to engage with the AI chatbot and take part in interviews may have been more open to technology or reflective practices than the average investment professional, potentially influencing the range of perspectives captured. Thirdly, the duration of the intervention was relatively short. Participants interacted with the chatbot over a limited period, which may not have been sufficient to observe sustained behavioural or well-being changes over time. Longer-term studies could provide deeper insights into how engagement patterns evolve and whether the benefits persist. Despite these limitations, this pilot study provides a meaningful foundation for understanding how AI coaching could support well-being and goal attainment in high-performance environments. The insights generated offer valuable directions for both future research and practice.

Future studies could extend this research by including a larger and more diverse sample of investment professionals across different geographic regions and financial sectors. Expanding the sample size and context would allow for broader insight from the findings and reveal whether similar themes emerge in other high-performance environments. Future studies could combine interviews with basic quantitative measures, such as stress levels or usage patterns, to provide a fuller picture of how AI coaching supports individuals over time. Following participants across a longer period would also help to see whether initial improvements in engagement or behaviour are maintained in the long run.