Conversing with one “difficult, abrasive” male colleague constantly interrupting made Vrnda Boykin sweat. “I was angry,” she says. Despite the discomfort, the senior programme manager at software company HubSpot repeatedly returned to the discussion.
For her interlocutor was not human but a chatbot — or “coachbot” — she had asked to act like a vexatious co-worker so she could rehearse prickly workplace conversations.
“I was able to practise scaling back my responses, so it was constructive”, Boykin says. The rehearsals allowed her to “pause and not be baited, I could think it through rather than saying, ‘That’s rude.’” It provided an opportunity to make mistakes “in private” and seemed far preferable to ending “in [an] HR dispute”.
Aimy, the chatbot, which is represented by a choice of avatars, and options to communicate via text and speech, was created by CoachHub, a coaching platform, and is part of a pilot at HubSpot. It repeated Boykin’s words back to her, suggesting alternatives to encourage her to work together with her combative colleague.
Aimy, which will go on general release this summer, is part of a new breed of coachbots, including Valence’s Nadia and Ezra’s Cai. They draw on generative artificial intelligence, offering a cheaper version of a service that has been largely reserved for senior executives.
AI-powered virtual coaches can provide advice on salary negotiations, suggest how a user’s habits might shape their work, or role-play conversations. They are just one way AI is being used in this area; others include supporting human coaches by summarising sessions, matching clients and coaches, or keeping clients accountable to their goals.
Valence markets itself as giving managers an always-on coach at “2 per cent of the traditional cost”. Parker Mitchell, chief executive, says demand is growing across industries from finance to travel, especially from “frontline leaders”. CoachHub’s chief executive, Matti Niebelschütz, says these one-on-one automated mentors are “democratising professional growth”.
But the rise of AI coaching also raises ethical concerns, and some in the industry question its purpose. Is it a marketing exercise to stimulate demand for human coaching, a helpful addition, or cannibalising the market for real-life professionals?
Such criticisms were crystallised in a recent article by Tatiana Bachkirova, professor of coaching psychology at Oxford Brookes University. She believes the “AI coaching invasion” represents an “enshittification” of the trade, and describes cheaper services as “an ersatz” not good enough for those at the top of the ladder. “Organisations’ developmental offerings to their employees are degraded, whilst keeping human coaching only for more important customers.”
Coaching is a wide category, ranging from teaching basic skills to expensive high-end executive development. Since the pandemic, many digital versions have become more accessible, with new platforms like CoachHub and BetterUp matching human coaches with workers online. According to Growth Markets Reports, the size of the global online coaching market was expected to grow to $6.79bn by 2031, from $2.19bn in 2022.
More recently, generative AI capabilities have made it possible for chatbots to meet some of the demand. Unlike generic AI such as ChatGPT, specialised coachbots are trained on specific company policies, tried-and-tested approaches to workplace challenges and information about a user’s experience or work stored in company systems. CoachHub says its service draws on structured coaching methodologies.
Carol Braddick, a coach who researches the industry, says chatbots are typically offered to middle managers, team leaders and individuals. Delta, Experian, Kraft Heinz, WPP and Novartis are among the corporate clients using Valence’s Nadia, to take one example.
Even the most enthusiastic advocates, however, say coachbots cannot replace high-level development for the most senior staff. A virtual coach may break a goal into achievable steps, but will not steer a client out of a full-blown professional existential crisis, for example. Bachkirova says goal setting — “the most important part of real coaching” where “projects are gradually identified by better understanding needs” — should not be left to a bot.
One advantage of a chatbot is that it is available all day and night. Some clients are more open with AI because they feel less judged than by a human who may be influenced by accents or appearances. For Boykin, a bot felt more objective, unfettered by emotions or biases of humans, who come with the risk that despite “best intentions . . . anything you say could be used against you”. Jordan Hochenbaum, vice-president of AI at BetterUp, says technology is particularly useful in role play. “It’s really difficult even for a trained actor. You can engage without worrying about what your hands look like.”
Nicky Terblanche, associate professor of leadership coaching and research methodology at Stellenbosch Business School, has been “surprised that people describe the chatbots they use as ‘my secret friend’ they tell problems they can’t share in the organisation. They didn’t care that the information was stored somewhere.” Valence’s Mitchell says data privacy is essential. “Coaching is about vulnerability, admitting what we don’t know, and sharing what we must struggle with. People only do that if they believe their AI coach preserves their confidentiality.”
The rise of AI has made some human coaches “nervous” that they will be replaced, says Terblanche. Such anxiety seems unfounded. The most recent report of the International Coaching Federation, a professional body, showed the number of coaches exceeded 100,000 for the first time in 2022, reaching 109,200, a 54 per cent increase on 2019. The biggest rises were in Asia (86 per cent) and the Middle East and Africa (74 per cent). The overall average fee per session rose to $244 an hour, a 9 per cent increase from 2019, with experienced coaches able to charge far higher.
Coaching is an unregulated industry; in theory, anyone can set themselves up as a coach without training, though professional bodies such as ICF offer accreditation. Even if the coach is good, the match might lack chemistry. Using AI to support real-life coaches could raise standards, and provide a more uniform experience, providing the data it is trained on is of sufficient quality.
However, Terblanche says some companies are “overreaching” with promises that are not substantiated. “We have yet to understand for which audiences and contexts AI works,” he says. “We have to understand which AI coaching and frameworks are useful. It’s so early.” Companies’ reluctance to make their algorithms transparent is the most significant hurdle.
One of the biggest dangers is that the technology becomes too embedded in our lives, says Carsten Schermuly, professor of business psychology at SRH Berlin University of Applied Sciences “We know from coaching with a real person that [it] can also produce side effects. One common side effect is that a relationship of dependency develops between coach and client.” If a coaching bot is available 24 hours a day, “I fear that the risk of dependency and addiction is much higher.”