Executives: AI is great, and will serve nicely as co-workers to humans.
Workers: Okay, but please tell us how this will work.
New research from Accenture finds a yawning gap between executives’ plans for AI, and the amount of preparation and training workers are getting to use the technology.
While 84% of executives expect genAI- and other AI-powered agents to work alongside humans within three years, just 26% of workers say they have received training on how to collaborate with AI, according to the survey of 14,000 workers and 1,100 executives across 12 countries.
Workers aren’t necessarily worried about AI impinging on their jobs – a large majority, 80%, view AI technologies, including gen AI, as more of an opportunity than a threat.
Credit the accessibility of AI and related digital technologies. “For the first time in history, people have developed accessible technologies that can learn and grow in dialogue with those who use them,” state the Accenture co-authors, led by Karalee Close, Kishore Durg, and Professor Majd Sakr. Today’s AI and genAI systems are capable of reasoning, planning and acting autonomously.
What is needed to make it all work is what the Accenture team calls continuous co-learning, which is a two-way learning loop between the human work and the AI agent. Current genAI assistants can be fonts of information, but only represent an early step toward true co-learning environments.
At this point, co-learning is still only seen in a minority of organizations, the survey shows. Only 11% of organizations are currently equipped to enable effective co-learning.
These forward-looking organizations are seeing tangible results. The survey’s authors define these organizations as those that “lead with curiosity and creativity, incorporate learning as part of the job, hardwire trust, and make gen AI work the way people work.”
They are achieving 4x faster skill development, 2x higher confidence in adapting daily work habits to collaborate with genAI, and are 1.4x more likely to report year-on-year
profitability increases, the survey finds.
Significantly, they also display 8x more trust in leadership.
The Accenture authors provide a working example of AI co-learning in action:
“Co-learning is a deeper relationship in which the individual and the AI continuously adapt to one another. In a call center, for example, co-learning can look like two teammates handling a call together, with a human representative in the lead and an autonomous AI agent listening in the background. As the rep speaks, the AI transcribes the dialogue and surfaces compliant, next-best responses in real time. When the rep edits a phrase, skips a prompt or rates a suggestion, the AI treats that as feedback—retraining its guidance for future calls.”
The Accenture team offers the following recommendations to build such a co-learning organization:
Make genAI work the way people work. The impetus needs to be on improving people’s day-to-day experience with AI. “If people are already struggling with the basics, what happens when AI tools become more complex and more agentic?”
Continuously improve the employee experience to enable effective AI-human collaboration. “Test usability regularly and addressing friction points through dedicated feedback and help mechanisms.”
Anchor genAI efforts in the outcomes that matter: better decisions, and employee and business impact. “Start small, but design for shared value across major business processes, not just siloed wins. Focus on horizontal outcomes that cut across business units and functions.”
Expand AI agent capabilities progressively. “Give employees clarity and autonomy in how they shape and interact with evolving agent workflows. But maintain transparent human oversight and monitor AI systems to be sure they are delivering up-to-date, context-relevant information.”
Transition to adaptive systems that learn from real-world use. This needs to “evolve in alignment with changing employee needs, workflows and business priorities.”
The key to successfully integrating AI into the workforce is enabling two-way learning. As this survey shows, we’re nowhere near that point yet. No matter how automated an organization, it still takes people to make things work.
