Skip to main content
Liverpool City Region Combined Authority Metro Mayor Liverpool City Region

14 Jan 2025

The Impact of Generative Agentic Artificial Intelligence in the Workplace

Dr Alison Lui, Reader in Corporate and Financial Law, Associate Dean Global Engagement at Liverpool John Moores University

Just as many people are getting to grips with generative artificial intelligence (AI), the latest technology that is beginning to be implemented in the workplace is generative agentic AI. Autonomous AI agents not only produce information to answer your questions, but they will also perform tasks similar to human beings, adapt and improve over a period of time. With the right prompts, AI agents can read through realms of data and enter information on software tools and platforms. We are thus seeing a shift from information providers to action agents.

Since AI agents are trained in foundation models, they can handle multiple tasks at the same time. Put simply, foundation models have the benefit of flexibility, because they are trained in large amounts of unstructured data rather than prescribed rules. The other major benefit of agentic systems is that people with no coding knowledge or experience should find them more straightforward to use than traditional agentic systems which relied on rules and codes. This is because agentic systems are built on natural language processing, which is programmed to understand human language in both written and oral forms.

Use of Generative Agentic AI in FinTech

For readers who are wondering why they should care about this nascent technology, generative AI agents can save even more time, boost productivity and efficiency than generative AI. McKinsey’s research shows that generative AI agents can reduce time spent on producing credit risk memos in loan underwriting by up to 20-60%. AI agents can offer more personalised customer service, tailored financial advice as well as enhancing fraud detection. Agents can think and act autonomously in the tasks they have been set to do. They are pilots rather than co-pilots in these specified tasks.

Impact of Generative Agentic AI in the Workplace

Studies on generative agentic AI are currently scarce due to the nascent nature of this technology. Nevertheless, the first piece of research in this area by Brynjolfsson, Li and Raymond reveals some interesting effects of generative AI agents in the workplace. The authors studied the staggered introduction of a generative AI-based conversational assistant using data from 5,179 customer support agents from a Fortune 500 software company. Their results revealed that using generative agentic AI increased productivity of human customer service staff by 14%. Staff are taking less time to deal with individual chats and increasing the number of chats per hour the staff can handle. However, the impact of productivity is different according to how skilled the customer services staff are. Agentic AI had the most significant impact amongst less skilled and less experienced customer services staff, where there was a 34% increase in the number of issues the staff could handle. In contrast, agentic AI had little impact on the experienced and skilled staff.

On the impact of following agentic AI recommendations, the authors saw a convergence in communication patterns between staff of different levels of experience. Less skilled staff were communicating more like their experienced colleagues. The benefits of generative agentic AI recommendations are the more personalised, real-time recommendations that staff can use. In customer service, customers often complain about the robotic nature of AI or the reliance on reading from a script. However, Brynjolfsson, Li and Raymond’s research shows that agentic AI recommendations to customer services staff are improving customers’ sentiment towards the staff. Customers treat staff better and are happier with the service. They are also requesting less to speak to the staff’s supervisor or query the competence of staff.

This raises the interesting question of the relationship between agentic AI and experienced staff. From the perspectives of personal development or productivity, experienced staff do not benefit much from agentic AI. In fact, they need to spend time training AI by sharing their best practices. Staff currently training the AI data are not generally compensated. In customer service, a worker’s productivity is measured by their ability to deal with customers’ queries. Therefore, an experienced customer service worker who spends time sharing best practice and training AI will be less productive and receive a smaller bonus under current pay practices. This raises the question of creating a different incentive model or new job roles such as AI trainers for top workers.

Key Takeaways for Businesses

Generative agentic AI can boost productivity, efficiency and customer satisfaction more than pure generative AI. Generative Agentic AI’s value as a tool can be truly harnessed by strategically implementing it in areas which can assist humans. Following agentic AI recommendations can improve the performance of less skilled and experienced staff, so that the gap between experienced and less experienced staff is narrowed. There is limited benefit to skilled and experienced staff though.

Businesses which are keen to implement generative agentic AI need to consider the pay model and new roles for the top workers who help with data training. Good AI knowledge in prompting, natural language processing and data analytics could propel customer service assistants to generative AI trainers. AI trainers are required in financial services to train data in wealth investment, loan applications, fraud detection…etc. Businesses should therefore consider upskilling their top workers or produce a new pay model that rewards them for training generative AI data. Upskilling can start from digital training and moving to AI Days and workshops on responsible use of AI. Particular efforts to upskill women and ethnic minorities are required, since the highest AI adoption is by educated, white men in financial services.

References

Cookies help us provide our services. By using this website, you accept our privacy policy.