Together we are stronger - Capgemini, agentic AI and human teaming

Despite a general wariness about the introduction of AI agents into the workforce Capgemini Research Institute’s recent report states that in 3 years, organizations expect to have AI agents as members within human-supervised teams. In order for this to be successful, humans have to view AI agents as valuable team members with a clear role to play. But according to the research 61% of organizations report rising employee anxiety about the impact of AI agents on their employment prospects, and over half believe AI agents will displace more jobs than they create. Despite this the study finds that few organizations are prioritizing re-skilling or workforce restructuring.

Popularising the human-agent team 

The acclimatisation to human-agent teams will be incremental but it will be swift. According to the research, in 12 months’ time, over 60% of organizations are expected to have human-agent teams in which AI agents serve as subordinates/tools or enhancements to human capabilities. In one to three years, AI agents are likely to evolve into members within human-supervised teams. Quoted in the report Itai Asseo, Head of Incubation and Brand Strategy (AI Research) at Salesforce, says: 

Human-AI interaction patterns are still taking shape and a seamless way to interact with AI agents is yet to surface. Just as the shift from BlackBerry to iPhone in early days of the smartphone era redefined how we engage with technology, the sooner humans can intuitively interact with AI agents, the faster will be the adoption of agentic AI and its subsequent benefits.

Steven Webb, Capgemini’s UK CTO explains that:

Workforces need to work alongside agents and managers need to be very clear about the task orientation agents are being asked to perform. Individuals need to make cultural changes in terms of understanding the purpose of agents around a team member.

Key hyperscalers such as Microsoft and Google are releasing products to enable data to be ingested by agents. Capgemini works with clients on selecting the platform and guardrails. You need to keep a human in the loop to ensure that data is used in the right context. We need to deconstruct processes in order to orchestrate end-to-end processes. We are starting to do this and need to understand where do agents fit here and how do we start orchestrating them? The Agent2Agent protocol (A2A) is the direction of travel for this in the vendor landscape. SAP is providing its own agents, whereas we will produce custom agents.

At Capgemini we are using agentic AI in software engineering and development and here we are seeing productivity gains. This is the journey we are on, using agents in the context of developing software.

An example that springs to Webb’s mind “is a retail app where it took six months to create a virtual try on space where the neck fits nicely into the avatar. This type of thing can now be done in two weeks.

He continues:

In the innovation space we are using the technology to get clients started in solving business problems. We can get to a Proof of Concept (PoC) very quickly, realising concepts and testing ideas. It is exciting the speed at which ideas can be taken and innovated on.

The particular use cases that Webb is seeing are in customer service where multi agent behaviour is being mapped to move to “next best agent”. And then also supply chain resilience using agents to support decision making when routes are closed to mitigate supply chain disruption.

Re-designing the org chart

Capgemini Research Institute’s survey reveals that 70% of organizations believe that AI agents will necessitate organizational restructuring. However, skill development, and organizational restructuring are not top strategic focuses of those surveyed. The report suggests that organizations should establish a team dedicated to AI resource management to systematically allocate and manage intelligent resources similar to human resources. It says:

As AI agents operate in real environments, it’s critical to manage how they work with humans, teams, and tools to achieve shared outcomes. Since AI tools are reusable and often shared across teams, tight coordination is essential. Companies should build an accessible ecosystem for AI agents, with platforms that track agent capabilities to reduce redundancy, support reuse, and standardize processes. Finally, managing also includes onboarding new agents, retiring outdated ones, and updating tools and databases as needed.

For human staff soft skills such as collaboration, logical reasoning, emotional intelligence, and ethical judgement are deemed crucial to working alongside AI agents.

The study warns that traditional metrics are not fit for purpose to capture the performance of human-agent teams. It states that organizations should consider introducing joint productivity metrics to capture the performance of human-agent teams; separate AI agent evaluation metrics; assessment of how well aligned human and agent roles are; as well as how adaptable employees are as they respond to working with AI agents. It concludes that “a combination of human feedback, agent performance logs, and business KPIs can help create a holistic view of hybrid team performance.”

My take

To manage the human-agent teams that are forecast to emerge in the next couple of years, organizations seem remarkably underprepared. It seems likely we need, as suggested in this report, new AI resource management teams. These presumably need to work closely with human resource management teams and both then report to a new role, possibly the VP Human-Agent Resources. Intriguingly it may well be the system integrators and business services vendors, companies such as Capgemini itself, whose operations are in the eye of this change management storm. The company’s recently announced acquisition of WNS to create a global leader in agentic AI arguably indicates the staffing direction of travel. As the company morphs into human-agent teams, will it heed its own research?