The era of AI realism is here, with business leaders increasingly looking to realize value in their gen AI investments. To help companies achieve the ROI they desire are two trends: AI agents and AI model customization by industry, according to a recently released study by Deloitte.
“Foundation model improvements – including domain and industry customization – and the promise of AI agents could help overcome inherent challenges and accelerate the creation of business value,” the State of Generative AI in the Enterprise report said. However, “it might be a multiyear journey for some organizations to reach full-scale deployment and achieve the ROI they are looking for.”
AI agents are software systems that can autonomously complete tasks with little or no human intervention. For example, it can make flight reservations or plan a vacation. It is related to generative AI, but the latter is semi-autonomous and was built to create new content based on the patterns it has learned. AI agents often use gen AI as a foundation, but they add reasoning and planning capabilities.
Companies are embracing agentic AI; it is the most sought after capability among emerging gen AI-related technological innovations, the report said. About 26% of respondents said they were already exploring AI agents to a “large or very large extent.”
The two most attractive areas in emerging technologies are agentic AI (52% of respondents) and multiagent systems (45%), a more advanced, complex form of agentic AI. Right behind is multimodal capabilities, in which the machine can understand and analyze images, audio and video, not just text.
Autonomous AI agents will be able to execute tasks given by the user, processing text, video, images and audio, while using various tools and interacting with other AI agents as necessary. They also remember what they’ve done in the past and learn from experience. In a way, they can be considered as an army of digital assistants that can work alongside human managers and employees.
AI agents can be trained to do specific tasks like sales research, helping sales teams gather critical information quickly, for example.
But the report also warned that challenges facing gen AI also confront agentic AI – including regulatory uncertainty, inadequate risk management, data deficiencies, employees lacking AI skills, and others. The authors argued that such challenges are even more important to overcome given the increased complexity of agentic AI systems.
Despite its drawbacks, agentic AI is here to stay. Organizations can best prepare for them by doing the following:
- Develop a strategic road map.
- Identify the tasks and workflows that fit agentic AI systems.
- Delineate specific goals and desired value.
- Map out risks.
- Start with low-risk use cases that use non-critical data, with human oversight.
“These early steps can help test and build the data management, cybersecurity and governance capabilities necessary for safe agentic AI applications. Once your organization is comfortable, it can then progress to applications that use more proprietary data, have access to more tools, and
operate more autonomously,” the report said.
Deloitte surveyed 2,773 director to C-suite level executives across six industries and 14 countries between July and September 2024.
Be First to Comment