How to determine whether now is the right time to implement agentic AI

The emergence of agentic AI: Agentic AI is positioning itself as the next evolution beyond generative AI, promising autonomous decision-making and action-taking capabilities that surpass traditional robotic process automation (RPA).

• Software vendors are increasingly shifting their marketing focus from generative AI to agentic AI, emphasizing its ability to follow instructions, make decisions, and take actions without human intervention.
• Agentic AI is being presented as a more advanced and adaptable solution compared to RPA, with the potential to handle complex, variable tasks that traditional automation cannot manage.

Key distinctions and potential applications: Agentic AI’s adaptive nature and ability to interpret data, predict outcomes, and learn from new information set it apart from traditional RPA systems.

• Paul Chada, co-founder of Doozer AI, likens RPA to a train on tracks, while agentic AI is compared to a self-driving car capable of navigating various routes and situations adaptively.
• Cameron Marsh, senior analyst at Nucleus Research, highlights agentic AI’s ability to handle unexpected data, unlike traditional RPA which often falters in such situations.
• Potential applications for agentic AI include roles such as claims adjusters, loan officers, and case workers, provided the AI has access to necessary data, workflows, and tools.

Current market offerings: Major tech companies are already introducing agentic AI solutions, indicating a growing interest in this technology.

• Salesforce’s Agentforce, Microsoft’s Copilot-based autonomous Agents, ServiceNow’s AI Agents, Google’s Vertex AI Agent Builder, Amazon Bedrock Agents, and IBM’s watsonx Agent Builder are among the current offerings in the market.
• These solutions aim to provide enterprises with tools to implement agentic AI in various business processes.

Reality check on implementation: Despite vendor claims of easy implementation, experts caution that adopting agentic AI is far from straightforward at this stage.

• Gartner analyst Tom Coshow predicts that by 2028, agentic AI will be available in only one-third of enterprise applications, enabling up to 15% of day-to-day work decisions to be made autonomously.
• Martin Bechard, principal consultant at Dev Consult Canada, describes agentic AI as being at the early-adopter stage, with initial offerings still containing flaws.
• Greg Ceccarelli of Tola Capital points out the lack of workflow-specific benchmarks to compare agent and human performance, indicating that the industry is still in its infancy regarding practical implementation.

Challenges in adoption: Implementing agentic AI involves more than simply replacing human decision-makers in existing workflows.

• Dion Hinchcliffe of The Futurum Group suggests that RPA workflows designed for human interaction will likely require significant re-engineering to work with agentic AI.
• Jason Andersen of Moor Insights and Strategy emphasizes the need for careful assessment and exposure of the right services, APIs, data, and controls to ensure agents have the necessary context and tools.
• Anil Clifford, founder of Eden Digital, highlights the need for enterprises to shift their overall approach to automation, given the probabilistic nature of agentic AI compared to traditional deterministic automation.

Development and integration hurdles: While some vendors offer low-code and no-code platforms for agent development, these solutions have limitations.

• Creating complex agents with customized integrations and nuanced decision-making abilities still requires technical expertise in data flows, machine learning model tuning, and API integrations.
• Enterprises with legacy applications face additional challenges due to limited or unavailable connectors, making integration more difficult.
• Shruti Dhumak, a cloud customer engineer at Google, suggests that cloud-native companies may find it easier to adopt agentic AI compared to those with legacy systems.

Strategic considerations for adoption: Experts offer varying perspectives on when and how enterprises should approach agentic AI implementation.

• Some view current spending on agentic AI as a bet on future potential rather than an immediate investment, suggesting that experimentation may be necessary to establish a strategic advantage.
• Others, like Sanjeev Mohan of SanjMo, recommend a wait-and-see approach, advising CIOs to understand the value of specific use cases before committing to implementation.
• A phased or layered adoption strategy is suggested by some analysts, using agentic AI as a complement to existing RPA rather than a replacement.

Weighing the decision: The decision to invest in agentic AI requires careful consideration of various factors.

• CIOs must weigh the costs in terms of money and time against potential benefits in enterprise agility, scalability, and operational efficiency.
• The possibility of RPA vendors incorporating agentic AI features into their existing solutions presents an alternative path for enterprises to explore.

Looking ahead: While agentic AI shows promise, its current state suggests a cautious approach to adoption may be prudent.

• The technology’s potential to revolutionize enterprise automation is clear, but the challenges in implementation and integration indicate that widespread adoption may take time.
• As the technology matures and more real-world applications emerge, enterprises will be better positioned to assess the true value and feasibility of agentic AI in their specific contexts.