Generative AI (GenAI) is an enabler for more automation — and is now getting embedded in many software vendors’ platforms with different names and labels such as AI agents, AI-led micro-automations, and autonomous workplace assistants (AWAs). All of these are somewhat overlapping terms for agentic automation, which is starting to sprout as generative AI promises to elevate automation to unknown-value heights.

Forrester is explicitly encouraging leveraging genAI in automation use cases of all sorts, but the recent emergence of agent-based automation also raises a concern: If robotic process automation (RPA) bots are getting replaced by AI agents or if AI agents are closing the next automation gap, this will lead to an unmanageable number of AI agents with overlapping functionalities, poor governance, and high run and maintenance costs. This development highlights many of the other challenges and risks that we have seen from scaled-up RPA bot environments. Let’s not repeat that! Instead, automation builders should:

  • Explore the technology’s opportunities, risks, and adoption challenges. This is the bedrock capability required for any further genAI adoption. It applies to any emerging technology. We have written a lot about emerging technology experimentation.
  • Identify the business problem behind the genAI automation use case. Very often, genAI use cases seem so obvious. When taking a closer look, however, building an AI agent to address an issue is more like a Band-Aid than a sustainable, long-term solution — similar to some RPA bots in the past. So our recommendation is to first identify and understand the problem before deciding if the best solution is an AI agent, an RPA bot, an API, or a better process.
  • Challenge the underlying process before improving a bad process. Autogenerating an email to a client or having an AWA search your product catalog for the best product match sounds like value-adding cases for AI. But wait a minute! What if the reason for still sending emails to clients and looking up products in product catalogs are due to badly connected application systems or information still sitting in documents instead of digital records? Improving the process first to understand if and where there might be a promising case for agentic automation will not only save on costs but will quite likely improve the customer and/or employee experience, as well.
  • Embed AI agents in orchestrated processes like any other automation technology. Treat genAI as another component in your automation toolbox that you use to orchestrate your processes. Currently, we are observing three patterns of how agentic automation is used in productive environments: 1) AI agents used in isolation or an AI agent replacing existing automation, mainly an RPA bot; 2) RPA bots calling AI agents, and vice versa, along an automated process; and 3) several AI agents orchestrated along a process.