KEY TAKEAWAYS:

• Beyond the rigid, brittle nature of RPA, and the limited, assistive capacities of Gen AI, the next big leap in enterprise productivity will come from the newest entrant in cognitive automation—Agentic AI.
• When applied to business processes in an efficient and secure manner, Agentic AI gives birth to Agentic Business Automation (ABA), where an intelligent, collaborative mesh of AI agents executes complex workflows end-to-end.
• ABA is transformative for a wide range of use-cases across industries and sectors, reordering job functions, economies, and the very nature of work.

The history of enterprise automation is a story of great ambitions and mixed results. Ever since humans started working, they also looked for ways to work more efficiently. And the solutions that were invented each had their own strengths and weaknesses; every new technology has had certain benefits but also inherent limitations and challenges.

The journey of Robotic Process Automation (RPA) began in the late 1990s, as a technology that would assist humans by doing the most repetitive, specialized tasks. These were mostly physical tasks that didn’t need knowledge, understanding, or insight, which RPA could perform through codified rules in specific workflows. Early applications included website scraping, data migration, and specialized sets of actions that would be executed the same way each time.

But with a technology so rule-based, RPA was also brittle and rigid. For these “dumb bots”, the slightest change in process would result in inaccuracies and broken workflows. RPA was thus found to have a widely variable RoI, with as many as 30%-50% RPA projects still failing today.

It was long believed that more cognitive automation would provide better results, as software would deal with natural language, reasoning, judgment, establishing context and possibly the meaning of things to generate insights. But even as late as 2018, technologists thought that cognitive automation was a while away.

Fast forward to 2024, and cognitive automation has arrived, and how. ChatGPT and other Gen AI products, which generate text, images, videos, interact with humans conversationally and even produce ‘insights’, have been among the fastest growing technological adoptions historically (at rates 2-3x the internet or personal computer). But in the enterprise, Gen AI has come with significant challenges and risks.

As we wrote in GenAI vs Agentic AI: From Assistance to Autonomy, Gen AI only assists humans within existing workflows, like customer support, HR, or sales and marketing. It struggles with inaccuracies and hallucinations; data privacy concerns; scalability and deployment issues; and sub-optimal RoI, often from organizations having no clear AI strategy. Given that its adoption is driven by integration into existing tools, mostly in a co-pilot, chatbot, or assistive capacity, the problems that plague traditional enterprise SaaS will only exacerbate. We believe that the real improvement in enterprise productivity lies in the next chapter of cognitive automation. The days of bloated, overpriced enterprise software are numbered—thanks to Agentic AI. Applied to business processes, Agentic AI is creating a new category, one that disrupts everything from how technology operates in the enterprise to how value is realized from it, and how humans and technology work together.

Since we at Ema are close witnesses to this new category, serving marquee customers with Agentic AI to automate their most complex business workflows, we have the unique opportunity to name, shape, and study this technological and societal leap.Welcome to Agentic Business Automation (ABA) — where Agentic AI executes complex business workflows across the enterprise, through a conversational, adaptive layer of intelligent, collaborative AI agents, who iterate and improve on their own output, work with each other and human colleagues, and access the right tools and systems in a secure, efficient fashion to complete tasks end-to-end and achieve tangible outcomes in line with strategic business goals. A revolution in enterprise productivity is underway.

What is Agentic Business Automation (ABA)?

Agentic Business Automation (ABA) is the next chapter in enterprise automation, where Agentic AI is embedded into everyday business processes. ABA involves autonomous, goal-driven software entities owning and executing workflows end-to-end, via a conversational, adaptive, intelligent mesh of AI agents, who work together and with human colleagues to execute tasks.

ABA is custom-trained on enterprise data to produce accurate and specific results, solving for common issues created by Gen AI in the enterprise, especially inaccuracy and hallucinations. These usually stem from a lack of high-quality training data and models that do not understand enterprise context, the latter of which ABA solves for. ABA can also access other tools, internal or external, relevant systems, applications, databases, and APIs, in a secure and efficient manner so as to produce the best results at the lowest costs.

Unlike traditional Robotic Process Automation (RPA), which relies on strict, rule-based instructions, or Gen AI copilots, which remain assistive rather than autonomous and strategic, ABA provides a more cognitive, holistic automation that acts across the enterprise and has continuous improvement baked into its design. Here are the key features that enable this, distinguishing ABA from everything that’s come before:

• Generating and executing workflows end-to-end: Instead of adhering to simple but brittle, rule-based scripts, ABA can spawn and execute its own complex, multi-step workflows to solve problems. For example, ABA in supply chain wouldn’t be limited to reordering stock at certain pre-set triggers; it would generate, execute, and iterate over workflows that span forecasting, procurement, and logistics.
• Autonomy and adaptive planning: ABA can act independently by planning, executing, and adjusting workflows in line with strategic business goals, specific enterprise contexts, and evolving external conditions. This autonomy and real-time adaptation are fundamental shifts from previous technologies, as ABA can make decisions and shift gears, reducing the need for constant human supervision.
• Self-improvement through iteration and feedback: ABA’s capacity to learn from feedback and real-time data, generated by each input-output cycle, enables it to get better with time, just as a human would. A customer support ABA would not only use historical data on what’s worked and what hasn’t, but also real-time sentiment analysis, and continuously update itself to provide better answers.
• Tooling and collaboration: Unlike prior automation embedded in data silos and complexity, ABA can access the right tools and systems securely, working with other agents and humans to refine its output. An HR ABA would involve multiple HR agents synchronizing data across ERP and CRM, working with each other to manage overall HR goals, creating a connected, intelligent automation experience.

ABA in the Enterprise: Transformative Impact and ROI

Because of its capacity for autonomy, execution, iteration, and tooling, ABA can provide a strategic yet hands-on form of automation across industries and functions.

In supply chain management, ABA can use a mesh of agents to continuously track stock levels (Inventory agent), forecast supply by studying historical as well as real-time data (Forecast agent), and route and ship inventory accordingly (Shipment agent). This entire agentic AI system could also work with a “human in the loop” for any critical or strategic decisions. According to Gartner, top supply chain organizations already use AI to optimize these processes at more than twice the rate of low-performing peers.

In finance and accounting, ABA can coordinate workflows across auditing, compliance, fraud detection, and financial analysis and forecasting. These are usually dependent on rule-based systems that only handle static data and conditions, with humans coordinating across them. But Gartner estimates that 90% of finance functions will deploy at least one AI-enabled solution by the year 2026, while only 10% will see headcount reductions. This indicates the potential of cognitive automation to correct for weaknesses and repetition in how humans work, while aiding their strengths in higher-order problem-solving.

In insurance and claims processing, ABA can handle the entire workflow from claim verification and validation to approval and disbursement. Instead of a human coordinating across these tasks, ABA will orchestrate document processing, data analysis, fraud detection, and final approval, autonomously verifying claims and speeding up the process for better customer satisfaction.

According to Mike Gualtieri, Principal Analyst at Forrester Research, “There’s a claims processing person who’s in charge of that process, [but] Agentic AI is where AI becomes in charge of that process. It’s full-on automation, and it’s not far-fetched for tens of thousands of business processes.”

In customer support, ABA’s stories of success already abound. Large companies like MoneyView, TrueLayer, and Envoy Global use Ema’s Customer Support AI Employee to automate as much as 80% of ticket resolutions, even solving for multilingual contexts and cyclical volume fluctuations (eliminating the complexity of seasonal hires), driving brand trust and customer satisfaction for these enterprises.

Ema’s Customer Support AI Employee ingests relevant enterprise-specific context by integrating with existing systems and tools, such as Notion, Zendesk, Freshdesk, or whatever else is in place. Domain-understanding allows Ema to soon operate with the vocabulary of a seasoned technical support agent, while also categorizing tickets that should be handled by humans and proactively assisting them in faster resolution.

In human resources, what traditionally requires human beings and several tools can now potentially involve one ABA HR solution that autonomously screens resumes, conducts initial assessments, follows up with a personalized screening process for those likely to convert, onboards new employees, and tracks their career over time. The top barrier for HR leaders in adopting AI has been integrating with existing systems. But because ABA can be custom-trained in enterprise contexts, it will better adapt to the organization’s goals and needs, without compromising accuracy and efficiency. ABA’s unique features and applications extend beyond these use-cases and functions, with the potential to improve efficiency and ROI in every department and business workflow. It will lower operational costs, increase the accuracy of output, break down data silos, and free up humans to work on more strategic, higher-value tasks.

The Future of Enterprise Automation with ABA

Agentic Business Automation (ABA) is poised to disrupt how companies approach work, and what to expect from their software and employees alike. Instead of simply optimizing limited routine tasks, Agentic AI systems will be integrated in business workflows to handle complex functions autonomously and end-to-end.

You will thus have autonomous AI employees working hand-in-hand with humans, as ABA grows normal in enterprises. Entire job functions and economies will get restructured as the division of labor between people and machines continues to shift towards machines, especially for repetitive tasks. Lower-order, rule-based work like data entry, once core to many positions, will become AI-driven, and more high-skilled labor will emerge and grow in demand. McKinsey estimates that between 30% to 50% of current work can be automated by 2030, predicting large job displacements in the global workforce.

Talent deriving from unique cognitive abilities, emotional and leadership skills will become more sought after. As humans and AI colleagues start to work together, organizations will relook at their internal structure and how work is allocated, creating systems for training and learning in an evolving workplace and world.

Several other key transitions await for companies embarking on this revolution in human productivity. To realize the full potential of ABA, their approach to technology will also have to change, with new questions that demand answers: How best can you source and procure ABA? How do you measure ABA’s ROI? How do you manage safety, compliance, and data security as ABA becomes the norm in enterprises? And how do you navigate the trade-offs between building custom solutions in-house, versus adopting more horizontal, conversational platforms like Ema, which already offer proven Agentic Business Automation as a Service (ABAaaS) to several customers? We are excited to answer these and other questions that impact enterprises and the future of work, as we shape and study the new category of Agentic Business Automation.