Artificial intelligence (AI) and robotic process automation (RPA) are the power couple of digital transformation and business process management. Just take a look at the numbers.
According to a report from Goldman Sachs, AI could boost global labor productivity by more than 1% per year and could attract investment exceeding $200 billion by 2025. Meanwhile, while active in a far narrower field than all-pervasive AI, the RPA market is set to exceed $14 billion by 2029.
Individually, these two independent technologies have redefined what’s achievable in process excellence within the workplace. Discussions of business process automation can sometimes be drawn into an ‘AI vs RPA’ side-by-side comparison, detailing the pros and cons of each as alternative solutions.
This is a valid, but ultimately limiting approach. Individually, AI and RPA can streamline processes and automate tasks to great effect. But together, the potency of their combined ‘intelligent process automation’ (IPA) to discover and deliver previously hidden value in business processes can be truly transformative. Many organizations have woken up to this fact, driving rapid IPA market growth. It’s expected to be worth around $37 billion by 2030.
This blog post will examine intelligent automation, the interaction between AI and RPA, the benefits they provide each other, and the intelligent process benefits they deliver to businesses. It will also reveal the single factor that will confer process superpowers to this power couple.
First, it’s important to define traditional RPA in order to illustrate the enhancements offered by intelligent automation.
Traditional robotic process automation deploys software robots (or ‘bots’) to perform high volume, repetitive, rules-based tasks common to many business processes or workflows. The tasks targeted for RPA tend to be data-heavy and therefore more liable to human error. They involve few exceptions or variations in processing methodology, and their data is consistently structured. Such tasks include data extraction and transfer, standardized reporting, or website scraping.
By deploying RPA technology these tasks are completed in greater volumes, at greater speeds, and with greater accuracy than is possible with a human worker. This can deliver significant productivity gains, drive down costs, and improve process scalability. Simultaneously, an RPA tool lifts team members out of the repetitive tedium of these routine tasks to focus on higher value work that requires their judgment and expertise.
Read a full account of RPA benefits here.
So traditional RPA has a lot to offer, but it also has clear limitations in scope. Aside from being restricted to relatively simple tasks, there’s another important consideration: RPA bots do not think. Instead they do exactly what they’re told, and only what they’re told.
They will neither think nor interpret information outside of their specific automation parameters. Similarly, the software robots won’t react to changes within the process ecosystem with which they interact unless instructed. This means even a minor process change upstream or downstream of the automation can dilute or derail its impact.
Intelligent process automation combines RPA’s ability to automate simple tasks, processes, or workflows with AI’s judgment and learning capabilities. While AI handles the key process decision points, the bots carry out the routine work needed to execute each step. By folding in decision making, this cognitive automation is capable of tackling far more complex tasks, far faster. An AI algorithm can be taught to reason out exceptions and variations in a process and to determine the appropriate action to take – instructing the software bots accordingly.
AI can be taught…and it can learn. AI’s application of contextual understanding and logical inferences enables fast, rational decision-making. But machine learning (ML) algorithms mean the AI system can learn to make better decisions over time – better decisions aligned with business success metrics. By detecting and addressing process anomalies efficiently, AI-informed intelligent automation ensures continual optimization of RPA bots’ effectiveness.
One of the great strengths that artificial intelligence brings to the process automation table is its ability to extract accurate understanding from unstructured data and inputs. The AI toolbox is crammed with innovative applications to do this, including computer vision, natural language processing (NLP), speech recognition, intelligent document processing (IDP), and optical character recognition (OCR).
So whether it’s a free text form field, an email, business document or even a real-time customer query, AI can pull out and organize the relevant information. This cleaned-up data can then be easily used by RPA bots (or to trigger them) to automate processes.
The number of use cases for intelligent automation is huge, and growing. However, some of the more typical broad applications, include:
Finance / Accounts: AI can read supplier or customer invoices and extract key details like amounts owed, due dates and purchase order numbers. On the accounts payable side, RPA bots then take this structured data to verify purchase orders, calculate payment totals, submit payments for approval, and process approved payments. For accounts receivable, the robots use the information to send automated payment reminders, reconcile incoming payments, and flag late payments to the AI system for collection. An AI-driven intelligent process is also highly effective for preventing non-compliance with strict financial processes. For example, the AI could detect if an invoice is checked and approved by the same employee and flag the anomaly – potentially dispatching an RPA bot to prevent payments being made.
Customer service: An increasingly common scenario sees a chatbot with NLP capabilities interact with customers to gather information and handle regular enquiries. Historically, it would pass complex issues to a human agent who would use relevant information gathered by RPA tools to handle the enquiry efficiently. However, advances in generative AI (GenAI), large language models (LLMs), language processing, and predictive analytics mean that in some cases the customer interaction can be handled by AI directly with near-human authenticity – again, fed with relevant contextual information from RPA robots.
Human Resources: Intelligent automation can streamline many HR processes. With recruitment and onboarding, for example, the AI performs background checks on new hires and surfaces any issues. Meanwhile, bots help onboard new employees by provisioning accounts, populating databases, and preparing orientation materials tailored to each worker.
IT support and security: IT professionals the world over face a perpetual battle to avoid being buried in help desk tickets for simple routine issues such as password requests, access provision and ticket update requests. With intelligent process automation, an AI chatbot can field many IT support requests and diagnose common problems. It can trigger RPA bots to reset passwords, provision access, and update help desk tickets, all while operating within the organization’s IT compliance protocols. IT system outages and malicious attacks are a major threat to business continuity. AI technology can be used as a first line of defense, monitoring system status and user behaviors in real time for any potential problems, anomalous data access, or suspicious activities – providing alerts for human action and triggering security protocols to protect key infrastructure.
As the previous sections have illustrated, artificial intelligence elevates the capabilities of RPA software enormously. With AI on board, the scope of an RPA bot’s effectiveness is multiplied, the bot evolves to something closer to a digital worker capable of decision making and geared to self-improvement.
But it would be wrong to think the benefits only flow from AI to RPA. Here are just some of the examples of how RPA provides key support functionality to make AI work more smoothly:
AI training data: RPA bots can quickly gather, clean, normalize and label training data from multiple systems to fuel AI systems and their decision-making capabilities. This saves a lot of time manually preparing data.
Bridging legacy systems: RPA can integrate legacy systems with newer AI tools that may lack connectors or APIs to access the older technology.
Explainability: One of the hot-button topics in AI circles is black box AI or decision transparency. RPA bots can trace the steps that an AI model took and explain how it arrived at specific conclusions.
Human-in-the-loop (HITL): RPA bots can be programmed to act as a safety net for key decisions being actioned by AI – specifically to flag potentially dubious AI outputs for human review. For instance, if an AI approved a loan to a customer with poor credit history, the rule-obeying RPA software set to review loan applications might flag this as non-compliant (and risky). By requesting human reviews, approvals or exception handling for dubious AI outputs, RPA robots help combine automated and human intelligence.
Monitoring AI performance: RPA bots can track AI system performance, watch for errors or bias creeping in, and flag data issues. AI systems make decisions based on the data they are trained on. Sometimes bugs or bias can creep into the AI’s logic over time if the data it is learning from changes. RPA bots can be programmed to continuously test and track how well the AI is working.
In summary, RPA software can enhance, guide, and monitor the AI technology that ultimately transforms it into an intelligent process automation solution.
The business benefits derived from the combination of AI and RPA are unquestionable. However, to maximize the ROI of an automation investment and unlock all the value opportunities hidden away in an organization’s processes requires a third player in the ensemble – what Celonis calls process intelligence.
Process intelligence combines detailed process mining insights with standardized process knowledge to provide AI with the language and learning materials with which to understand, optimize, and automate end-to-end processes. Celonis supplies this process intelligence through the Process Intelligence Graph.
The Process Intelligence Graph produces a ‘digital twin’ of an organization’s processes across every application, business function, and location within its ecosystem. It provides real-time, data-driven insight into how a business actually works and how processes interact. This is consolidated with process best practices compiled from thousands of Celonis customer deployments.
If an AI implementation is only as strong as the data that feeds it, the Process Intelligence Graph ensures an ideal, evolving data foundation from which an AI system can orchestrate and activate RPA bots. The Celonis system not only pinpoints automation opportunities offering the greatest return against business success metrics, but also the means to measure RPA (or IPA) deployment success.
Going a step further, Celonis’ approach can provide a level of situational awareness that will take process automation to the next level. With this hyper detailed, real-time process insight, an AI can detect and analyze process anomalies and trigger RPA bot countermeasures automatically.
Rudy Kuhn, Lead Transformation Evangelist, described the process: “You’ll see exactly the different systems, you’ll see how the process runs, you’ll see the bots working in between, and you can really orchestrate everything. And if you discover that something is wrong in your process, because we are capable of analyzing the data in real time, so you can immediately trigger a workflow, you can trigger an RPA bot, you can do whatever is necessary to fix the problem we just discovered by looking at the process, by looking at the data, and make sure that we basically heal the process as it happens immediately.”
Individually, AI and RPA have the power to transform business processes, but they’re a more formidable force taken together – with the right process intelligence foundation. Intelligent automation solutions marry the repetitive task expertise of RPA bots with the cognitive skills of AI. This amplifies the strengths of both technologies, resulting in automated processes that are faster, smarter and more accurate.