Rewriting processes executed by Humans, Robots and AI Agents

Rewriting a process for a world in which humans, software robots and AI agents all do part of the work means deciding, task by task, who is best placed to execute each step and what rules govern the result. The methods that managers have relied on for decades, such as Lean, Six Sigma and SIPOC, were built on the assumption that every task belongs to a person. That assumption no longer holds, and the leaders who redesign their processes first will set the pace for their industries.

Imagine a head of quality at a financial institution who opens the process manual for credit approval. It is carefully written, certified and current as of last year. Yet on a normal day a software robot already copies the applicant's data between three systems, an AI model already pre-screens the documents, and a pair of analysts handle only the exceptions. The manual describes none of this. When an auditor asks who is accountable for a wrong decision taken at two in the morning, the honest answer is that the document does not know. The process exists in practice, but it has not been designed, and what has not been designed cannot be governed.

This scene is becoming common, and the trend is accelerating. Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value and inadequate risk controls, and expects at least 15% of routine work decisions to be made autonomously by agentic AI by 2028, up from none in 2024. Both sides of that forecast point to the same need. Companies will give more work to machines, and many will fail to do it well unless the underlying processes are redesigned with clarity. For readers who want to understand the technology itself, our article on what agentic AI is is a good starting point.

This article addresses the new context in a direct and professional way, aimed at managers and leaders of process and quality areas. The central question is whether they are truly updating and redefining their processes for this landscape, identifying clearly which tasks should be performed by humans, which by robots and which by AI agents. Many quality and process departments still limit themselves to documenting procedures in traditional formats, such as printed diagrams, manuals or static files, and overlook the need for standard languages such as BPMN (Business Process Model and Notation) and for platforms such as Appian, UiPath or Interfacing, which support comprehensive process automation and orchestration.

Process managers and leaders should also position themselves as agents of change, the first to identify automation opportunities with statements such as "this process should be executed by a robot" or "this task could be handled by artificial intelligence." In many organizations, however, these professionals remain in reactive roles, far from the strategic leadership that the context demands. The effective incorporation of robots and intelligent agents changes not only who executes a task, but also how performance is evaluated, risks are managed, traceability is ensured and quality is guaranteed. With the right tools and mindset, process and quality leaders can go beyond their traditional roles and become drivers of digital and operational transformation. For a foundation in the discipline itself, see what a business process is and how it is modeled.

How did process management get here?

Historically, process management emerged as a response to the need to optimize resources, reduce errors and improve the quality of products and services. The first methodologies and tools, such as Lean Manufacturing and Six Sigma, were developed in a predominantly industrial context, where processes were linear, repetitive and entirely executed by humans. That human-centered approach made it possible to control each task, identify failures directly and improve operations through manual interventions, constant audits and incremental adjustments.

Over time, organizations adopted more structured ways of documenting processes, among them the SIPOC model, which stands for Supplier, Input, Process, Output and Customer. It brought clarity and standardization, and together with the other methods it became a global benchmark for ISO certification and quality assurance, placing process and quality departments among the strategic areas of the company.

The current technological revolution is transforming this traditional view. The emergence of Robotic Process Automation (RPA), generative artificial intelligence and intelligent automation platforms has changed the operational landscape. Many activities that were once performed by people can now be carried out more quickly and accurately by technological agents. This forces quality and process managers to rethink not only how they document or improve processes but how they design and execute them, because the introduction of robots and intelligent agents changes who performs a task and also how it is managed, measured, controlled and optimized.

For this reason the main challenge is not merely to adopt new technologies, but to redefine the structure and logic of processes. Professionals in quality and process management must take a proactive and strategic role, deciding where automation belongs and how artificial intelligence can be used well. Adaptation is not optional. It is a requirement for competing in a digitized and dynamic business environment, and it is a core part of any serious digital transformation.

What happens to Lean, Six Sigma and SIPOC when machines do the work?

Traditional methodologies for process management, such as Lean, Six Sigma, DMAIC and SIPOC, have been essential in reaching high standards of quality, efficiency and continuous improvement. Lean has enabled organizations to identify and eliminate activities that do not add value for the customer. Six Sigma has focused on reducing variability, controlling defects and driving improvement through rigorous statistical analysis. Both have proven effective, but both were developed on the assumption that tasks would be performed by people, which makes them incomplete in an increasingly automated context.

RPA and AI agents can execute repetitive and cognitive tasks with great precision, consistency and speed, and that obliges us to rethink some basic concepts. The notion of value added in Lean, for example, must now consider that a task performed by a person, however efficient it looks, may be better executed by a robot or an algorithm. The statistical precision pursued by Six Sigma gains new dimensions when it is combined with predictive analytics and machine learning models that reduce human error. Waste itself changes form. A person copying data between screens was always a form of motion and over-processing, and a robot that does it faster has not eliminated the waste, it has only hidden it. A sound redesign removes the need for the copy.

The SIPOC model also changes. Traditionally centered on human suppliers, manual inputs and observable outputs, it must evolve to include digital suppliers, automated inputs and outputs generated by algorithms. This new technological dimension changes how processes are documented, managed and audited. The key is to see this change not as a break with the past but as a logical evolution of good practice in process management. Companies that already have a disciplined process culture, with owners, indicators and standard work, are in the best position to absorb robots and agents, because they already know what they want each step to achieve.

Who should execute each step, a person, a robot or an AI agent?

Proper assignment of executors within business process design is fundamental to efficiency, quality and productivity. In the digital era this assignment goes beyond the traditional division of human roles to include four distinct categories, which are human executors, traditional automations, software robots and AI agents.

Human executors remain essential for tasks that require intrinsically human skills such as empathy, critical judgment, creativity, complex decision-making and strategic leadership. They should concentrate on interactions that are complex or emotionally significant, while repetitive or predictable tasks move to other types of executor. Traditional automations are systems configured to perform linear, repetitive tasks with minimal or no human intervention. They integrate existing IT systems through scripts and predefined flows, and they are ideal for processes with strictly defined rules and little variation in context.

Software robots, known as RPA, imitate the actions that people perform on IT systems. They execute repetitive and structured tasks based on clear rules, such as data entry, transaction processing, migration of data between systems and automatic report generation. Notable platforms include UiPath and Automation Anywhere. AI agents go further. They have capabilities such as machine learning, natural language processing and autonomous decision-making in complex contexts, and they are well suited to predictive analytics, advanced customer service assistants, automated validations and support for complex business decisions.

The practical rule is to match the executor to the nature of the work. If the step is rule-based and stable, a traditional automation or a robot is usually enough. If it depends on interpreting language, images or patterns, an agent may be appropriate, provided that its output is validated. If it requires empathy, negotiation, accountability or judgment in unfamiliar situations, it should stay with a person, supported by information prepared by the machines. Choosing correctly improves efficiency, optimizes cost, increases accuracy and frees talent for more strategic functions. To make the choice systematically, managers and analysts can use tools such as the CX Matrix, advanced BPMN and robust platforms like Appian or Interfacing.

Gartner's warning about agent washing is relevant here. Vendors are rebranding existing assistants, RPA tools and chatbots as agentic AI, and Gartner estimates that only about 130 of the thousands of agentic AI vendors are real. A leader who classifies executors carefully, by what each technology actually does, is better protected against buying an agent when a robot would do, or a robot when a simple integration would be enough.

How do you redesign the process in practice?

Once the context and the types of executors are understood, the next step is the implementation of the redesign. The first move is a detailed diagnosis with tools such as the CX Matrix. The diagnosis determines which existing activities should be maintained, which can be automated and which need specialized human intervention. It makes the current state visible and allows a clear projection of the desired future state, with advanced execution technologies integrated.

The second move is to document the new design in a standard notation. The Object Management Group, which publishes the specification, describes BPMN as a notation designed to bridge business and technical teams, precise enough for diagrams to be translated into software process components. Its advantage in this context is that it can represent workflows with several types of executors, defining tasks, decisions, events and responsibilities. A BPMN diagram in which each lane is a person, a robot or an agent is the clearest way to show everyone what the new process is.

The third move is the selection of the technology platforms. Tools such as Appian, UiPath, Interfacing, Automation Anywhere and Bizagi offer environments for the execution, automation and orchestration of hybrid processes, combining human decisions, robotic interventions and tasks driven by AI. The fourth move, and the one that most often decides the outcome, is change management. Those involved must understand their new roles and the value that automation brings, and quality and process managers should lead the transition by communicating expected benefits such as fewer errors, shorter times, higher productivity and a better experience for the user. A role that is not explained becomes a rumor, and a rumor becomes resistance.

The fifth move is continuous monitoring and improvement. Specific metrics, clear indicators and analytical tools allow the performance of the automated process to be evaluated in real time and adjusted when needed, which keeps the process aligned with strategic objectives and generates sustainable value. The process owner who is responsible for these indicators plays the role that the business process manager has long had, which is why many companies are discovering that the role matters more, not less, in an automated organization.

How are governance, risk and quality assured in automated processes?

The implementation of processes with robots, automations and AI agents requires a clear vision of governance, risk management and continuous quality assurance. In a highly automated environment governance takes on new dimensions and demands a proactive approach to emerging risks. It involves clear policies, defined responsibilities and strict control and supervision protocols. Organizations must define who is accountable when a task executed by a robot or an AI fails, and how exceptions that fall outside the automated logic will be handled. That calls for a clear organizational structure and a culture that prioritizes transparency, accountability and agility in responding to incidents.

Risk management includes the systematic identification and evaluation of the risks associated with technology, such as technical failures, algorithmic errors, biases in AI models or security vulnerabilities. Solid contingency plans and rapid recovery strategies are indispensable. Continuous quality assurance must also adapt. Traditional auditing, manual review and performance evaluation should be complemented by predictive analytics, real-time monitoring and automated validations. Modern platforms allow continuous monitoring systems that provide early alerts, so that any deviation or anomaly is identified and addressed promptly.

Full traceability becomes crucial. It means documenting the decision logic of each robot or AI agent and establishing protocols to validate the accuracy and consistency of their outputs. Platforms such as Appian and UiPath offer features for automated traceability and continuous process auditing. A useful test for any executive is to ask whether, for any decision taken last week by a machine, the company can reconstruct what data was used, which rule or model applied, who approved the design and who answers for the result. If it cannot, the process is not under control, however well it performs.

Finally, ongoing training and awareness are fundamental. Quality and process managers should provide targeted training and promote a clear understanding of the strategic and operational impact of automation. The goal is not only operational efficiency but also the trust of the organization, its customers and its regulators in the move to highly digitalized, hybrid operating models.

What is the new role of the quality and process manager?

The role of the quality and process manager must shift from its traditional function. It is no longer enough to document, audit and manage existing processes. Process professionals must become active and strategic agents of change. Their first responsibility is to promote a culture that embraces innovation and technological change, which involves challenging the operational status quo, identifying tasks that could benefit from intelligent automation and communicating those opportunities across the organization.

To do so, the manager needs a deep understanding of emerging technologies and their application in process redesign, including platforms like Appian, UiPath, Interfacing and Automation Anywhere, as well as agile methods and low-code or no-code approaches that allow fast and adaptive change. The role also requires analytical skills to interpret the data that automated systems and AI agents generate, and to use it for continuous improvement. In addition, the manager must lead multidisciplinary teams that bring together experts in technology, operations and customer experience, so that automation decisions reflect real strategic needs.

The manager also has a critical role in education and training, making sure that people at every level understand the purpose, benefits and challenges of automation. Clear and consistent communication mitigates resistance and fosters adoption. Finally, the modern quality manager should contribute to the future operating strategy, working closely with senior leadership so that automated processes align with business objectives and create a sustainable competitive advantage. Those who accept this role move from the back office to the leadership table.

What do real cases show?

The successful implementation of automated and AI-assisted processes is already a reality in several industries. The following cases come from the original analysis and are presented as illustrations of the order of magnitude that organizations report, not as audited results. In the financial sector, a global bank adopted UiPath to automate repetitive processes in credit management and internal audits. The reported result was a 70% reduction in processing time, a significant decrease in human errors and the reassignment of people to more strategic and analytical tasks.

In the insurance industry, a well-known company implemented intelligent agents for predictive analysis and automated claims management. By using AI it reduced the average claim processing time by more than 60%, improved the accuracy of fraud detection and significantly improved the customer experience. In manufacturing, a multinational automotive company used intelligent automation platforms such as Appian and Interfacing to optimize production and logistics processes. The hybrid approach was reported to reduce operating costs by 40% and production times by 50%, while maintaining high quality standards.

Four lessons emerge from these cases. The first is the importance of a precise diagnosis with tools such as the CX Matrix. The second is the need to integrate technology platforms with the processes that already exist. The third is the relevance of clear and proactive leadership from process and quality managers. The fourth is the strategic value of continuous change management and effective communication. Notice that none of these lessons is about the technology itself, which confirms the central message of this article. The advantage belongs to organizations that design their processes well, not to those that simply buy tools.

What this means for the C-level

For executives who want to grow and attract customers, redesigned processes shorten the time between a customer request and a response, a quote or a delivery, and that speed is visible in the market. For those responsible for loyalty and retention, hybrid processes allow people to spend their time on the conversations that matter while machines deal with the routine, which improves both consistency and the quality of human attention. For those accountable for efficiency and risk, the main point is that automation moves risk rather than removing it. Errors that used to happen one at a time now repeat at machine speed, so controls, monitoring and accountability must be designed before the robots go live.

There is also a strategic question about capability. The companies that will lead are not those that run the largest number of automations but those that know how to decide which work belongs to which executor, and how to change that decision as technology matures. That is a management capability, and it can be built. It begins with an inventory of processes, a common language such as BPMN, a small group of process owners with authority to redesign, and a governance forum that reviews the portfolio of robots and agents as seriously as it reviews financial investments.

How can a director start next Monday?

Select one process with high volume and visible pain, such as onboarding, claims, invoicing or customer requests. Ask the people who run it to walk you through it step by step and mark each activity as human, rule-based or interpretive. Draw the result in BPMN with a lane for each type of executor and note where the decisions and exceptions are. Then define for each automated step who owns it, which indicator measures it and what happens when it fails. Run the redesigned flow with a limited volume, measure for a month and adjust. A single process redesigned with discipline teaches the organization more than a dozen pilots launched without a method.

To document the activities, owners and decisions of that first process, the ICX Define Activities and Processes toolkit offers a practical template that a team can complete in a working session.

How ICX helps

ICX Consulting helps organizations rewrite their processes for hybrid operation, starting with the diagnosis and the design of the process and only then selecting the technology. Our business process consulting brings together Lean, SIPOC, BPMN and the CX Matrix, and our AI agents and robotic process automation capabilities help to implement the redesigned flows on platforms such as Appian. If you want to know which of your processes should be performed by people, which by robots and which by agents, request a process diagnostic with an ICX consultant.

Frequently asked questions

What is the difference between an RPA robot and an AI agent?

A robot follows explicit rules to repeat actions that a person would perform on a system, and it behaves the same way every time. An AI agent interprets context, can handle unstructured information and may decide among options, which makes it more flexible but also less predictable. For that reason agents need stronger validation, supervision and traceability.

Do Lean and Six Sigma still matter in an automated company?

Yes, and perhaps more than before. Automating a process with waste or variability multiplies the problem at machine speed. Lean and Six Sigma provide the discipline to simplify and stabilize a process before and after automation, and they need to be extended to consider digital executors.

Why is BPMN recommended for hybrid processes?

BPMN is a standard notation that business and technical teams can both read. It can show tasks, decisions, events and the responsibility of each executor, including robots and agents, and its diagrams can be translated into software process components, which reduces the gap between design and implementation.

Who is accountable when an automated process fails?

Accountability must be assigned by design, not discovered after an incident. Each automated process should have a named owner, defined limits for what the robot or agent may decide, an escalation route for exceptions and records that show how decisions were made.

Processes that are written for people alone will keep working for a while, but they will not carry the company into a future in which machines do a growing share of the work. If you want to design that future on your terms, talk with ICX about the processes that deserve to be rewritten first.