Business Process Automation: Where to Start

What Is Business Process Automation

involves the use of technology to automatically execute and coordinate tasks, data exchanges, and steps in a workflow, thereby reducing manual labor and making the process more efficient and controllable.

How Can Automation Improve a Business Process?

Depending on the problem being addressed, automation can have different effects, which can be categorized into four main areas.

• Operational efficiency: Reduces the time spent on repetitive tasks and manual interventions, freeing up resources that can be dedicated to higher-value activities.
• Scalability: It enables the handling of larger volumes of cases, requests, or transactions without proportionally increasing the workload and required resources.
• Quality and reliability: minimizes rework, data transfer errors, and redundant steps, making the exchange of information between systems and process stages more seamless.
• Visibility and control: It makes it easier to monitor timelines, volumes, and exceptions; identify bottlenecks; and track results over time.

They also represent the starting points for understanding where automation can have a tangible impact.

Preliminary Analysis: Identifying and Understanding Candidate Processes

The first step in implementing process automation is to identify the activities that offer real potential for improvement. The most obvious signs often emerge in day-to-day operations: repetitive tasks, manual transfers between systems, recurring checks, or volumes that require an increasing amount of effort.

Once a potential area for improvement has been identified, it is important to examine it within the context of the entire workflow. Let’s take document classification as an example: Which system does the document come from? What information does it contain? Where should the extracted data be recorded? What task is triggered next? Which cases require verification?

The same logic applies to an email, an administrative task, a CRM update, or order processing. Analyzing the end-to-end workflow helps us understand where to intervene, which dependencies to consider, and whether automation truly resolves the inefficiency or simply shifts it to another point in the workflow. To reconstruct the workflow, it’s helpful to involve those who carry out the process every day, as they are familiar with exceptions and undocumented steps; and, when volumes warrant it, to analyze system logs (process mining) to observe the actual flow, which often differs from what’s described in the procedures. If redundant steps emerge, it’s best to simplify them before automating them: automating an inefficient process simply means making it inefficient more quickly.

Selecting the Process to Start With

Once you have identified the activities with the greatest potential for improvement, you need to determine which ones to prioritize. For an initial project, it is best to choose an activity that is relevant, well-defined, and sufficiently narrow in scope to allow you to assess its feasibility and results. Four considerations can help narrow down the options.

Does the volume justify the effort?
A frequent task can be very time-consuming even if each individual instance takes only a few minutes. Therefore, frequency, time spent, people involved, and rework must all be considered: a task performed only a few times a month remains a low priority even if it is time-consuming on its own, while a short task performed hundreds of times can consume more resources than it appears at first glance.

Is the logic of the workflow clear?
The more understandable the steps, decisions, and exceptions are, the easier it is to identify what to automate and what approach to take. When the workflow is still poorly defined, clearly mapping it out is already a necessary step in the project, even before choosing a technology.

Is the data available?
It is necessary to verify where the necessary information is located, in what format, and how accessible it is. Data quality and availability directly affect feasibility: information that exists but is scattered across different systems or requires manual steps to retrieve can make the project more complex than it appears.

Is it possible to monitor results and exceptions?
For an initial initiative, it is helpful to choose an area where results and any exceptions can be clearly monitored, while maintaining human oversight where necessary—especially during the most delicate or high-impact stages.

A good starting point, therefore, combines operational impact, feasibility, and the ability to monitor progress.

Selecting the Most Appropriate Process Automation Technology

Once the process has been selected, you can define the technological approach that best aligns with the workflow structure, the systems involved, and the level of variability in the activities. The approaches described below are not mutually exclusive; in most projects, they are combined.

• Application integrations, APIs, and orchestration components: These connect systems, data, and applications so they can exchange information and trigger actions in a coordinated manner. They are useful when a process spans multiple tools and manual steps between systems need to be avoided.
• Workflow engines: They manage the sequence of activities in a process by defining rules, approvals, responsibilities, and exceptions. They are particularly useful in workflows involving multiple people or multiple operational phases.
• Robotic Process Automation (RPA): automates repetitive, rule-based tasks by replicating the actions a person would perform on existing applications and interfaces. It can be particularly useful when systems cannot be easily integrated via APIs. However, because it works on interfaces, it is more sensitive to changes in those interfaces and requires maintenance; when API integration is possible, it is generally the more robust choice.
• Artificial Intelligence: Extends automation to tasks that require interpreting, classifying, or analyzing less structured information. In email sorting, for example, a rule can handle messages identified by specific codes or fields; however, when routing depends on the meaning of the request, AI can interpret the content and assist with classification. Since a model’s output is probabilistic, it is good practice to define confidence thresholds: above the threshold, the case proceeds automatically; below it, it is routed for human review.
• AI Agents: systems based on language models that, given a goal, plan and execute multiple steps autonomously, querying applications and data via APIs and dedicated tools. They are well-suited for complex and variable tasks, such as the end-to-end management of a request, but require clear rules regarding what they are permitted to do, traceability of actions, and human oversight.

Once AI becomes a permanent part of operations, it becomes important to design AI systems that are integrated into business processes and connected to the data and applications that drive the workflow.

Integrating Automation into the Workflow

Once the technology has been selected, automation must be incorporated into the existing workflow by designing proper integration between systems and applications. Automation must interface with applications, systems, and processes already in use, avoiding the creation of isolated solutions.

If, for example, a system classifies a request or extracts information from a document, the result must correctly feed into the next activity: updating a management system, triggering a workflow, generating a notification, or initiating a check.
Integration is therefore part of the use case design: input, processing, and output must function as components of the same process.

Governance aspects are also part of the design process: who is responsible for the automation once it goes into production, how the operations performed are tracked, how personal data is processed in compliance with the GDPR, and—for AI components—what requirements arise from the AI Act. Automation without a designated person in charge and without maintenance tends to lose effectiveness over time, as systems, rules, and data change.

Defining KPIs and Measuring Results

Before launching the project, it is necessary to determine what results you want to achieve and which KPIs will be used to measure them, ideally focusing on the areas of benefit identified at the outset: time, operational capacity, quality/reliability, or flow control.

The starting point is a baseline—that is, a reference point that allows for a comparison between the situation before and after. Depending on the case, factors such as processing times, volumes processed, manual effort required, rework, SLA compliance, or the number of cases handled with the same resources may be considered. In projects that use AI, specific metrics can be added, such as classification accuracy, output quality, or the percentage of cases requiring verification.Finally, it is advisable to proceed in phases: a pilot project with a limited scope and success criteria defined in advance allows you to verify results against the baseline and gradually involve staff before extending automation to other volumes or processes.