Artificial intelligence
Setting up an RPP program
Publiée le November 18, 2025
Publiée le November 18, 2025
A common mistake is to launch an RPA project by selecting a tool without having clarified the objectives. Before looking at the technology, you need to answer the following questions: What are the strategic objectives? Reduce costs, improve customer experience, ensure compliance? Which processes contribute most to these objectives? Processes need to be mapped, and their automatibility (volume, variability, stability) assessed.
The BFSI sector is an instructive example: it accounts for almost 29% of the RPA market. However, automation is advancing faster in branches where management has defined a clear roadmap and articulated precise objectives (reducing loan processing time, reducing the risk of error, improving customer service).
Effective RPA governance is based on :
An executive sponsor: a member of the management committee who sets the vision, arbitrates investments and facilitates communication between business and IT.
A Center of Excellence (CoE): a unit dedicated to automation, comprising RPA developers, data analysts, business experts and IT architects. The CoE defines standards (process modeling, robot nomenclature, security), provides training and supports local teams.
Business referents: for each area (finance, HR, production, etc.), a referent identifies automation opportunities, validates needs and evaluates gains. They work with the CoE to design and test the robots.
A change management framework: communication, training and support for teams to encourage acceptance. Automation should not be perceived as a threat, but as a means of enriching missions.
An impact/feasibility matrix can be used to prioritize processes. Criteria include :
Business impact: number of hours saved, effect on customers, risk reduction.
Process complexity: number of systems involved, variability of cases, exceptions.
Dependencies: planned IS evolution, data availability, current projects.
In the start-up phase, we recommend choosing quick wins: processes that are simple to automate, with a visible impact. For example, automating invoice reconciliation in an accounting department, or generating compliance reports.
A robot generally follows the following steps:
Process capture: document analysis, interviews, task mining. This phase models the flow, variants and exceptions.
Design: writing the functional and technical specification. It describes inputs, outputs, business rules and interactions. It’s important to take exceptions into account right from the start.
Development: implementation of the workflow in the RPA tool, creation of scripts, integration of API or AI calls, variable configuration.
Unit and integration testing: checking the robot’s operation on a test environment, handling simple cases and exceptions.
User test (UAT): validation with the businesses and any adjustments required.
Production launch: planning and deployment via the orchestrator, configuration of accesses and parameters.
Supervision and maintenance: monitoring logs, resolving incidents, optimizing the robot, adapting to changes (new functions, interface upgrades).
To measure the effectiveness of the RPP program, we generally follow :
Automated hours: time saved per year thanks to robots.
Error rate: comparison between manual and automated errors.
Processing times: before/after automation.
Robot availability rate: proportion of time robots are operational.
Return on investment (ROI ): financial gains generated by robots compared with development and maintenance costs.
Underestimating maintenance: interfaces evolve, volumes increase, and business logic can change. You need to anticipate the need for resources to maintain robots, integrate updates and manage versions.
Neglecting security: robots access sensitive data. Separate service accounts need to be defined, rights managed by role, and access tracked (activity logs). AI models must also be secure.
Failure to involve the business: the success of RPA depends on a detailed understanding of processes, and user buy-in. Businesses must co-construct the robots and take ownership of the results.
Limiting RPA to silos: isolated departmental automation limits gains. The ambition must be to cross functional boundaries and orchestrate end-to-end processes.
Building an RPA program is a strategic project that goes far beyond simply installing robots. It’s about rethinking processes, bringing business and IT teams closer together, and instilling a culture of continuous improvement. Feedback shows that automation can free up 20-30% of capacity in organizations (marutitech.com) and significantly improve service quality. However, success depends on sound governance, intelligent prioritization and a long-term vision in which automation becomes a lever for innovation rather than a mere cost-cutting tool.