Digital technologies are changing both the world that labour inspectors are called upon to regulate and the way inspection services themselves can operate.

But digital transformation does not begin with technology. It begins with the question: what problem does a labour inspectorate need to solve?

This question shaped Agile Labour Inspection: Digital Transformations for Workplace Compliance, a 60-hour blended learning programme delivered by the ITCILO in June 2026. Following an online phase, 24 labour inspection professionals and related specialists came together in Turin for an intensive residential week, before continuing with applied work in their own institutional contexts.

The course formed the labour inspection pillar of the wider Digital Governance Initiative, which brought together 86 participants across five specialized learning tracks spanning labour inspection, occupational safety and health, social protection, social dialogue and labour migration.

The learning journey was organized around six interconnected dimensions of digital governance: purpose, data, tools, process, methodology and people.

Participants began by defining why an inspectorate might digitalize at all. Only then did they examine the information available to support that objective, the tools that could put those data to use, the processes that would need to change, the methodology governing decisions and, ultimately, the people responsible for making the system work.

An institution, therefore, should not begin by asking which AI application or digital platform it wants to acquire. It should first identify the labour inspection problem that needs to be addressed.

That might mean reaching workers who rarely submit complaints, improving the targeting of undeclared work, reducing the time between notification of a serious occupational accident and an inspection response, improving consistency in the treatment of employers, or reducing administrative work so that inspectors can devote more attention to workers and workplaces.

Labour inspectorates already generate large volumes of information.Inspection histories, complaints, accident reports, detected violations, sanctions and follow-up measures all contain signals about where risks may arise and how effectively inspection systems are responding to them. Other public institutions may hold complementary information through business registers, social security systems, tax databases and occupational safety and health records.

The challenge is often less about producing new data than about making better use of what already exists.

During the online phase, participants examined the data points needed for strategic labour inspection and the importance of using common definitions and structured information. They distinguished between data used for targeting (helping decide where inspection resources should be directed) and data used to assess performance and impact (helping determine whether inspection interventions actually improve compliance and worker protection). 

The learning deliberately started from imperfect institutional realities. Some inspectorates still depend heavily on paper files or spreadsheets. Others operate digital case-management systems but have limited interoperability with other public databases. More advanced administrations are experimenting with predictive analysis, machine learning and integrated risk models. Digital transformation therefore cannot follow the same pathway everywhere.The relevant question is what meaningful next step can be taken from the institution’s existing level of digital maturity.

International practice gave these questions a concrete dimension. Experience from Albania showed how an existing electronic inspection system and a substantial database of previous inspections could be combined with information from other administrative sources to strengthen the targeting of undeclared work. Rather than beginning with an abstract AI project, the approach started with a defined compliance problem and an existing institutional data asset. 

Spain’s Labour and Social Security Inspectorate provided another example. Its Anti-Fraud Tool uses large quantities of administrative information and analytical techniques to support inspection planning and identify entities presenting indicators of possible fraud or non-compliance. Participants also examined predictive, network and sector-specific models and the importance of adapting analysis to geographic and socio-economic contexts. 

Brazil introduced the concept of labour inspection intelligence. Its experience demonstrated how inspection authorities can draw on smart reporting channels, geospatial information, satellite and drone imagery and cross-referenced public information to map supply chains, locate workplaces and support the investigation of serious violations, including forced labour. 

These examples illustrated that digital solutions become useful when they are connected to a specific institutional mandate, reliable information and a clearly defined decision-making process.

Artificial intelligence featured prominently in the course, but it was consistently treated as a supporting capability rather than an autonomous decision-maker. AI can summarize, suggest, identify patterns and support preparation. But, ultimately, the inspector remains responsible for interpreting information, making decisions and exercising statutory authority.

Participants explored machine learning, large and small language models, agentic AI and open-source solutions, while also examining questions of privacy, cybersecurity, transparency, explainability and institutional accountability.

One Brazilian case explored the development of a locally deployed language model designed to summarize administrative appeals arising from labour inspection. The system was conceived specifically to avoid transferring confidential government information to public AI services, illustrating the trade-offs that public institutions must consider between convenience, data protection, computational requirements and specialization. 

The wider Digital Governance Initiative also placed these questions within the responsibilities of public institutions. The keynote discussion on AI for public practitioners emphasized that effective technological adoption depends on integrated strategy, public-sector capability, appropriate governance and the ability to adapt as technologies continue to evolve. 

Participants were also involved in a sequence of AI Clinics and immersive exercises. The clinics began with real institutional problems brought by participants themselves. Instead of presenting a catalogue of technologies, facilitators asked participants to diagnose the inspection challenge first and then assess which data, methods and tools might realistically support a response.

This process continued throughout the week, moving from problem definition to field operations and ultimately to the design of participants’ own digital transformation proposals.

People, the final dimension of the methodology, brought the different strands together.

Digitalization creates a dual challenge for labour inspection.

First, technology can augment the inspector. AI can help summarize previous inspection records and complaints, support the preparation of a tailored checklist, organize information before a workplace visit or assist with drafting questions. Used appropriately, such tools can reduce repetitive administrative work and allow inspectors to devote more time to interviewing, observation, analysis and prevention.

Second, inspectorates must prepare inspectors for a changing world of work. The workplaces they regulate are themselves becoming more digital. Algorithmic management, AI-supported recruitment and dismissal, worker-monitoring systems, smart devices, robotics and digital labour platforms are creating new questions for working conditions, occupational safety and health, worker data and fundamental rights. 

This creates a need for capabilities that cannot be reduced to technical literacy alone.

Inspectors need to understand digital systems sufficiently to question them, but they also need the professional skills that technology does not replace: interviewing, listening, critical judgement, empathy, communication and the ability to interpret incomplete or conflicting evidence.

Digital transformation therefore requires investment in institutions and people, not only in tools.

The strongest evidence of the learning journey came from what participants did with it.

During the residential phase, every participant presented an individual proposal on how digitalization could strengthen labour inspection in their own context. Proposals were discussed and assessed by peers as well as tutors, making feedback itself part of the learning process. 

Participants then developed these ideas further through an Applied Final Project. Rather than designing technology for its own sake, they were asked to identify one concrete labour inspection challenge, assess their institution’s existing data and digital maturity, propose a feasible solution, define safeguards and human oversight, and set out a realistic implementation pathway. 

The resulting projects covered remarkably different institutional needs. Participants explored the use of open-source intelligence to identify risk signals in online job advertisements; digital monitoring of statutory labour levies; predictive analysis for labour formalization; risk-based planning for undeclared work; digital transformation of occupational accident investigations; electronic management of high-risk work permits; integrated inspection platforms; OSH knowledge verification; and digital consultation and information systems. 

One of the strongest features across the projects was restraint.Several participants proposed improving existing systems rather than creating entirely new ones. Others began with small pilots, limited datasets or a clearly bounded institutional workflow. In some cases, the most appropriate digital transformation did not involve AI at all, but better case management, interoperability or structured information.

The residential week represented only one stage of the programme.

The complete learning journey combined online preparation, the five-day Turin phase and post-course assignments between 29 June and 20 July. Assessment integrated self-guided learning, participation, peer-reviewed presentations, an applied project and an AI-supported interview in which participants had to explain and defend their own proposals. 

Course evaluation results were also strong. The final report recorded an average satisfaction score of 4.49 out of 5, with 97.26 per cent of responses rating the course either 4 or 5 and all respondents recommending the course. 

By the end of the programme, the conversation had moved well beyond the question of what artificial intelligence can do.

Participants were instead asking more demanding questions:

What does our inspectorate need to achieve? What information do we already have? What evidence are we missing? Which method is appropriate? What should remain a human decision? And what is the smallest, realistic step we can take towards improvement?

These questions lie at the heart of responsible digital governance. Ultimately, digital transformation depends on strong institutions, informed professionals and a commitment to protecting the people they serve. In labour inspection, digital transformation is only as effective as the people and institutions that put it into practice.