AI Is Moving from Pilots into Processes

Artificial intelligence is increasingly becoming a practical tool for improving business processes rather than remaining a standalone technology experiment. It is particularly relevant to small and medium-sized companies that do not have large internal development and optimisation teams, although the same principles apply to corporations. “We can offer solutions that significantly improve the development of small and medium-sized businesses, and corporations as well,” Mislav Galler, Management Board Member and Chief Commercial Officer at Telemach Croatia, tells ICTbusiness Media - ICTbusiness.info. The point is not to hand an entire business over to AI, but to define precisely which tasks it can perform and where it creates measurable value.

For smaller organisations, AI can partly compensate for a shortage of specialist capacity in activities that are not part of the strategic core. A company may not be able to employ a top specialist for every operational function, but a well-designed system can provide a capable digital collaborator. “For jobs that are not of crucial or strategic importance, you can get an excellent collaborator,” Galler says. The purpose is not to eliminate expertise, but to free specialists from part of the routine workload.

The same principle applies in large organisations. Repetitive accounting, technical and operational tasks are among the first candidates for automation because their patterns are easier to describe and expected outcomes can be defined more clearly. “In a large corporation, repetitive accounting or technical-operational work can already be improved substantially,” he says. AI then becomes an additional tool for improving productivity rather than an autonomous substitute for the organisation itself.

The boundary becomes clearer when processes involve sensitive databases, confidential information or decisions with major business and legal consequences. Galler therefore stresses the human-in-the-loop model, in which people remain involved in supervision and decision-making in sensitive areas. “You would not place it into very delicate functions involving certain databases and sensitive information without human oversight,” he says.

Even with those limits, he sees the potential acceleration of business processes as enormous regardless of company size. Automation can shorten processing times, take over repeated steps and allow employees to focus on work where experience, judgement and business context matter. “The acceleration of the business process, whether in a small, medium-sized or large company, is immeasurable,” Galler says.

To achieve that benefit, however, a company must first know exactly what it wants from the system. A poorly defined process does not improve simply because AI is added to it. Expectations, responsibilities, data sources, and verification methods all need to be specified. “We have to be very precise about what we are asking for and clearly define the expectations,” he stresses. In many cases, this preparation is more important than the technical connection itself.

His experience is that integration with existing IT systems does not necessarily represent the hardest part. Technology and integration mechanisms have advanced enough for connectivity to be solved faster than organisational and regulatory questions. “The smallest problem is implementation and connecting to existing systems; that can often be resolved much faster,” Galler says. The focus therefore shifts from whether something can be built technically to the conditions under which it may be deployed.

Legal and regulatory prerequisites are therefore among the current bottlenecks. Companies need to understand which data enter the system, who can access them, what happens to them and whether parts of the process require additional restrictions. “We have to be very careful to resolve the legal and regulatory prerequisites, and that usually takes longer,” he says. A technical solution may therefore be ready while deployment into production still depends on governance issues being clarified.

This is particularly important when AI is connected to personal data, confidential business information or other sensitive content. The relevant question is no longer only how capable the model is, but whether the complete system preserves organisational control over data and responsibility. Galler sees this combination of technology, process and regulation as the real implementation challenge.

Despite those constraints, his conclusion is that the period of waiting is over. Repetitive processes are already mature enough for serious use, provided that companies understand what they are automating and why. “I think it is already here, and I would suggest that everyone seriously informs themselves and starts,” Galler concludes. The next phase will therefore be less about whether AI has a business use case and more about how well organisations can embed it into existing processes with clear controls and human supervision.