Improper AI deployment can lead to staffing crises and productivity losses for businesses.

After years of heavy investment in AI, many businesses are facing a paradox: instead of improving efficiency and competitiveness, they are inadvertently weakening their core operational foundations. According to Datatonic, a cloud-based data and AI consulting firm, the cause lies not in the technology itself, but in the disconnected implementation between humans and AI.

 

Datatonic suggests that in the next phase of enterprise AI, success will belong to organizations that build 'human-in-the-loop' (HiTL) systems – that is, AI operating within a controlled environment and in close coordination with humans, rather than operating in isolation.

The company's research indicates that businesses that don't integrate AI into their actual workflows are falling behind the competition as productivity stagnates. According to Scott Eivers, CEO of Datatonic, AI is not simply a new tool but an opportunity to redesign how work is done. The biggest risk in the market today is the 'productivity leak' that occurs when AI exists independently, not connected to the people directly running the business.

 

The pressure to demonstrate the effectiveness of investments is increasing. However, many AI initiatives remain stuck in the testing phase due to a lack of user trust. As a result, businesses fail to leverage AI analytics and recommendations to improve decisions and processes, meaning the performance benefits never materialize.

According to Datatonic, the HiTL model is key to the future because it combines the processing speed of AI with human judgment and accountability. This is clearly demonstrated in the field of AI-assisted software development: AI systems can generate code from rudimentary requirements, but humans still define objectives, review requirements, and validate plans before implementation. Once the direction is clear, AI agents then begin building each component in a modular structure.

This trend is also spreading to finance and operations. In back-office departments, AI-powered document processing systems are said to reduce invoice processing costs by up to 70%. However, the final decision still rests with the finance team to ensure accuracy and compliance.

Andrew Harding, CTO of Datatonic, believes this is a story of collaboration, not replacement. Humans build systems for evaluation, validate plans, establish control barriers, and make decisions; while AI handles execution at speed and scale. It is this combination that creates real value for businesses.

However, Datatonic warns that many organizations are still not ready to fully deploy automated agents due to a lack of appropriate security mechanisms and governance frameworks. Automation can only scale when businesses establish approval points, performance evaluation standards, and continuous monitoring systems, especially in the context of ever-changing AI models. Ignoring governance will not increase speed but only increase risk.

Harding emphasized that once trust is built, businesses can gradually delegate more responsibilities to AI. But building trust takes time and a well-structured control mechanism.

Datatonic predicts that over the next two years, workloads will accelerate dramatically with deeper involvement of AI actors, particularly in preparation and validation. AI could even be used to test and eliminate ineffective decisions before businesses allocate resources.

Scott Eivers believes the future will be of lean yet highly specialized departments – from finance and human resources to marketing – whose capabilities are amplified by AI. The winning businesses will be those that know how to train people to work with AI, rather than trying to avoid or isolate this technology.