The corporate AI market has spent the past few years putting increasingly capable assistants in front of employees. Chatbots can write reports, summarise documents and answer questions, while copilots have been added to the software people already use. The next stage may be less visible, with companies using AI to handle the processes that sit behind those interfaces, from moving information between systems to processing documents and identifying problems that would otherwise require someone to notice them manually.

South African technology group 4Sight Holdings is positioning itself in that part of the market. The company, which is listed on the Johannesburg Stock Exchange, has been expanding its 4Sight Automated Intelligence platform, or 4AI, around business-process automation, while recently adding Clark Fourie as chief information technology officer, Jeandré du Randt as chief business environment officer and Essich Wassenaar as group frontier operations officer. The appointments are part of an effort to bring technology, operations and business processes closer together as the company develops its AI offering.

4Sight describes its approach as a combination of data, process, training and organisational change. Its platform is designed to work across areas such as enterprise resource planning, customer relationship management, human resources, payroll, reporting and other core business systems, rather than functioning only as a general-purpose AI assistant. The company has also described a progression from relatively simple task automation towards process automation, decision support and, eventually, more autonomous systems.

Related ArticleThe Race to Make AI Work in African Languages→

The distinction is becoming more relevant as businesses move from experimenting with generative AI to deciding where it can produce measurable operational gains. An employee asking an AI system to summarise an invoice is useful, but it leaves the rest of the process intact. A system that can extract the information, compare it with existing records, identify an anomaly and send the matter to the appropriate person is doing something different: it is participating in the workflow rather than simply assisting the person performing it.

That is the type of application 4Sight is pursuing. One example used by the company involves documents such as utility bills. An AI system could extract information from a bill and compare it with previous consumption, potentially flagging an unusual increase that requires investigation. The technology becomes more useful when the information it extracts is connected to another action, rather than ending with a summary on an employee's screen.

Much of the work required to make that possible has little to do with the spectacle surrounding generative AI. Companies need reliable data, clearly defined processes and systems that can communicate with one another. They also need to know who is responsible when an automated process produces the wrong result. These requirements can become difficult in large organisations where workflows have developed over years and information is spread across different software platforms.

4Sight's own description of its AI strategy reflects that problem. The company says its automated-intelligence solutions are intended to operate across existing business systems, including ERP, CRM, HR and reporting platforms. Its 2025 annual report also describes AI projects involving procurement, operations, finance, employee experience and customer engagement, suggesting that the company is treating automation as an organisational project rather than a standalone software feature.

The economic case for this approach is straightforward. If software can reduce the amount of manual work involved in repetitive processes, a company can potentially lower costs, process more transactions or allow employees to spend more time on work that requires judgement. But the same technology can also reduce demand for particular tasks, and whether businesses use the resulting efficiency to expand operations or reduce their workforce is ultimately a management decision rather than a technical one.

Research from the International Labour Organization suggests that the effect of generative AI is likely to be uneven. Its 2025 global assessment found that one in four workers are in occupations with some exposure to generative AI, but concluded that transformation of jobs is more likely than complete replacement because most occupations still contain tasks requiring human input. The organisation also points to infrastructure, skills, cost and operational difficulties as barriers to widespread adoption.

That makes the human side of enterprise automation important. An organisation cannot simply install an AI system and expect employees to adjust themselves around it. Workers need to understand what the system is supposed to do, how its decisions should be checked and what they are expected to do when it fails. The World Economic Forum's Future of Jobs Report 2025 found that 86% of employers surveyed expect AI and information-processing technologies to transform their businesses by 2030, while 77% expect to pursue reskilling or upskilling as part of that transition.

For companies building these systems, this creates a different problem from the one faced by consumer AI products. A chatbot can be useful even when a user occasionally receives an imperfect answer. An automated business process has a narrower margin for error because the system may be connected to money, customer records, inventory, payroll or other operational information. As AI moves closer to taking action rather than simply producing text, questions about permissions, monitoring and accountability become increasingly important.

4Sight has described a longer-term vision in which AI systems could operate as digital counterparts to employees, handling routine work continuously while people retain responsibility for reviewing the results. The company places this idea within a broader progression towards autonomous intelligence, where systems could eventually manage processes with much less direct intervention. That remains an ambition rather than an established feature of enterprise software, and the practical difficulty will be determining how much authority businesses are willing to give systems whose decisions may affect their operations.

The company is also pursuing a model in which technology developed for its own operations can become part of its commercial offering. That approach allows 4Sight to use its internal processes as a testing ground for automation before adapting successful applications for customers. It also reflects a wider trend in enterprise software, where companies increasingly build specialised AI tools around existing business systems instead of attempting to compete directly with the general-purpose models produced by the largest AI companies.

The strategy does not depend on 4Sight developing the most powerful foundation model. Its opportunity lies further down the stack, where AI has to interact with the less glamorous machinery of a business: databases, documents, approval chains, enterprise applications and employees. That is also where some of the industry's biggest claims about productivity will eventually have to be tested.

The appeal of enterprise AI has never really been the chatbot itself. It is the possibility that software can take over enough repetitive work to change how an organisation operates. 4Sight is betting that the companies that make the most of AI will eventually spend less time asking what the technology can say and more time deciding what work they are prepared to let it do.