AI Agents Are Changing Information Technology: Are Businesses Ready for Software That Can Act on Its Own?

For years, businesses have understood software as something people operate. An accountant opens an application to prepare financial reports. A customer service officer responds to complaints through a digital platform. A manager checks a dashboard before making a decision. A programmer writes instructions that tell a computer what to do.

Artificial intelligence agents are beginning to challenge this familiar relationship between people and technology.

Instead of simply waiting for a person to enter a command, an AI agent can receive an objective, examine available information, decide what actions are necessary and carry out several steps using connected digital systems. In practical terms, we are moving from software that people use towards software that can increasingly perform work on their behalf.

This change could become one of the most important developments in information technology since the arrival of cloud computing.

An ordinary generative AI system may answer a question or draft a document. An AI agent goes further. It can potentially search a database, compare information, prepare a response, update another system, send a notification and continue monitoring the task until it is completed. Some agents can also communicate with other specialised agents, creating networks of digital workers responsible for different parts of a business process.

The implications for organisations are significant.

Consider a company receiving hundreds of customer enquiries every day. An AI agent could examine each request, retrieve information about the customer, check previous communication, identify the appropriate response and resolve simple cases without waiting for an employee to handle every step. More complicated cases could then be transferred to a human officer.

In banking, an agent might examine transactions for unusual behaviour and prepare cases for investigation. In logistics, agents could monitor inventory, delivery schedules and supplier information before recommending or initiating appropriate actions. In healthcare administration, they could assist with appointment management, documentation and routine communication. Universities could use them to support admissions, student services, research administration and academic information management.

Software development itself is also changing. AI agents can already assist programmers with analysing code, identifying problems, writing tests and completing portions of software development workflows. This suggests that the influence of agents will not be limited to business users. They may change the information technology profession itself.

Gartner reported in September 2026 that enterprise AI assistants are increasingly becoming a major entry point through which employees conduct research, analyse information, create content and delegate actions. The research firm has also argued that AI agents could substantially reshape how organisations buy and use enterprise software.

The attraction is understandable. Businesses constantly search for ways to reduce repetitive work, improve response times and allow skilled employees to concentrate on activities requiring judgement, creativity and human relationships. Properly designed agents could help organisations achieve these objectives.

But giving software the ability to act creates a very different problem from giving software the ability to answer questions.

If an AI chatbot produces an incorrect answer, someone may recognise the mistake and disregard it. If an autonomous system makes an incorrect decision and immediately acts on that decision, the consequences can be more serious.

Imagine an AI agent incorrectly cancelling a customer's order, approving an inappropriate transaction, deleting important information or sending confidential material to the wrong person. The more systems an agent can access, the greater the potential consequences of an error.

This is why the discussion about AI agents must move beyond excitement about automation.

The important question for businesses is not simply, “Can we deploy AI agents?” It is, “Which decisions should an AI agent be permitted to make, and under what level of human supervision?”

Recent evidence suggests that many organisations have not fully answered that question. An IBM study involving 2,000 technology executives found that only 11 percent believed their organisations were completely prepared for the scale of AI agent deployment they expected. The same study found that 77 percent said AI adoption was already moving faster than their existing governance capabilities.

That should concern business leaders.
An organisation would not normally employ a new worker, immediately provide access to financial records, customer databases, payment systems and confidential documents, and then allow that person to make unlimited decisions without supervision. Businesses should be equally careful when deciding what an intelligent software agent is permitted to access and execute.

Every organisation adopting agents therefore needs clear boundaries.

The first requirement is data quality. An intelligent system cannot consistently make good decisions when the information available to it is inaccurate, incomplete or badly organised. Many businesses still operate with records scattered across spreadsheets, emails, paper documents and disconnected software platforms. Deploying sophisticated agents on top of poor information infrastructure may simply automate existing weaknesses.

The second requirement is cybersecurity. AI agents may require identities, passwords, application permissions and access to sensitive business information. Each additional connection can create another security concern. Microsoft has acknowledged this challenge in its own enterprise work, noting that increasingly autonomous agents require careful attention to security, governance and operational control.

The third requirement is accountability. Businesses must know who remains responsible when an agent acts incorrectly. Technology should not create an accountability vacuum where employees blame the system and managers cannot determine why a decision was made.

For important decisions involving money, employment, safety, health, legal obligations or personal information, meaningful human oversight remains essential.

There is also the question of employment.

Whenever a technology becomes capable of performing tasks traditionally done by people, concerns about job losses naturally emerge. Some routine responsibilities will almost certainly be automated. But history also shows that technological change frequently transforms occupations rather than simply eliminating them.

The more realistic challenge for many organisations may therefore be redesigning jobs around cooperation between people and intelligent systems.

Microsoft's 2026 Work Trend Index describes an emerging environment where agents increasingly handle portions of execution while people retain responsibility for directing work, exercising judgement and determining outcomes. The study was based on a survey of 20,000 workers using AI across ten countries alongside large scale workplace data.

This distinction matters particularly for developing economies.

For Ghana and other African countries, AI agents could provide an opportunity for businesses to overcome some longstanding limitations in administrative capacity and digital service delivery. A small company may eventually use intelligent agents to provide customer support, analyse sales information, prepare business reports and manage routine administrative activities that previously required a much larger workforce.

Government institutions, universities, hospitals, financial institutions and local technology companies could also benefit from carefully designed agent based systems.

But Africa should not become merely a consumer of intelligent software created elsewhere.

Universities and technology institutions should train students not only to use AI tools but also to understand software engineering, data management, cybersecurity, machine learning, responsible AI and digital governance. Local developers should be encouraged to build systems that understand African institutions, languages and economic realities.

Businesses also need to resist the temptation to automate simply because automation has become fashionable.

The most sophisticated organisation will not necessarily be the one with the largest number of AI agents. It will be the organisation that knows exactly where intelligent automation creates value and where human judgement remains indispensable.

Gartner's guidance to business leaders in 2026 reflects this principle. It argues that organisations should establish limits on autonomy, improve data and identity management, redesign work before automating it and continuously govern agent behaviour rather than treating governance as an activity performed only when a system is first deployed.

This is the conversation business leaders should be having now.

The future of information technology may no longer revolve entirely around people opening applications and manually performing every digital task. Increasingly, people may describe objectives while intelligent software performs parts of the work across multiple systems.

That future offers enormous possibilities, but autonomy without accountability would be dangerous.

Businesses should therefore begin with controlled applications, establish strong security and data governance, keep people involved in consequential decisions and measure whether agents are actually improving productivity and service quality.

The question is no longer whether software will become capable of acting.

It already is.
The more important question is whether our businesses, institutions, laws, skills and systems of accountability are developing quickly enough to ensure that when software acts on our behalf, it acts within boundaries that we understand and control.

AI agents may become powerful digital workers. But businesses must remember one principle as they enter this new era of information technology: intelligence can be delegated, but responsibility cannot.

Philipa Serwaa
University of Energy and Natural Resources P.O. Box 214 Sunyani, Bono Region

E-mail: [email protected]