“Old Mutual group chief technology and data officer Dhesen Ramsamy has urged South African businesses to invest in employee skills, data quality, organisational culture and responsible governance alongside AI technology, arguing that sustainable competitive advantage depends on trust and measurable business outcomes. The report highlights opportunities to accelerate software development and improve financial services, while warning that poor data, inadequate oversight, cybersecurity threats and the underrepresentation of African communities in global AI datasets could undermine the technology’s benefits.”
South Africa’s AI Future Depends on Skills, Trust and Better Data, Old Mutual Technology Chief Says
Moving beyond the excitement surrounding AI
South African businesses are facing an important turning point as artificial intelligence becomes increasingly integrated into financial services, software development, customer support and business decision-making. While companies worldwide are investing in advanced AI tools, questions about the practical value of those investments are becoming just as important as the technology itself.
A report published by Business Day on 21 October 2026 highlights this challenge through the perspective of Dhesen Ramsamy, group chief technology and data officer at Old Mutual. Ramsamy argues that organisations cannot secure a lasting advantage simply by purchasing new systems or adopting AI faster than their competitors. Instead, companies must develop the skills, data foundations, governance processes and workplace cultures necessary to use the technology responsibly and effectively.
His message is particularly relevant to South Africa, where businesses are exploring digital transformation while navigating skills shortages, unequal access to technology and the need to protect sensitive customer information.
The central question is therefore changing. Rather than asking whether an organisation should adopt artificial intelligence, leaders increasingly need to determine which problems AI should solve, what resources are required and how the resulting benefits and risks should be measured.
Why businesses need to rethink AI investment
Artificial intelligence has attracted substantial interest because it can analyse information, automate repetitive activities, assist software developers and help organisations make decisions using large datasets.
However, purchasing AI software does not automatically translate into higher productivity or improved customer experiences. A company may introduce sophisticated tools without having reliable information, trained employees or clearly defined procedures for using them.
Ramsamy’s argument is that investment in technology must be accompanied by equivalent attention to human capabilities and organisational readiness.
For example, an insurer might introduce AI to help process claims. The system could identify missing documents, classify incoming applications and highlight cases requiring further investigation. Yet the benefits would be limited if customer records were incomplete, employees did not understand the system or nobody was responsible for checking its recommendations.
The same principle applies to banks, retailers, telecommunications companies and government departments. Technology must fit into a functioning business process rather than operate as an isolated experiment.
This distinction matters because companies need evidence that AI is improving outcomes. Useful measures could include processing times, error rates, customer satisfaction, operational costs and the quality of decisions.
A successful implementation should demonstrate measurable improvement against a defined starting point. Otherwise, businesses risk spending money on impressive demonstrations that fail to deliver lasting value.
Financial services offer significant opportunities
Financial services are among the sectors where AI could have a substantial practical impact. Banks and insurers process large volumes of information, assess risk and communicate with customers through multiple channels.
According to the Business Day report, Old Mutual is exploring technology and machine learning to improve areas including underwriting, premium calculations, claims processing, risk assessment and pricing.
These applications could make some processes faster and more consistent.
For instance, AI-assisted claims processing could help employees organise documentation, identify missing information and prioritise cases. Underwriting teams could use analytical tools to examine relevant information more efficiently, while customer-service departments could use AI assistants to answer routine questions.
Nevertheless, financial decisions require particular care because mistakes can affect people’s access to insurance, credit and other essential services.
A system that incorrectly classifies a claim or produces an unfair risk assessment could harm customers if its recommendation is accepted without review.
Consequently, automation should not eliminate accountability. Financial institutions need clear procedures for reviewing consequential decisions, investigating complaints and correcting errors.
Customers should also understand when AI is being used in interactions that materially affect them. Transparency helps people know whether they are dealing with an automated system, how their information is being used and how to seek human assistance.
For South African financial institutions, the opportunity is not simply to process more transactions. It is to deliver more reliable services while preserving the trust on which long-term customer relationships depend.
Data quality is the foundation of reliable AI
One of the strongest themes in Ramsamy’s remarks is the importance of data quality.
AI models depend on information supplied during their development and operation. If that information is inaccurate, incomplete, inconsistent or poorly organised, the resulting outputs may be unreliable regardless of how advanced the underlying technology appears.
Businesses frequently hold information across different databases, legacy systems and departmental applications. Customer details may be duplicated, formats may differ and important records may be missing.
Before introducing AI at scale, organisations therefore need to understand what information they possess, where it comes from, who can access it and whether it is suitable for the intended purpose.
This preparation can be expensive and time-consuming, but ignoring it creates additional risks.
Consider an organisation attempting to build a customer-support assistant using outdated product information. Even a technically sophisticated model could provide incorrect answers if its reference material is inaccurate.
Similarly, an insurance system trained on incomplete historical records might produce misleading assessments or perform differently for customer groups that were poorly represented in the data.
Companies should establish data-quality standards, access controls, documentation requirements and procedures for correcting errors. They should also monitor how information changes over time.
Good data governance is therefore not merely an administrative responsibility. It is a practical requirement for achieving dependable results from AI.
African representation in global AI datasets
Another important issue raised in the report is the limited representation of African communities in global AI datasets.
Many widely used AI systems are developed using information collected from markets with different languages, social conditions, cultural practices and economic circumstances from those found across Africa.
As a result, systems that perform well in one setting may struggle in another.
A language model might handle widely represented languages effectively while producing less accurate responses in African languages. A financial assessment system could also misunderstand local economic circumstances if the data used to develop it does not adequately reflect those realities.
These limitations matter in South Africa, a country with 12 official languages and substantial differences in income, access to digital services and everyday experiences.
AI systems designed without sufficient local representation could unintentionally exclude communities, reproduce existing inequalities or provide lower-quality services to particular groups.
Addressing the problem requires more than translating an existing interface. Developers need representative datasets, appropriate evaluation methods and participation from people who understand local languages and communities.
South African universities, technology companies, public institutions and research organisations could contribute by supporting locally relevant research, developing datasets responsibly and testing AI products against real-world needs.
Such work must also respect privacy and the rights of individuals whose information is used.
The long-term objective is to ensure that African organisations can adapt and develop technology rather than relying exclusively on systems designed elsewhere.
From technology consumers to technology creators
South Africa has an opportunity to strengthen its role in the wider AI economy by investing in local expertise and innovation.
Using international AI products can provide immediate benefits, particularly where developing a system independently would be prohibitively expensive. However, excessive dependence on external providers can create challenges involving cost, control over data, technical expertise and the ability to customise products.
Ramsamy has argued that African institutions should move beyond consuming foreign tools towards training local models and strengthening data sovereignty.
This does not mean every organisation should build its own large AI model. For many businesses, adapting existing models or using specialised applications will be more practical.
The broader priority is to develop enough local capability to evaluate systems, manage risks, customise solutions and build products suited to African markets.
Universities and training institutions can help by expanding education in computer science, data engineering, cybersecurity, machine learning and AI ethics.
Businesses can support this process through apprenticeships, professional development, partnerships with universities and opportunities for employees to work on practical projects.
Public policy also has a role to play in supporting research, improving digital infrastructure and creating conditions in which responsible innovation can flourish.
If these efforts succeed, AI adoption could create opportunities for local technology firms and skilled workers while improving services across established industries.
Productivity gains should not be measured only through job cuts
The debate about AI frequently centres on whether automation will eliminate jobs. Although some tasks may become less labour-intensive, the immediate effects can differ substantially between industries and occupations.
The Business Day report cites Ramsamy’s account that engineering teams using AI-assisted development reduced delivery times by 30% to 50% on particular tasks. One example involved reducing the development time for a user journey from several months to approximately six weeks. These are examples reported by the executive, not a verified industry-wide productivity estimate.
Such results illustrate how AI can help employees produce and test software more quickly.
However, faster initial development does not necessarily mean an entire project can be completed in the same proportion of time. Security reviews, testing, integration, compliance checks and maintenance still require attention.
AI-assisted software development can also change the nature of technical work. Developers may spend less time generating routine code and more time reviewing outputs, integrating systems, identifying vulnerabilities and ensuring that applications behave as intended.
Organisations should therefore evaluate productivity across the complete workflow rather than focusing on isolated tasks.
The same principle applies to other professions. AI may help employees summarise documents, analyse information or prepare initial drafts, allowing them to devote more attention to complex problems and customer needs.
Whether these changes produce better outcomes depends on how organisations redesign their work.
Businesses that invest in retraining and practical support may be better positioned to help employees adapt than those that introduce new tools without explaining their purpose or providing opportunities to learn.
Cybersecurity and customer trust remain essential
AI adoption also creates security and privacy challenges.
Organisations must protect confidential information, restrict access to sensitive records and prevent employees or automated systems from sharing data inappropriately.
Financial institutions face particular pressure because they handle personal information, insurance records and other sensitive customer details.
The report describes Old Mutual’s use of controls that include staged deployment, inventories of AI tools, pre-deployment checklists, data-loss prevention measures and human sign-offs for use cases.
These measures reflect an important principle: AI systems should be assessed before they are deployed, monitored while they operate and reviewed when risks change.
Companies also need to consider how AI may be exploited by attackers. Criminals can use AI tools to produce more convincing phishing messages, impersonate people or automate aspects of fraud.
Ramsamy additionally raises the longer-term concern that attackers may collect encrypted information today in the hope of decrypting it when future computing capabilities make that possible.
The precise timing and implications of future quantum threats remain uncertain, but organisations should monitor developments in cryptography and plan for appropriate security upgrades.
Protecting customers requires a combination of technical safeguards, employee awareness, incident-response planning and clear accountability.
Ultimately, trust is difficult to restore after a serious breach or an unexplained automated decision. Security and transparency must therefore be built into AI projects from the beginning.
Why company boards must understand AI governance
AI is no longer solely a technical matter for software developers and IT departments. Decisions about its use can influence customers, employees, financial performance and legal responsibilities.
For this reason, company boards and senior executives need sufficient knowledge to oversee AI investments, even if they do not need to understand every technical detail of individual models.
Leaders should ask several fundamental questions before approving a project.
First, what specific customer or societal problem is the system intended to solve? Second, is the data suitable, accurate and responsibly managed? Third, who is accountable for the risks? Finally, is the expected benefit substantial enough to justify the costs?
These questions help distinguish practical investments from projects driven mainly by enthusiasm for new technology.
Boards should also require appropriate reporting on performance, security incidents, model limitations and the handling of complaints.
For high-impact applications, organisations may need additional independent assessments, documented human oversight and processes through which affected individuals can challenge decisions.
Clear governance does not have to prevent innovation. Properly designed controls can help organisations identify problems early, protect customers and scale systems with greater confidence.
South Africa’s wider AI readiness challenge
The concerns raised by Old Mutual’s technology chief reflect broader developments across Africa.
A report published on 21 October 2026 by Lifestyle & Tech, summarising Boston Consulting Group’s AI Maturity Matrix 2026, identifies skills development, infrastructure, investment, policy and research as important factors in determining how effectively countries adopt AI. The report argues that governments must turn strategic ambitions into coordinated action.
Reliable connectivity, access to computing resources, affordable digital services and relevant education all influence who can benefit from AI. Small businesses may face different barriers from large financial institutions, particularly when they lack specialist employees or the resources to experiment with expensive systems.
Public services also present opportunities, including administrative automation, improved access to information and better analysis of operational data. However, these applications require appropriate privacy protections, human accountability and careful assessment of potential harm.
Policymakers must balance the need to encourage innovation with the responsibility to protect citizens.
Businesses, universities, government agencies and civil society can contribute by sharing expertise, supporting skills development and establishing practical approaches to responsible AI.
The goal should be broad-based economic and social benefit rather than technology adoption for its own sake.
Conclusion: A more disciplined approach to AI investment
South Africa’s AI opportunity will depend on more than access to advanced software. It will require organisations to develop reliable data foundations, strengthen employee capabilities, improve cybersecurity and establish clear rules for making and reviewing AI-assisted decisions.
Dhesen Ramsamy’s message to business leaders is that sustainable competitive advantage comes from combining technology with capable people and trustworthy institutions.
Financial services provide a useful example of both the opportunities and the responsibilities involved. AI could accelerate claims processing, support software development and improve access to information, but poorly governed systems could expose customers to unfair decisions, privacy breaches and costly mistakes.
For South African businesses, the practical response is to begin with clearly defined problems, measure outcomes and expand successful projects only when the necessary safeguards are in place.
Education and workforce development must also remain central. Employees need opportunities to learn new skills, while local researchers and developers need the resources to build and adapt systems for African circumstances.
Government and industry should work towards dependable infrastructure, proportionate oversight and stronger local innovation.
Artificial intelligence may eventually reshape many aspects of the economy, but its value will not be determined by the sophistication of its algorithms alone. It will depend on whether organisations can apply those algorithms responsibly, fairly and effectively to real human needs.
For South Africa, the challenge is not simply to adopt AI. It is to build the capability and trust required to make the technology useful for businesses, workers and communities.
