Most organisations across SA have already decided to adopt AI. But adoption is only the beginning of the conversation. The more difficult questions are whether AI can be trusted, whether the data underpinning it can be trusted, and whether that data can be recovered cleanly and accurately when something goes wrong.
The recent ITWeb and Veeam survey set out to measure how ready African organisations are to adopt, govern and scale AI. For Tahir Latif, data trust and governance lead for EMEA East at Veeam Software, and Mark Govender, manager for systems engineering, Africa at Veeam Software, the results reveal a gap between how confident organisations feel and what they can prove.
“Confidence is multifaceted,” says Latif. “Confidence in adopting AI is very high because the potential returns are too significant for organisations to sit on the sidelines. However, confidence in the readiness of the underlying data is much lower.”
Drawing on responses from 140 senior IT, data and security leaders at medium and large organisations across 14 African countries, the survey paints a picture of confidence that begins to thin under scrutiny.
When asked whether they could recover clean and usable data after a ransomware attack, 52% of respondents said they were very or extremely confident, while 33% described themselves as moderately confident. However, almost a third (32%) have either never run a full recovery test, have not done so for more than a year, or cannot say when their last test took place. Among those who have tested the recovery process end to end, 45% experienced only partial success, significant gaps or outright failure.
“A lot of organisations are confident that they have good visibility across their data estate,” says Govender. “In reality, many do not. Do you know where sensitive data resides? Do you know which data contains personal information? Organisations need to understand all of that before they move ahead with an AI initiative.”
Visibility is the starting point for data and AI trust. Organisations cannot govern, protect or responsibly use data if they do not know what they have, where it is located, who can access it or how it is being used. As AI increases the speed and scale at which data can be accessed and processed, that understanding becomes even more important.
Access is the next layer. AI agents can interact with data differently from human users, introducing new governance and security considerations.
“Every employee is given access according to their role,” Latif explains. “But an agent is not human. If an agent is given unrestricted access, it will not question whether that access is appropriate. It will simply ingest the data.”
Managing this risk means applying least-privilege access to AI models, systems and agents as rigorously as it is applied to people. Organisations need visibility into which systems and agents are interacting with their data, alongside clear governance over what each one is permitted to access and use.
Recoverability is another essential component of data and AI trust. More than half of survey respondents rated their backup and recovery capabilities as mature, yet far fewer regularly test whether their backups can be restored successfully.
“Having backups does not necessarily mean that an organisation is resilient,” says Latif. “There may be corruption within the data, so an organisation could complete a recovery only to discover that it has restored an inaccurate or compromised version.”
Govender adds: “It is not enough to have successful backups. Organisations need to be able to prove that they can recover clean, accurate and usable data.”
This is where Veeam supports the wider data and AI trust agenda. By helping organisations gain greater visibility across their data estates, understand how data is being accessed and maintain a verified copy of business-critical information, Veeam enables organisations to recover to a trusted data state when something goes wrong.
This is increasingly important as AI agents and applications become more deeply embedded within business operations. An incorrect, unauthorised or malicious action can move quickly through an organisation’s data estate. The ability to identify what has changed and restore a trusted version of the affected data helps organisations contain the impact and maintain confidence in the information driving AI systems and business decisions.
“If an agent or system does something malicious or incorrect, we can help the organisation undo it,” concludes Govender. “Ultimately, data and AI trust is about knowing that the data powering the business is visible, governed, protected and recoverable, so it can be relied on to support accurate decisions and trusted AI outcomes.”
- ITWeb TV Biz episodes are sponsored.
Share
