// FEATURE //
This is why data classification and categorisation must come before AI adoption and be made an ongoing priority thereafter.
Upfront, all internal data must be audited. It sounds daunting, but it has to be done. For AI to be leveraged properly without risk, all data needs to be classified in a manner that gives AI systems an indication of what it’ s looking at and how sensitive it is.
Greg Jones, Senior Vice President of MSP Success( EMEA & North America), Kaseya
Data should also be categorised by severity. Organisations should consider what’ s genuinely confidential, such as IP and HR data, as well as what’ s low-risk or irrelevant. A useful test here is to consider what would happen if each file, document or piece of information is released for public consumption – would it affect the company, or not? Start applying that lens retrospectively, as well as going forward.
Thereafter, anything that truly matters should be ringfenced away from the AI.
The most sensitive data should reside in a demilitarised-zone-equivalent that AI simply cannot reach. Not restricted by permissions, but genuinely walled off from any model, be it internal or external.
Equally, organisations should begin to practice data reduction. Most data only needs to be retained for a defined period. Beyond that, if you don’ t need it, get rid of it. If it’ s not providing value, it may just be sitting there as a risk.
Start small and create AI sandboxes before you scale
None of this means slowing AI adoption to a crawl. It simply means being deliberate about how you start your AI journey.
The key is to not run before you can walk. Early AI projects should be treated as proof of concepts that are well tested for both effectiveness and potential issues before scaling them up into full blown tools and solutions.
Sandboxes are genuinely useful in this process. Using dummy data sets that mirror your real environment will ensure you don’ t risk exposing anything sensitive, while also letting you trial, test and refine AI models before they go anywhere near live systems.
The US $ 2.59 trillion being poured into AI this year represents genuine ambition. But ambition without control is how projects fail, budgets get wasted and businesses end up exposed to risks they never intended to take on.
Data needs to be at the centre of this control. Audit everything, categorise everything, scrutinise everything. Get the foundations right and the AI built on top will stand a genuine chance of becoming a genuinely useful tool for years to come. �
Intelligent SME. tech
27