The importance of preparing data for AI integration

Despite the importance and timely arrival of the EU AI Act, there remain some major compliance concerns and the impact it will have on AI adoption and governance strategies. In fact, a recent survey found that having the proper AI governance in place is a top priority for 41 percent of business decision-makers. However, around one-quarter of UK firms have yet to make preparations for AI, and this is partly due to lingering confusion over their obligations. 

Yet, the requirements set out by the Act are specific, particularly for “businesses or public authorities that develop or use AI applications that constitute a high risk for the safety or fundamental rights of citizens.” This high-risk category can include anything from law enforcement and employment systems to those used by life sciences and critical infrastructure organizations. 

Fundamentally, organizations developing or adopting AI technologies must take a risk-based approach based on a range of considerations, from transparency and accountability to oversight and compliance. 

In this context, one of the priorities leaders need to focus very clearly on is preparing their data, such as what datasets they can and can’t use, their sensitivity levels and accuracy, among others. But this is just the start. Other crucial steps in the overall preparation process include creating policies that also meet compliance requirements. Don’t forget that these systems must be taught which data is sensitive as well as the limitations that must be applied when responding to a user request.

A three-point data preparation plan

The big question to ask at this point, however, is how? What are the foundational steps that organizations should take to ensure their data is correctly prepared for integration with AI? Generally speaking, there are three key priorities that should be addressed:

1 -- Assess data maturity

The process should begin with a thorough assessment of the existing levels of data maturity. Key criteria include the extent to which the organization’s datasets are trustworthy and automated across any relevant areas, from financial and CRM to inventory data. In particular, are these datasets consistent, reconciled and reliable? 
The assessment should also identify to what extent manual, error-prone processes, such as those reliant on tools such as Excel spreadsheets, remain in use. Organizations in this situation are less likely to be ready for AI and will need to improve their data management before moving forward.

2 -- Implement a modern data platform

Based on the outcome of the data maturity assessment, the next step is to create an action plan that prioritizes the most critical improvements needed to prepare for AI.

The focus here should be on building or enhancing a data platform that delivers trustworthy and insightful analysis. This platform should be capable of fully automating data collection, processing and reporting, while also eliminating the need for manual intervention. This includes considerations such as transitioning from descriptive and diagnostic analytics (understanding what happened and why) to a situation where the organization can confidently base decisions on their data.

Organizations looking to update or replace their data platform should choose a solution that can scale with their needs, particularly one that can support more advanced analytics and AI as their level of data maturity grows.

3 -- Leverage AI and advanced analytics

Once the data platform is in place and delivering reliable, automated insights, the organization is in a strong position to begin leveraging AI and advanced analytics.

Assuming that the relevant datasets are trustworthy and well-managed, the process of developing advanced analytical models and AI-driven applications can get underway. This can address a wide variety of requirements, from integrating AI with existing data systems to enhance decision-making, optimizing operations or offering predictive insights. In each situation, the data platform used for current analytics should also be capable of supporting these advanced use cases.

The key point to remember is that organizations should systematically prepare their data infrastructure rather than jumping straight into AI integration without first doing the necessary groundwork.

What’s the hurry?  

Among the various discussion points that come with the growth of AI is the need for urgency, which is also true for data strategy. For organizations that have the time and resources, this process should already be underway, and if it isn’t, it should begin immediately. 

Granted, those who delay getting their data organized or don’t move to modern, cloud-based platforms can certainly come back to it in three to five years' time, for instance. The problem is that by then, they will be starting from scratch and will still need to go through the various important data preparation and integration processes. They might also be competing with rivals who have built experience, momentum and a head start that could deliver a range of significant advantages.

The end point of true AI integration therefore may feel like a utopian aspiration, but as with everything it can and should be broken down by organizations into the steps to take them there. They should address the necessary fundamentals today, to enable them tomorrow to capitalize on the AI technologies available.

Reuben Barry is Practice Director -- Data & AI at Node4.