What Is Data Modernization And How Does It Affect My AI Strategy?

Play Audio Version

Artificial Intelligence (AI) has consistently been one of the top trends in technology over the past few years. If you look at Google Trends to see the history of what people have searched for online then it is clear that in 2022 there was very little interest in AI – interest started growing in 2023, but more recently it has grown exponentially.

However, several major studies, including one from last year published by MIT, suggest that many companies are struggling to implement AI in their business. They can’t scale up from a pilot to the enterprise – it just doesn’t work on a larger scale.

The MIT research suggested a 95% failure rate.

This is not addressing the general use of AI. Nobody is suggesting that they can’t get Microsoft Copilot or ChatGPT to work. The issue is the bespoke use of Generative AI to create new services or a new way to access data inside the business.

The most important problem is data.

Anyone can fire up ChatGPT and ask it some questions. OpenAI have already done the training and they are managing the database that sits behind the chatbot.

But what about if you want to use AI to analyze your internal inventory data, or delivery history, or customer service requirements? When you need to start training an LLM on data that is specific to your business then you need to think more carefully about how your data is organized.

This is even more important if you want to progress to use some agentic processes. If you plan to automate some internal business processes with Agentic AI then you effectively need to create a data layer for your business. It’s not going to work if some data is stored in the marketing database and some is in the CRM system used by the sales team.

Most companies face this problem. Important business processes are decided based on data that only exists inside an Excel spreadsheet. Even where the data is a little more open and structured, it remains common for different departments to have no access to the information each one is storing.

How do you automate processes that involve your suppliers, employees, and customers when the data exists in a dozen different databases that are not connected to each other?

This is why so many companies have tried a limited pilot using AI, but then found it difficult to scale the benefits across an entire business. It looks impressive when your AI system is generating insight into what clients need, but once you try scaling that up to the entire enterprise it just doesn’t work.

If your business has experimented with AI, even managed to successfully run some pilot projects, then you need to think about these different subjects before pushing AI to the entire enterprise:

• AI is more than just a software tool: don’t just bolt a chatbot onto existing processes. Think about how you can change your business processes, especially using new data flows.
• Data foundations: you need clean, structured, and accessible data – especially once you start embracing agentic. How do you move on from the departmental silos and systems that are not connected?
• Ownership: AI pilots are often led by an innovation team or an individual department exploring change only in their area. How are you building a roadmap and governance model for real improvement?
• Change: if people are ignoring your AI tools or even fearing their job may vanish if they use AI then you need to develop training and explanations that can encourage team members to improve what they are doing – with AI.
• Security: once you start designing governance for your AI strategy then you can start crafting a security and compliance strategy.

Each business is different, but the most concrete area where you can usually start building a beachhead is data. How can you start connecting systems together and building a data layer that encompasses everyone and every system in your enterprise?

As you start on this journey of data modernization, you can start building out the ownership and security strategy, but without this planning for the foundations of an AI strategy you will be lost and just a new statistic in that MIT report on project failures.

Embracing AI requires much more than a willingness to run a few limited pilots. This sounds negative, but on the positive side there are many case studies available now that demonstrate the best way to reorganize data in a way that supports the rapid enterprise-wide adoption of AI tools. It has been done. There is a roadmap.

Modernize how your enterprise uses data today and you can help it step into the future of AI tomorrow.

Click here for more information on data management, analytics, and AI services from IBA Group.