By Charlie Farah, CTO, Qlik
It’s that time of year again when businesses reflect on the past 12 months as well as planning and strategising for the next and trying to predict what will be the key business hacks to increase efficiency and competitive advantage. In 2025, as foundational Artificial Intelligence models reach performance plateaus, my prediction is that the greatest potential for AI in business won’t lie in the next big external tool but within the data businesses already possess.
As organisations have raced to adopt AI, many have overlooked the wealth of internal, unstructured data – emails, call transcripts, and product documentation – that can drive truly transformative insights. Business data will become the fuel that propels AI forward, but it’s not just any data – it’s proprietary, real-time, and well-integrated data that separates leaders from the rest. Relying on gains from foundational model performance won’t cut it anymore. Today’s smartest companies are funneling proprietary data directly from dozens of sources for instant impact. Embracing this strategy will build AI systems that are more adaptable, responsive, and aligned with their specific business goals, giving them a competitive advantage in their respective industries.
Unstructured data: The hidden resource
Unstructured data, which makes up around 80% of the world’s data and is set to double by this year alone, has traditionally been undervalued because it doesn’t neatly fit into the tabular forms that analytics tools prefer. However, advances in large language models (LLMs) and Retrieval-Augmented Generation (RAG) are making it possible to finally bring this information to life, providing unprecedented opportunities to mine insights and create value from internal data. In today’s data-driven landscape, building internal capabilities isn’t just a choice; it’s becoming a strategic imperative.
The vast majority of valuable information in companies is unstructured – spread across countless emails, Slack messages, PDFs, and other formats that hold a record of the business’s activities, decisions, and knowledge. While this data was once just a byproduct of everyday operations, today it has a lot of value as a strategic resource. Tapping into it can reveal insights previously locked away, offering everything from product feedback to internal expertise and industry trends. Integrating unstructured data for AI applications requires careful planning. It’s not just about feeding this information into a model and hoping for insights. The data needs to be cleaned, categorised, and formatted to be ready for analysis by AI tools. This is where tools like Qlik Answers, powered by generative AI and Retrieval-Augmented Generation (RAG), become essential, helping organisations handle unstructured data in a way that produces relevant, reliable insights.
Building in-house vs. relying on third-party LLMs
When it comes to leveraging AI, companies face an important choice: build in-house or use third-party. While third-party solutions might offer speed and ease of implementation, they often fall short in terms of control and security. Outsourcing AI to a third-party model can expose companies to competitive risks, especially if those tools are embedded in platforms that may also serve competitors.
By developing in-house solutions tailored to their specific needs and built on their data, companies can maintain greater control over their information and enjoy a customised approach. The path to success involves selecting technology that not only meets short-term needs but also aligns with long-term business objectives. Businesses should choose a platform designed to bring the benefits of advanced AI while allowing them to keep data within their walls, reducing dependence on external vendors and improving security.
Strategic value versus speed of delivery
When deciding between in-house development and third-party solutions, businesses must weigh strategic value against speed of delivery. For rapid deployment, third-party LLMs offer a quick path to market – often helping startups reach product-market fit. However, consider a business aiming to reduce customer support load by 25% over the next year. Rather than outsourcing, an in-house model trained on the company’s own support tickets, product information, and customer interactions can yield more relevant, accurate responses, directly impacting the bottom line. By focusing on creating internal AI resources, companies can ensure that each insight is tailored to their precise goals, offering a competitive advantage that third-party solutions can’t provide.
Risk tolerance and data privacy
For many companies, the decision to build or buy AI solutions is driven by their tolerance for risk, particularly around data privacy. In highly regulated industries like finance, healthcare, and government, protecting customer and employee data is non-negotiable. If exposing sensitive data to a third-party API could lead to serious consequences, an in-house approach is likely the safer choice. Building internal AI capabilities allows companies to enforce strict privacy standards and control data access. There are solutions designed with security and compliance in mind, reducing the risk of data exposure and ensuring that sensitive information remains under the organisation’s control.
The future of data-driven innovation
As businesses increasingly look to unlock the potential of their internal data, we’re seeing the beginning of a new era where data truly drives innovation and smart decision-making. Looking ahead to 2025, businesses that prioritise internal data and build AI capabilities that are both responsive to their unique needs and secure in data privacy will be best positioned to navigate the future. My opinion is that the decision to look inward is more than just a strategy – it’s a necessity for businesses aiming to build resilient, innovative futures in a data-driven 2025.