The Real AI Investment – Prioritizing Data Quality Over Model Hype

Forbes

15-05-2025

AI Hype
We are almost three years past the fanfare of ChatGPT's big debut and the emergence of generative AI (GenAI) models in the workplace. With broader accessibility to the technology, and the bell curve of the hype cycle pushing early operational innovations – like GenAI-enabled virtual assistants – over the hump of inflated expectations, it is worth taking a thorough look at where the market is today and what can be improved.
Worrying over models distracts from data issues
Enterprise adoption of GenAI has surged in these last few years, with more organizations integrating GenAI tools into their operations and deploying them in production environments. Yet, even as usage expands rapidly, many companies continue to grapple with uncertainties and concerns tied to the arrival of the latest GenAI models.
When a new, exciting GenAI model gets released, many businesses fall distressed over which model is best and if they should jump ship to a 'better' one. In fact, according to recent survey findings from Qlik, a global data, analytics and AI company, nearly half (47%) of AI professionals now worry their organizations has overinvested in costly, inefficient models, leading them to reassess their AI approach. However, placing significant weight on choosing one model over another is a misstep.
Industry disruptions like DeepSeek will continue to arise with offers of 'better' or 'cheaper' models. For a business to reassess its current model that satisfies its needs, just because a new, shiny option comes along, would be a material misuse of resources – and an infinite loop. Furthermore, GenAI models are quickly becoming commoditized. As this trend continues, the importance of whether a business leverages a model from OpenAI, Meta or DeepSeek will continue to diminish.
Rather than expending effort on model comparisons, enterprises should acknowledge and focus their time on the true foundation of all good AI: data quality. A company can sign up for the most state-of-the-art technology to power business processes and strategies, but if its data foundation isn't sound, it will be impossible to make any meaningful progress.
'As AI projects continue to drive excitement, growth, and possibility in the enterprise, it can be easy to get swept up in the hype,' empathizes Mike Capone, CEO of Qlik. 'However, one bad, rushed experience can erode organizational trust in AI and deflate enthusiasm for future innovation.' To succeed well into the future, leaders need to first understand GenAI, its capabilities and its limitations, and then take a bottoms-up approach to discovery and implementation. This starts with a strong data foundation.
Use cases beyond operations
According to a recent Prosper Insights & Analytics survey, over 40% of executives report they're already using GenAI.
Prosper - Heard of Genetrative AI
The current scope of GenAI in the enterprise today, however, largely focuses on operations – acting as a virtual assistant, automating repetitive tasks, amplifying customer service, and improving logistics. The aforementioned Prosper Insights & Analytics survey reveals that research and writing are other popular areas where GenAI comes into play. This is where the cautionary tale of an 'operational trap' comes in.
Prosper - Use Generative Artifical Intelligence For
Businesses will limit the technology simply out of comfort, lack of understanding, or a little of both. While streamlining operations is a legitimate business goal, GenAI has the potential to provide much more value – especially when considering use cases like augmented analytics to drive better decision-making at the executive level.
A business can apply AI techniques and algorithms to automate analysis processes, analyze, and interpret data, derive insights, and make recommendations. AI-augmented analytics has two main advantages over traditional analytics: first, it can leverage advanced algorithms and machine learning models to handle complex and unstructured data, identify patterns, and make accurate predictions. Second, it can accomplish analytics tasks significantly faster than a human could. Things like data processing will cease to take up the majority of a data professionals' day – freeing up their time to focus on more strategic tasks. This nirvana state, however, can only be accomplished if issues like data quality are addressed first.
The current state of AI data quality
'The foundation of any strong AI investment is data quality, which business leaders unfortunately under-prioritize,' Capone commiserated. The aforementioned Qlik survey showed that AI initiatives in the enterprise are at serious risk due to poor data quality and ROI expectations. While AI adoption is accelerating and AI investment is increasing, most companies do not have the proper data foundations to fully capture the value and capabilities of AI.
This isn't just a management problem – it is one that can be a company-wide plague. Four out of five (81%) professionals who deal with data, analytics, and AI models know their organization has more to do to overcome data quality issues, with either a moderate or a large amount of work. Furthermore, concern about data quality is the top reason organizations are limiting their employees' usage or outright access to GenAI tools today, according to a 2025 research report from Avanade, global professional services company.
Building AI on flawed data will, at best, produce unreliable results. At worst, companies will endure financial waste and increased business risk from biased models, unreliable insights, and poor ROI. According to Gartner, at least 30% of GenAI projects will be abandoned after proof of concept by the end of 2025, due to poor data quality, inadequate risk controls, escalating costs or unclear business value.
Those closest to AI implementation are taking heed of data quality issues, but in a lot of cases, without the serious support of executives – 90% of directors and managers believe leadership is failing to focus on the issue, Qlik found in its aforementioned AI survey. This juxtaposition between a workers' lack of confidence and executives' overconfidence in AI strategy is creating a tension that must be addressed to harness everything that GenAI has to offer the enterprise.
One parting piece of advice
The single most important action that business leaders can implement today if they truly want to make gains from their GenAI investment is focusing their energy on data quality first. Think of it this way: if you want to ride a bike, you had better first make sure that there's air in the tires and grease on the chain.
We have yet to tap into the full power of AI. Those who make strategic and decisive moves to build a trusted data foundation today will be the ones that harness its advanced applications tomorrow, driving a competitive edge and a sustainable, more efficient business well into the future.