While there is no shortage of hype surrounding artificial intelligence (AI), businesses are beginning to realise just how difficult it can be to deliver something useful with the technology.
The headline-grabbing revenue of chipmaker Nvidia is an indicator of how much money is being spent on building infrastructure to support AI inference and machine learning workloads, but in reality, many businesses are struggling with the technology.
Nvidia’s latest financial results show that its datacentre business posted $26.3bn revenue in the last quarter, much of which came from sales of AI acceleration hardware to cloud providers, consumer internet businesses and enterprises riding the AI wave.
However, a survey from Sapio Research for Hewlett Packard Enterprise (HPE) found that only one-third (32%) of IT leaders in the UK and Ireland believe their organisations are fully set up to realise the benefits of AI.
The poll of 400 IT leaders in the UK and Ireland for HPE’s Architect an AI advantage report found that while commitment to artificial intelligence shows growing investments, businesses are overlooking key areas that will have a bearing on their ability to deliver successful AI outcomes. These include low data maturity levels, possible deficiencies in their networking and compute provisioning, and vital ethics and compliance considerations.
Matt Armstrong-Barnes, chief technologist for AI at Hewlett Packard Enterprise, said: “Businesses are investing in AI without first taking a holistic view of the technology and how to implement it. Diving in before considering whether they are set up to benefit from AI and who needs to be involved in its roll-out will lead to misalignment between departments and fragmentation that limits its potential.”
The survey is among a number of new pieces of research that show the gap between AI hype and reality in terms of business outcomes. A separate KPMG poll of UK business leaders reported that despite the promises of generative AI (GenAI), there are significant concerns about its implications for business performance.
The business leaders polled in the KPMG survey cited the inaccuracy of results, including hallucinations, as the biggest concern when adopting GenAI (60%). According to KPMG, boards are also worried about errors in the underlying data and information skewing the model’s outputs (53%), as well as problems related to cyber security (50%).
KPMG also reported that only 30% of the directors polled said responsible GenAI usage guidelines have been published and communicated throughout their organisation to mitigate these potential issues.
It’s important that companies thoughtfully define a clear AI strategy rather than merely chase the next technological innovation. This strategy should balance the value, cost and risk associated with AI use cases
Leanne Allen, KPMG UK
Leanne Allen, head of AI at KPMG UK, said: “Given boards’ concerns, it’s important that companies thoughtfully define a clear AI strategy rather than merely chase the next technological innovation. This strategy should balance the value, cost and risk associated with AI use cases. This strategic equilibrium is crucial for both progress and stakeholder trust.”
One challenge in making a success of AI is the quality of data. HPE’s research showed that data maturity among survey respondents remains at a low level. It found that only a small percentage (6%) of organisations can run real-time data pushes and pulls to enable innovation and external data monetisation, while just 29% have set up data governance models and can run advanced analytics.
HPE also reported that fewer than six in 10 respondents said their organisation was completely capable of handling any of the key stages of data preparation for use in AI models – from accessing (57%) and storing (51%), to analysing (54%) and processing (52%). According to HPE, this discrepancy not only risks slowing down the AI model creation process, but also increases the probability the model will deliver inaccurate insights and a negative return on investment (ROI).
Further research, from Vanson Bourne for Fivetran, presented in the AI in 2024 – hopes and hurdles report, found that almost all (97%) of the 550 organisations surveyed faced barriers in their AI adoption. The study reported that only 40% of the IT leaders polled measure the ROI of their AI programmes fully.
Significantly, the Vanson Bourne survey reported that on average, businesses lost 6% of their global annual revenue due to misinformed business decisions based on AI systems that use inaccurate or low-quality data.
However, Vanson Bourne also found that organisations in the starter phase of AI adoption are seeing a 62% ROI, on average, indicating the financial benefits start high before suboptimal data leads to underperforming AI models and negatively impacts financial return.
The study reported that, on average, financial ROI is slightly higher (56%) for those who build their own AI models, compared with those using third-party suppliers, be that open source (52%) or closed source (51%).
“If businesses continue their current approach to AI, it will adversely impact their long-term success,” said HPE’s Armstrong-Barnes. “They must adopt a comprehensive end-to-end approach across the full AI lifecycle to streamline interoperability and better identify risks and opportunities.”
Read more data quality stories
AI’s effectiveness is limited by data quality. Building strong data management and governance programmes is crucial to handling the challenges that AI presents to organisations.
Without quality data, creating and managing AI systems is an uphill battle. Methods such as zero-copy integration and primary key consistency can ensure trusted data for better AI.
Read more on Data quality management and governance
