Interest in implementing AI is truly global and industry-agnostic. However, few companies have established the foundational building blocks that enable AI to generate value at scale. While each organisation and industry will have their own specific challenges that may impact AI adoption, there are four common barriers that all companies tend to encounter: People, Control of AI models, Quality, and Cost. To implement AI successfully and ensure long-term value creation, it’s critical that organisations take steps to address these challenges.
Accessible upskilling
At the forefront of these challenges is the impending AI skills gap. The speed at which the technology has developed demands attention, with executives estimating that 40% of their workforce will need to re-skill in the next three years as a result of implementing AI – outlying that this is a challenge that requires immediate attention.
To tackle this hurdle, organisations must provide training that is relevant to their needs, while also establishing a culture of continuous learning in their workforce. As the technology continues to evolve and new iterations of tools are introduced, it’s vital that workforces stay up to date on their skills.
Equally important is democratising AI upskilling across the entire organisation – not just focusing on tech roles. Everyone within an organisation, from HR and administrative roles to analysts and data scientists, can benefit from using AI. It’s up to the organisation to ensure learning materials and upskilling initiatives are as widely accessible as possible. However, democratising access to AI shouldn’t be seen as a radical move that instantly prepares a workforce to use AI. Instead, it’s crucial to establish not just what is rolled out, but how this will be done. Organisations should consider their level of AI maturity, making strategic choices about which teams have the right skills for AI and where the greatest need lies.
Consider AI models
As organisations embrace AI, protecting data and intellectual property becomes paramount. One effective strategy is to shift focus from larger, generic models (LLMs) to smaller, customised language models and move toward agentic or compound AI systems. These purpose-built models offer numerous advantages, including improved accuracy, relevance to specific business needs, and better alignment with industry-specific requirements.
Custom-built models also address efficiency concerns. Training a generalised LLM requires significant resources, including expensive Graphics Processing Units (GPUs). Smaller models require fewer GPUs for training and inference, benefiting businesses aiming to keep costs and energy consumption low.
When building these customised models, organisations should use an open, unified foundation for all their data and governance. A data intelligence platform ensures the quality, accuracy, and accessibility of the data behind language models. This approach democratises data access, enabling employees across the enterprise to query corporate data using natural language, freeing up in-house experts to focus on higher-level, innovative tasks.
The importance of data quality
Data quality forms the foundation of successful AI implementation. As organisations rush to adopt AI, they must recognise that data serves as the fuel for these systems, directly impacting their accuracy, reliability, and trustworthiness. By leveraging high-quality, organisation-specific data to train smaller, customised models, companies ensure AI outputs are contextually relevant and aligned with their unique needs. This approach not only enhances security and regulatory compliance but also allows for confident AI experimentation while maintaining robust data governance.
Implementing AI hastily without proper data quality assurance can lead to significant challenges. AI hallucinations – instances where models generate false or misleading information – pose a real threat to businesses, potentially resulting in legal issues, reputational damage, or loss of trust.
By prioritising data quality, organisations can mitigate risks associated with AI adoption while maximising its potential benefits. This approach not only ensures more reliable AI outputs but also builds trust in AI systems among employees, stakeholders, and customers alike, paving the way for successful long-term AI integration.
Managing expenses in AI deployment
For C-suite executives under pressure to reduce spending, data architectures are a key area to examine. While a recent survey found that Generative AI has skyrocketed to the #2 priority for enterprise tech buyers, and 84% of CIOs plan to increase AI/ML budgets, 92% noted they don’t have a budget increase over 10%. This indicates that executives need to plan strategically about how to integrate AI while remaining within cost constraints.
Legacy architectures like data lakes and data warehouses can be cumbersome to operate, leading to information silos and inaccurate, duplicated datasets, ultimately impacting businesses’ bottom lines. While migrating to a scalable data architecture, such as a data lakehouse, comes with an initial cost, it’s an investment in the future. Lakehouses are easier to operate, saving crucial time, and are open platforms, freeing organisations from vendor lock-in. They also simplify the skills needed by data teams as they rationalise their data architecture.
With the right architecture underpinning an AI strategy, organisations should also consider data intelligence platforms to leverage data and AI by being tailored to its specific needs and industry jargon, resulting in more accurate responses. This customisation allows users at all levels to effectively navigate and analyse their enterprise’s data.
Consider the costs, pump the brakes, and take a holistic approach
Before investing in any AI systems, businesses should consider the costs of the data platform on which they will perform their AI use cases. Cloud-based enterprise data platforms are not a one-off expense but form part of a business’ ongoing operational expenditure. The total cost of ownership (TCO) includes various regular costs, such as cloud computing, unplanned downtime, training, and maintenance.
Mitigating these costs isn’t about putting the brakes on AI investment, but rather consolidating and standardising AI systems into one enterprise data platform. This approach brings AI models closer to the data that trains and drives them, removing overheads from operating across multiple systems and platforms.
As organisations navigate the complexities of AI adoption, addressing these four main barriers is crucial. By taking a holistic approach that focuses on upskilling, data governance, customisation, and cost management, companies will be better placed for successful AI integration.
- Data & AI
