Analytics is more than just crunching numbers
Organisations continue to invest heavily in Artificial Intelligence (AI) and Advanced Analytics (AA) to sharpen decision-making and reduce response times. From processing petabytes of information to reducing manual effort, the potential is enormous. Yet, despite these innovations, 70–80% of AI projects fail to deliver on their promise (1).
A solution in search of a problem
Many AI and AA projects are scientific experiments in search of a problem – organisations just want to ‘use AI’ and are not sufficiently focussed on solving a real business problem. If the problem is not understood, how do you measure success? Today, as few as 5% of AI proof-of-concepts successfully scale into production, often because the use case lacked a business purpose and was unable to demonstrate sufficient value to the business (2). Analytics must be about more than just the ability to crunch numbers.
Considering data through a human context
We need to put ourselves in the shoes of real people – the users or recipients – to understand the exact insights they seek. For example, two different users may require access to the same data. The executive user may require the data aggregated and presented visually, while a power user may require raw data sets for interrogation. The core data is the same, but the outputs differ significantly.
And we also need to understand how, where, and when they seek the insights. In some cases, insights are time-sensitive, and we must prioritise speed over completeness. In other cases, insights are required on mobile devices or in remote locations with low connectivity which impacts format and presentation considerations. Understanding context allows analysts to better translate mathematical possibilities into pragmatic and seamless user experiences. Focussing on the human element brings relevance, context, and richness.
The technology exists, but the ability to craft an impactful use case which can deliver value still eludes many.
How much is enough data?
AI and AA systems rely on high volumes of data—but that’s just the beginning. Data must be collected with purpose. Just because we can gather it doesn’t mean we should. Data quality is key - incomplete, irrelevant, or biased inputs can skew models and lead to flawed decisions. Worse still, relying on poorly prepared training data or small trials can create models that collapse under pressure.
Good analytics is not about big data—it’s about smart data.
Investing in ‘good’ not ‘evil’
AI systems don’t just automate—they amplify. If biases exist in the data, AI will reflect and reinforce them—quickly and at scale. Moreover, models are never complete—they require ongoing updates, monitoring, and tuning. AI should be treated as a living system, not a set-and-forget feature. Without planning for maintenance, models drift and become irrelevant (3).
Appropriate investment
There is a common misconception that AI and AA projects run with minimal resources – this is like running a spaceship on AA batteries (4). In reality there is a significant investment of time, money, infrastructure, and a human investment. Lack of infrastructure is a commonly reported reason why AI trials never progress into production - the infrastructure simply cannot support the levels of compute required. Similarly, small data sets may be sufficient in trials, but the massive data sets required for enterprise release may be unattainable.
From use case definition and user training, to infrastructure and data acquisition, planning to ongoing model maintenance – successful AI and analytics projects require ongoing investment.
Is design thinking the key to supercharging outcomes?
While completely separate and unrelated disciplines, combining AA and design thinking principles could be the missing trick to supercharging data and analytics outcomes.
AA’s purely analytic approach could benefit from a more human perspective which highlights intangibles like context and the behaviour of users in the process. Additionally, considering solutions as part of a wider unified system, rather than a narrow component-by-component view, could add value.
Simple steps like introducing multidisciplinary teams of data and design experts to work on projects and moving to an organisation-wide view of how we use data could deliver measurable benefits.
How do I trigger a user-centric AI and analytics evolution?
Understand your users: An understanding of both users and the business itself is crucial to a successful AI or data and analytics program. Getting an unbiased assessment of the maturity of your current state and clarifying your future needs is the first step.
Develop a clear and actionable strategy: A clear and actionable strategy is your roadmap for the future. It will help you build the capability you need to deliver your business goals.
Design AI and analytic practices that can evolve with your business: To remain competitive, you must constantly evolve and respond to the changing environment. Design models, solutions, and systems which are agile, flexible, and have the capacity to rapidly scale.
Conclusion
The future of AI and analytics isn’t about crunching more numbers—it’s about solving real problems in human-centred ways. Those companies who can successfully blend advanced analytic capabilities with the needs of the business and their users will succeed. That success won’t come from the latest algorithm or shiny new platform. It will come from a cultural shift where the business and technology come together with the primary purpose of meeting user needs.
References
Let’s create solutions that deliver and last
Work with an experienced team to plan, design and deliver secure, user-centred solutions. From modernising services to improving efficiency and solving complex challenges, we help you achieve results that matter.