Perhaps this is heresy, but is there a risk that some companies are pursuing data analytics and AI projects without a clear enough business objective?
Let me give you an example. We recently heard about a marketing use case where the data analytics team came up with some truly interesting customer micro-segmentations. They saw the clear possibility to open up new services and new ways of marketing. They also saw the potential for additional revenue.
However, those possibilities were never actually accessible: The organization’s global marketing strategy had embarked on precisely the opposite course – simplifying customer segments as a route to cost savings.
You might imagine such a set of circumstances could never happen where you work. But is that such a safe bet? Larger enterprises have been investing in the technology to access and interrogate the ‘data lakes’ created by digitalization for several years. Typically, they have teams of data scientists who are highly skilled at extracting meaning and value from the data. The temptation to pursue the possibilities – because you can – is very strong.
Stop. Think. Talk.
Considering my job title, you might not expect this, but I think we need to step back from data analytics for its own sake. It’s time to focus on converting data insight into meaningful improvements to business processes rather than pursuing new technical possibilities because the technology is available and proven.
There are two dimensions here. The first – and theoretically simpler of the two – is clarity of purpose. Does the project fit your company’s broader strategies and objectives? If it doesn’t, it’s time to reconsider. This is only theoretically simpler because show-stopping obstacles sometimes only come into full view after the event.
The second dimension is the ability to do something with your insights when you find them. This means integrating data, usually in real-time, into business processes invented in other parts of the organization. These were probably designed for very different purposes. A further challenge is to fit into systems you might not have developed or own – very likely in the cloud. Clearly, it helps if any hurdles to achieving this – technical, budgetary or even political – have been anticipated in advance.
One reliable way to accelerate and de-risk the process is through planning and discussions with partners in both cases. They will have seen risks and bottlenecks in other contexts and with other customers, although these might be new to your team.
MLOps – a better way to integrate data analytics and AI insights into business processes
Something else you will want to consider is MLOps.
Improving decision making as well as businesses operations through AI and analytics it is not just about how to apply algorithms to solve certain problems. It’s also about how you embed them into your core processes. Fortunately, there is a rapidly maturing body of tools and processes to ensure this happens quickly and successfully. It is based on collaboration and communication between data scientists and operations professionals to help manage production ML. This is MLOps. As its name implies, it is similar to DevOps or DataOps and approaches and enables the application of agile principles to ML projects.
MLOps started as a set of best practices and is evolving into an independent approach to ML lifecycle management. It aims to increase opportunities for automation and improve the quality of production ML while also focusing on business and regulatory requirements.
As well as upping the pace of development, MLOps together with cloud hyperscalers, makes it easier to take data insights and AI into the next level, mainly meaning we can now scale them and build a data intelligence factory for mastering business efficiency and decision-making.
Data analytics in business use
A framework approach is something we advocate at Fujitsu and we have applied this to our own data analytics and AI offerings. Our framework for computer vision and natural language processing recognizes that customers will be on a learning journey with AI implementations and do not want mono AI capabilities that only do one thing. A framework provides the tools to adapt and expand the application of AIs into the future as new possibilities come to light.
Fujitsu’s framework approach lets customers access the necessary technology at a sensible cost, using modular components. They build experience and confidence with successful value creation projects and then adapt the components – often using their own people – to new uses without significant new investment.
Data always lies at the heart of these projects. Recently, a security services multinational chose Fujitsu to assist the company in industrializing AI uses cases being developed by different analysts and data scientists in various business areas and embedding them to existing business processes.
Fujitsu is also helping a European drug administration agency in scaling up and industrializing NLP processes which structure data coming from clinical notes using the cloud and analytics automation engines. The results are used for pharmacovigilance purposes (tracking the side effects of the COVID-19 vaccination process).
As these examples show, the technology behind data lakes is no longer the main issue. We have reached the point where business understanding is equally important. Therefore, ecosystem selection must refocus on partners with track records delivering services at scale, integrated into complex, wider business processes. That’s an area where Fujitsu excels. If you would like to explore the possibilities to take data analytics and AI insights into your organization's business processes, contact us today, or for more information visit our Sholark webpage and our AI webpage.
