Setting Up Your AI Pilot Project for Success

AI is automation taken to a new level. Automating jobs that people now perform using supervised learning technology will be a valuable source of inspiration for AI initiatives. You’ll discover that it is easier to employ AI to automate chores than jobs. The low-hanging fruits in the short run are therefore made up of parts of the individual duties that humans perform. This and an analysis of whether any of them can be automated should thus be the first step.

If, for instance, an oncologist’s duties include interpreting X-ray images, operating imaging equipment, consulting with co-workers, chemotherapy treatments and organising surgery, you would need to consider automating or partially automating only one of the duties to start with rather than attempting to automate the entire job.

Prior to launching an AI pilot project, we would advise outlining the expected timetable and the intended objective(s) in painstaking detail and subsequently giving the AI team a suitable budget to work with such that they will not have to cut any corners.

Selecting a leader who can be the project champion is also important. You will need to pick a person who believes in the technology and in the benefits it offers and who can bridge the gap between domain specialists in your industry and cross-functional AI. In this manner, if the project is successful, it will have an impact on the company as a whole. Their overarching objective here should always be to create a successful project that will have a profound impact on the way the organisation views and understands AI. This should be the first step towards creating future AI initiatives, and not to create an AI start-up.

Performing business value and technical due diligence is important. Make sure the business executives concur that the project will provide enough value for the company if it is carried out properly. You should also confirm the project’s viability. A technical team must review the data you already have and maybe conduct small-scale tests as part of the technical diligence process, which might take a few weeks.

In order to engage with consumers and manage everything in the company, from customers and clients to inventories and products, every big company currently handles a proliferation of websites, applications, and technological platforms that are continuously generating data. However, firms still struggle to leverage such rudimentary data to boost fundamental business operations that could give them a competitive edge, cut costs, and enhance customer service, despite numerous generations of investments and millions of euros in digital transformations. AI was envisioned as a solution for this. However, every big businesses in almost all economic sectors have spent at least a few million euros on unsuccessful AI programs and it is very important to ask why.

From our experience, the promises made by many AI service providers are ineffective unless a company’s data infrastructure is adequately ready for AI. Data is very often, if not invariably, not organised in a way that can be readily utilised as the fuel for AI since it is locked up in separate silos, is unavailable, has poor structure and – most importantly – is not readily accessible. Instead, businesses must invest heavily in developing a complete data ontology in order to be able to partake in the benefits that AI can bring about. This is a comprehensive characterisation of the architecture of all of the organisation’s data, and it is usually a huge task – much bigger than most organisations, even those who know their data quite well, are ready to acknowledge.

You may have been given recommendations to launch AI projects slowly and with restricted budgets and this might also work for some time, but the limitations of dispersed trials become increasingly painfully obvious when AI programs eventually proliferate throughout the enterprise. This approach will get entangled in complexity as soon as you start feeding AI algorithms a variety of (sometimes incompatible) data sources. Your current data systems will soon be connected to a plethora of one-off AI pilots that might have scored their intended goals but that bring very little added value at the higher, strategic level and will thus fail to serve your organisation in a broader, more strategic sense where the biggest benefits are likely to arise. This is why earlier on in this insight, we have suggested that starting small could work but should always be undertaken in view of a complete roadmap, which can never be thought of on a small scale.

AI has now advanced to the point where a comprehensive strategy – a key that unifies all of your organisation’s data – is required. That is where the data ontology comes in. A model of all the components that go into and connect your various information systems, including the goods and services, solutions and processes, organisational structures, protocols, customer characteristics, manufacturing techniques, knowledge, content, and data of all types, is called an ontology. It serves as the organisation’s knowledge masterplan. AI systems can only evolve in a piecemeal, fragmented fashion without a consistent, intentional approach to creating, using, and expanding a data ontology; they will lack the foundation that would enable them to be intelligent enough to have an impact. An investment that will continue to pay off as AI becomes more prevalent is the ontology, which is at the core of the information design of any AI-powered organisation.