Development of the Internet Of Things (IOT), advances in Natural Language Processing (NLP) and Generation (NLG), spread of automation strategies (Robotic Process Automation, RPA), … At the start of this new year, many Artificial Intelligence (AI) experts have been asked what the most significant trends in AI will be in 2020.

 Is AI becoming mainstream ?

Without a doubt, the power and capabilities of AI tools will continue to grow significantly in the short term. However, the most noteworthy shift in AI might not come from the release of better tools, but from a better business integration of what is currently existing. One of the most significant trends in AI for the next year, and certainly for the next decade, will be a wider adoption of AI in the decision-making processes of companies. Indeed, we are currently located at a critical junction in terms of adoption of AI technologies, due to the existence of several factors:

• The need for intelligent analytical tools: The ever-increasing competition in many sectors incentivizes companies to look for innovations that can enhance their decision-making. On the other hand, the growing digitization of our society provides a massive amount of data that can enrich this decision-making. Companies crucially need tools capable of harnessing these data and extract relevant insights.

• The maturity of (some) AI technologies: Many AI projects have been implemented these past years inside organizations. Some successfully delivered added-value, others did not. Nevertheless, case studies now exist. They enable to critically assess what AI can and cannot do, and the key factors for a successful AI project. In addition, the increasing use of AI in daily life (e.g. chatbots, spam filtering, audio/video content recommendations, autonomous cars, etc.) and the growth in the digital maturity of decision makers are easing barriers to adoption.

• The availability of AI tools: The right tools to successfully implement AI in companies exist: from libraries that facilitate the development of algorithms, to frameworks and connectors that ease the integration of these algorithms into existing business processes. Experimented data scientists have nowadays a toolbox that allows them to quickly and efficiently solve business needs and to deploy AI projects in any business environment.

Key success factors for implementing AI in a business context

A more generalized adoption of AI tools in companies can only occur over the long term if the first wave of new projects generates of a sufficient return on investment. We are convinced that several elements are crucial for an AI project to generate added value:

• Think long and hard about how the project will be integrated within the existing IT and business processes: One of the keys to a successful AI project resides in an efficient management of its whole lifecycle. In the past, too many AI projects have remained stuck in their prototyping phase without never been put in production. To avoid that, it is vital to conceive an AI project by considering its lifecycle beforehand, from the definition of its scope (and the goals to be achieved, realistically) to its deployment, integration, and maintenance. While the deployment phase was not paid a lot of attention in the past, this phase is not understood as primordial since its goal is to bring the value of the AI modeling to the project’s end users. Nowadays, many tools and platforms enable a quick and efficient deployment. For instance, on Qlik, AI algorithms developed on Python and R can be easily deployed using the built-in Advanced Connector. As a result, AI projects can be put to work in a reduced amount of time, while being seamlessly integrated with the existing business tools of a company. This in turn provides an ideal ground for a wide-scale adoption of the developed tool in the company.

• Build transparent AI solutions: One of the most important hurdles in the business adoption of AI is the blackbox sentiment that end users can feel. It is therefore paramount to develop AI solutions that consider the human factor. One way of doing so is to build algorithms that are explainable and provide results that are understandable by end users (Explainable AI paradigm). On the other hand, a way of promoting the adoption of AI solutions is to enable users to easily interact with the solutions, without resorting to data scientists as intermediaries. This can be achieved by integrating the AI tool in a business tool known by the end user (e.g. Qlik), or by implementing chatbots or virtual assistants that can bridge this interaction.

• Use high quality data and an integrated data stack: Outputs of an AI project can only be as robust as its inputs. A successful project relies on high quality data. It is therefore key to integrate a new AI project into a company wide data strategy and a holistic analytical approach. Optimally, an integration with existing Business Intelligence tools can facilitate this integration.

Do not hesitate to contact us if you wish to know more about these topics. We can organize a free brainstorming session to think together about how AI can be applied to your environment and the return on investment you can expect!