There is perhaps no emerging technology that has generated as much attention in recent years as artificial intelligence (AI). Over the years, AI technology and underlying machine learning (ML) models are matured enough to be one of the mainstream technologies that enterprises are looking to leverage. The majority of organizations have started exploring AI/ML in some way, including investments in low code/no code platforms and expensive data scientists. However, the majority of effort is limited to research and development work that tends to be shelved within 6 months, and does not yield the anticipated business value given the complexity of the technology. The what and the how of this problem-solving technology remains to many as mysterious as magic. This leaves enterprises in a tough position.
On one hand, there is an incentive for greater investment into intelligent automation. Yet articulating an enterprise AI strategy that maximizes ROI through targeted, high-value use cases remains difficult. This challenge is known as the magic wand trap. When AI is viewed as a magic wand that can be waved over problems, it becomes harder to identify the areas and processes in which AI can be successfully applied to achieve maximum business value. Thus, despite high levels of interest and investment in AI, the question for many enterprises very much remains: “How can AI be leveraged in systemic fashion to provide tangible benefits to our organization?”
There is a clear need for enterprises to shift their perception of AI from a “magic wand” to an “intelligence toolkit”. Like many emerging technologies, AI is not a catch-all solution to an organization’s inefficiencies. In fact, AI is not even really a single technology at all. It is an umbrella term used to describe a set of similar tools and techniques that mimic human learning in solving problems. Different tools in the AI toolkit can solve different problems in different contexts, and these tools work best used in the contexts for which they were designed. One wouldn’t use a hammer to work with screws. That said, when these tools are combined and customized thoughtfully with human centric design approaches, AI solutions can offer tremendous business value across an end-to-end business process. As with every tool – intelligent or not – AI’s value comes with where and how you use it.
At CGI, we often say that we are not in the business of AI models, but providing solutions to the business challenges that AI combining with other technologies can solve more efficiently. Rarely in practice do we see a situation where a single model can be applied to a complex business process. Rather, it is most often a collection of different AI models, data processing, business rules, and human-in-the-loop validation, combined into a single end-to-end workflow that constitutes an intelligent automation of a business process.
This solution-centric approach exposes another common misconception of enterprise AI; namely, the common (yet misguided) preference for a general-purpose rather than a customized solution. While a truly general-purpose AI would indeed be transformative when harnessed correctly, no such system yet exists or is likely to exist in the near future. As the technology currently stands, it is depth rather than breadth that offers the highest performance. In other words, a one-size-fits-all solution is really no solution at all. It takes both domain and technical expertise to customize and tailor a solution with a human-centric approach in such a way that it automates complex business processes with a high degree of performance and ease of use. A generic, general-purpose model will do little to offer enterprises real productivity gains, as the accuracy of the models in production will be too low for a human to take their eyes off them.
CGI’s AI-based solutions have been developed with a clear understanding of these pitfalls. To counter the magic wand trap, CGI’s Enterprise AI Framework classifies AI’s value to enterprises across three key dimensions:
Operational AI: Focuses on cutting costs and improving operational efficiencies by automating operational processes.
Strategic AI: Focuses on generating revenue and growth opportunities by personalizing customer experiences and orchestrating customer journeys.
Transformational AI: Focuses on generating insights from the wealth of unstructured data within organizations.
From a technical perspective, these AI dimensions leverage different tools in the AI toolbox in addition to other tools like RPA and BPM. For instance, operational AI is oriented around unstructured data processing though text, images and audio analysis, while strategic AI is more concerned with decision augmentation/automation, personalized customer interactions, and next-best-action predictions. CGI’s emphasis on solution-centric AI has manifested itself in some other exciting developments, including the release of CGI PulseAI, an AI-based IP for accelerating the development and deployment of high-performing business solutions leveraging AI across both operational and strategic domains.
At the end of the day, AI solutions are not merely technical solutions. Industry-specific expertise plays just as critical a role in a project’s success as high-quality engineers. In proper application, AI is far from magic. AI’s success is a function of both the quality of the tool and the expertise of the person using it. That is why, for CGI: “Any sufficiently advanced technology, while it may seem like magic, needs thoughtful humans and human-centric approaches behind it.”