Composite AI and Decision Intelligence – The Competitive Edge You Need

What is Composite AI?

Composite AI is applying various AI techniques to improve learning efficiency, develop higher levels of knowledge representation, and solve varying business problems more effectively.

Gartner found that Composite AI has a 3 to 6 year adoption time frame, since it necessitates adopting decision systems platforms. This new market will take time to mature and is at least three years away. Although Composite AI techniques have been available for some time, the uptake has been sporadic.

However, many organizations are accelerating their efforts because of the ability of Composite AI to learn from incomplete or partial data sets, thus requiring less energy, which means lower costs; and it can handle “small data” workflows (high velocity) giving its users a competitive edge. This increase in activity is accompanied by capital inflows and the proliferation of businesses using Composite AI methods out of necessity. According to Gartner, composite AI currently makes up 5 to 20 percent of the way toward early majority adoption.

Companies can also use Composite AI for operational control across the entire value chain where the “smarts” of an enterprise reside. Composite AI enables us to build intelligent machines that combine human information with other forms of machine learning into one comprehensive system. The Intelligent Assistant acts as a virtual chief executive, applying Composite AI to the organization’s existing data to steer it in the right direction.

The goal of composite AI at this point is to link together “connectionist” types such as machine learning (or its subset, deep learning) with “symbolic” and other AI methods like agent-based modeling, rule-based reasoning, optimization, and graph analysis. The principles behind Composite AI are not new, but they have only recently started to materialize.
Furthermore, Composite AI understands that deep learning, graph analytics, or other “more traditional” AI approaches are not silver bullets.

Gartner found that, in addition to startups, “traditional” AI providers within established marketplaces like the data science and machine markets and knowledge graph vendors have started to integrate more AI capabilities, increasing the number of composite AI product groups available in the market.
Many business problems, especially in banking and insurance, require the capture of know-how, compliance rules, and other known knowledge from human experts in a structured manner and automatically reason with those real-world knowledge elements as a basis for new decision making.

Machine learning, rule-based approaches, and optimization methods can improve the ability to detect trends and patterns in the data. Thus, creating a composite AI system is the result of combining several technologies (among others) such as machine learning, rule-based systems, and optimization procedures.

Mass

The Mass is high because organizations need to make decisions faster or risk becoming irrelevant within their markets in a more complex business landscape. Time is money and the ability to automate repetitive aspects of back-office operations will free employees up to take on more value-adding tasks.

Traditional AI applications bring many benefits that Composite AI can also deliver across business functions, including:

• Customer service rep automation
• Reduced operational costs related to compliance/regulation updates
• Faster market research insights than competitors thanks to machine learning capabilities applied over fast data streams.

In addition, composite AI is required to deal with complicated logic or provide answers that are hard to express in natural language. In some cases, the best method to describe (and code) the solution necessitates using a variety of techniques to both capture business logic and learn from experience. Thus, composite AI is the next stage in AI’s growth.

Composite AI functions are helpful for any industry, sector, or company process. They do not take the place of current investments but rather enhance them quickly, hence justifying their usage. However, integrating AI technologies into existing application ecosystems is a difficult task. Despite that, combining AI methods lowers technical debt while simultaneously allowing developers to work at higher abstraction layers and benefiting from better segmentation of the knowledge gathered or required. This, in turn, opens the door to a larger range of stakeholders who can “code” their expertise more effectively, therefore, widening Composite AI’s appeal.

The days of relying on a single type of AI are drawing to a close. Software and service vendors that can’t provide solutions that incorporate various AI methods (such as machine learning, rule-based systems, optimization procedures, knowledge graphs, natural language technologies) will find themselves at a disadvantage compared to those who can. The introduction of composite AI methods into current products will have a significant impact on their capabilities.

Recommended Actions

One of the key takeaways from the report about Composite AI is to apply decision management with business rules, knowledge graphs, or physical models in conjunction with ML models to leverage domain knowledge and human expertise to give context and complementation to data-driven insights.

Use Cases for Decision Intelligence

Common use cases for decision intelligence in banking and insurance include portfolio optimization, personalized customer experiences, and risk management. Even though these decisions are often already supported by data, adding more data delivers greater value. For example, even though an insurer may have excellent data about the type of products offered to customers, this information alone is not detailed enough to provide personalized service. The missing link is an insight into whether or not any given customer has an affinity to certain products. So, the insurance company can set up a “data lake” containing information on, for example, whether the customer recently bought a professional camera or has been active in a Facebook group focused on “wildlife photography.” Based on these kinds of real-life touchpoints, Machine Learning DI systems can determine whether the client has an affinity for certain insurance products such as Equipment and Camera Gear Insurance and Personal Accident. And it works well in many situations. It would be virtually impossible for a person to write the necessary business rules manually with many data sources.

The unpredictability of company decision models’ long-term viability, which is based on the degree of transparency, is one of the significant problems that Decision Intelligence faces. Therefore, to avoid negative consequences and greater expenditures, autonomous decision models require more attention at design time.

Mass

The confluence of several technological clusters, including composite AI, smart business process, decision management, and advanced personalization platforms, is establishing a new market for decision systems platforms that support the DI discipline.

Many companies operate as siloed, disconnected teams. Local optimizations are sometimes favored over broader efficiency. Decision intelligence methods promote the transparency, interpretability, fairness, dependability, and accountability of decision models, essential for human and machine collaboration.

Risk management, particularly in the banking and insurance industries, is becoming increasingly essential as regulations tighten, with privacy and ethical standards to new laws and government demands, and decision intelligence aiding organizations in fully comprehending the risk consequences of their choices. Decision intelligence allows for an explicit representation of decision models, lowering this risk.

It can be applied to any organization of any size and industry because the discipline of making consistent and repeatable decisions is relevant to all. However, in practice, larger organizations with more heterogeneous landscapes and higher technological debt will have a greater need to adopt decision intelligence practices to ensure consistency and repeatability in decision making.

Decision intelligence as a discipline pushes the company to focus on value and prevents executives from being tempted to focus on nice-to-have but ineffectual data because all of the analytical attention is focused on enhancing decisions. It also increases the consistency of decisions, which, while present in operational matters (such as financial credit decisions), is often absent in strategic ones (such as where to open a branch or which low-profit services to cut). Improving the repeatability of important decisions is critical for organizations to learn from challenging situations and help them become truly data-driven.

Long-term strategic decisions will require a different collection of technology components that enable operational decisions via real-time intelligence. As a result, there is a demand for a wide range of technology components to assist decision intelligence across tactical and strategic judgments.

DI’s primary benefit will be to help organizations leverage data and analytics and AI-related components (machine learning) by providing a broad framework for businesses.

Recommended Actions

The key recommended actions on DI from the Gartner report include:

• Consider focusing on expense-cutting and increasing automation in low-value, low-risk decisions as a precursor to more advanced decision intelligence.
• The capacity to learn from past actions and apply knowledge will be well served by a composable framework, which allows for composability principles.