05.01.2022 | Blog Three AI Trends for 2022

Artificial intelligence and especially the subfield of machine learning have become indispensable helpers in some application areas. The key point here is the amount of data available for AI learning. While the platform giants collect huge amounts of data and use it to feed generalized AI methods available for generalized use cases, most companies have limited ability to use these generalized models in their specific context of use. They have to train and optimally adjust the AI with the real data, which is only available in small quantities, and with regard to the respective, usually very specific, context. Therefore, we see three main AI trends for this year:

1. More focus on Small Data and Wide Data

Big Data has long been an absolute necessity when it comes to artificial intelligence training. The problem, however, is that in practice only a few companies and developers have access to sufficient amounts of training data. As a result, much of the business community is largely excluded from the technologies of tomorrow. New trends such as Small and Wide Data, which make it possible to make AI and ML accessible to smaller companies, will therefore become more of a focus in the future.

Small Data approaches aim to create value even from smaller data volumes with machine learning methods optimized for them using new analysis techniques. Wide Data is about creating synergies from a wide range of different data sources and types to improve the context for AI applications. With these approaches, more organizations will be able to effectively and profitably leverage their own treasure trove of data in the future.

For example, according to a study by the market research and consulting company Gartner, around 70 percent of all companies will shift their focus from Big Data to Small and Wide Data by 2025.

2. Intelligent Document Processing on the Rise

Intelligent document analytics can make text-based work processes in companies much more efficient by partially or fully automating these processes. Especially in public authorities and large companies, vast amounts of data are available and new ones are added every day. Often, a number of employees are tasked with filtering the relevant information from documents that is required for further processing. This takes a lot of time, and the human factor makes for a comparatively high susceptibility to errors. Intelligent Document Processing (IDP), i.e. the use of AI-based software for processing documents, enables the digitization and automation of workflows.

The potential applications are diverse and range, for example, from application reviews and order acceptance to updating customer and payment data. Therefore, this technology will gain in importance.

3. Conversational AI makes progress

Conversational AI refers to smart, AI-based dialog systems that act as virtual assistants and attempt to interact with customers in at least a partially automated manner. The whole range of Natural Language technology, i.e. machine learning and also symbolic AI, is used.

The goal is to recognize the intention of customer inquiries and to act accordingly in an automated manner. A distinction must be made between "I have paid my bill" and "I will pay my bill when..." in order to respond correctly to customer emails. Easy for the human, not easy for the machine, but possible to master. So the customer's question has to be interpreted and "understood" and for that possible answers have to be identified.

This is a new challenge for intelligent search engines equipped with the full range of speech and text-based AI functionalities. Together with Question Answering, this takes chatbots to a new level of quality. The better the conversational AI system works, the more customer inquiries companies can process automatically. This not only saves employee resources and reduces costs, but also makes customers less dependent on business hours.

However, development here is still in its infancy and it would be reckless to place too high expectations on the performance of conversational AI systems or chatbots based on them at this stage. Providers often act with full-bodied promises that cannot be kept in practice. Clear expectation management is therefore very important for a successful project with achievable goals. The optimization of AI processes for the ability to achieve robust results with the often sparse training input is what distinguishes market-leading manufacturers here.

Conclusion: Artificial intelligence is certainly not a panacea. But we should seize the opportunities and use the added value that AI can offer us: Relieving employees of time-consuming standard tasks, making many business processes more efficient, and providing employees and customers with access to the information and services they need in the shortest possible time.