Business process automation (BPA) is the strategy a business uses to automate processes in order to contain costs.” This definition from Wikipedia needs expansion beyond cost containment, I feel, in today’s digital world to also include the notion of processes being automated, e.g. in order to build customer bases and grow revenues. In the extreme and not too distant future, BPA will allow automation of workflows under both notions to the point where human intervention becomes unnecessary – in part or in full resulting from the introduction of artificial intelligence (AI). But (today?) not every business process is a good fit for automation yet. So it’s important for any business to clarify which processes are best suited to automation and which ones are not and remain best handled by humans. Of course routine and repetitive tasks that are predictable and manageable have been automated first, like assembly work on the shop floor. But increasingly, manual tasks that are particularly risky or may cause injury are being taken over by robots, too. Think of cleaning of radioactive spoils for example. And companies are also beginning to exploit IoT technology to drive their automation and after sales services by sensor-based tracking and alerts. But also difficult decision-making has become machine-addressable as of late such as loan approvals in banking – at least with some standardized products. Or think of document management systems that allow an end-to-end paperless workflow, or automated database searches that use both traditional records and big data. But any company should think very carefully before introducing automation to its customer-facing processes. Not every customer will be appreciative e.g. of complicated chat-bot lines with multi-layered phone trees, and at worst a phone disconnect at the end of the process. Voice- and optical-based automation systems are not yet that infallible that they can handle all situations properly.
Growing numbers of specialist companies are entering the BPA market and offering solutions based on AI technologies that can work with big data and unstructured datasets, interact with human beings, and adapt to new types of problems in self-learning mode. With the exception of arago, most BPA providers in Germany like Sota Solutions, 5Analytics, N-Join, KONUX or Micropsi Industries are not claiming to develop general or human level let alone super artificial intelligence but so-called narrow AI. This is focused on solving single tasks as opposed to general AI that is able to apply experience gathered in one area to a different set of problems, just as humans would do. But their underlying approach tends to be similar in that they will attempt to provide the shortest route to automation by exploiting the user interface layer rather than going deeply into the application code or databases sitting behind them. They also simplify their own interface to the extent that these tools can be used directly by non-technically qualified staff. The main advantage of these toolsets is, therefore, their speed of deployment, the drawback is that it brings yet another IT supplier to the organization.
In an interview on July 28, 2017 with Stefan Söhnle, co-founder and operations director of 5analytics UG from Stuttgart, I asked Stefan about their approach and what they are bringing to the table? 5Analytics, founded in spring 2016, focuses on the integration of AI into business processes, i.e. what Stefan calls ‘operationalising of algorithms’ which in other words enables algorithms to recognize complex connections in data and deliver new basis for decisions. But what is new or artificially intelligent about this? Take car repair and maintenance as a simple example: in the early days, automobiles drove until something needed fixing, e.g. the motor broke down or a tyre went flat (no rules), then fixed intervals for inspection were introduced, i.e. drivers needed to check certain parts say every 10,000 km (fixed rules), and artificial intelligence now introduces individualization and flexibility, in that each individual car can be monitored seamlessly and its collected data used for predictive maintenance triggering action that may again be automated in full or in part or involving human workforce. Take another example, some online shopper who buys his first Harry Potter book is likely to buy also the other volumes, here a so-called recommendation engine –based on data analytics of thousands of other customers- automatically pops up on his computer screen: “Other clients also purchased … Harry Potter volumes 2-7 or similar fantasy novels x, y, z.” 5Analytics does not specialize on identifying the influencing parameters and translating them into algorithms, although it provides this as a consultancy service to clients with no own AI expertise, but the company specializes on converting all these algorithms into real action in the daily operation and above all into revenues!
According to Stefan, contacting SME clients is relatively easy, the AI topic and IoT being ‘en vogue’. Explaining what 5analytics is doing and how its’ product offering could help reduce cost and/or generate more sales in practice takes more time. Once contracts are won, problems which need mastering are again more of a standard technical nature, such as checking of how the algorithm need to look like, ascertaining data access and data collection wherever stored, integrating the data connections, etc. … and 5analytics praises itself that in the end all that works with many parameters and users at the same time and most importantly in real time.
Depending on customer need, 5Analytics’ artificial intelligence platform ADA offers clients a number of solutions for specific work streams in the fields of production, predictive maintenance as well as sales & marketing. The latter for example include a dynamic pricing system as well as customer behaviour models for cross- and up-selling or improving customer retention.
Clients can choose to buy or rent their platform which comes with standard interfaces for all the most common systems, and claims to be highly scalable and easily integratable into existing systems landscapes building on existing R or Python scripts.
Although competition in the BPA and AI market is likely to heat up in the coming years, the landscape of international and German AI today is so fragmented that there is room for many software and hardware companies to grow and become successful. How successful and intelligent they will be remains to be seen, but Stefan is convinced that (narrow) AI at least will be as widely used in 20 years as smartphones today.