The stage is being set for machine learning (ML) to be adopted by enterprises. With the convergence of data storage, computing hardware, and new algorithms upon us, everyday business processes are set to undergo meaningful evolution. However, much of ML’s success in the business context will hinge on two key pillars:
- What business problems it will solve?
- How those business problems are technically setup for ML to steal the show
Chatbots, digital assistants, and other natural language-based applications are being increasingly adopted and for good reason. As companies scale over time, their operations and processes will evolve in complexity, requiring a corresponding evolution in workforce capabilities. This gap can be bridged by conversational AI - enabling machines to learn from users and not vice versa ultimately empowers the end user. With natural language interactions, employees can engage with business systems through a compelling user experience - say it how they want, use it where they want and access it whenever they want - and generate meaningful value.
In the context of IT, let’s think about the business case for an IT help desk. Customer facing companies especially are flooded by customer requests through every channel: phone, emails and chats. They struggle to keep up with all of these inbound requests while still providing excellent customer care to other clients. Typically, there are three levels of support that are provided by a help desk, where level one and level three require low touch and high touch actions, respectively. Running a help desk is challenging - enterprise users require timely and accurate help to ensure they can easily execute daily responsibilities. The high volume of requests places a lot of strain on help desk agents.
Cue the chatbot: a virtual agent can work upstream of IT agents to field level one and some level two requests from users as best as it can before falling back to an agent for further help or creating a support ticket. Chatbots are the multilingual, 24/7 agent that can manage basic questions autonomously, saving up to 20 percent of the time of human customer support agents. With this help, human agents no longer are burdened by answering the same question over and over again. Instead, they can focus on providing personalized service or focus on prospects and sales to generate revenue. Unsurprisingly, this setup is particularly compelling for all stakeholders. Users are able to come away with responses in a potentially short timeframe; agents are able to focus their time on the most pressing issues without being burdened by the volume; and the company benefits overall as its primary stakeholders are happy.
It is important to note that conversational technologies extend beyond just chatbots. For example, many exchanges are conducted through email, static messaging, and more. These situations might require different solutions: does the agent want technology to automatically answer emails or simply classify the messages so it is easier for them to handle? Does the agent want the solution to only answer specific emails (e.g. password reset) while handing other emails (e.g. broken software) by themselves? It is entirely dependent on the situation. Therefore, the previously-discussed business case and ROI initiative can help determine the sweet spot for a conversational solution.
In general, there are many applications for ML in the context of IT outside of support bots. Machine learning spans fields from time series analysis to image recognition - the possibilities are limitless! However, this is why validating the business case is so important. Organizations should identify the most inefficient end-to-end process, drill down into the critical steps (operation protocol, user journey, etc.), curate the problem-need-solution focus, and then turn to its ML arsenal to see if it can be used in the solution. Processes that are repetitive in nature, data-intensive and time-intensive (especially on users) are some of the ones where ML can offer the most value. Above all, how organizations condition the problem - operating environment, desired capabilities and more - may largely affect the outcome of a ML-based course of action.
Though it is on the upswing, ML has just started to make its presence known in much of the enterprise space. A recent study by SAP and the Economist Intelligence Unit (EIU) shows that 48 percent of companies who have embraced ML have experienced profitability as the top benefit, however, if companies can adopt and implement it in the course of business, it will yield not just tangible bottom-line effects. Beyond profitability, ML can contribute to more abstract goals as well such as improved productivity and happiness of workers. Being on the cutting edge of technology and using it in innovative ways to solve problems will foster creativity and ingenuity within the business, and also shape the external perception that the company is a capable leader.
With respect to company-to-company relationships, success will further cultivate customers’ confidence in the organization’s abilities, raising the prospects of the company through increased business, and also strategic activities like partnerships. Similarly, in company-to-individual relationships, businesses will not only be more attractive to current employees but potential talent as well; technological savviness and openness are highly-valued by today’s recruits. The best part is, none of these transformations are temporary and companies that incorporate purpose-driven machine learning, can drive success for years to come.
As more companies begin to invest in ML, it is important to remember that the outcomes are only as good as the data that is input. With the availability of an overwhelming amount of data, companies need to ensure their data is clean and consolidated before implementing ML. Integrating data between systems is critical to making sure datasets are accurately and efficiently analyzed. Without this alignment, it is difficult to gain valuable insights. For businesses to succeed with ML, they must understand that it is not just one isolated project but an entire journey. Ultimately, the organizations that are willing to put in the time, and learn from ML trial and error, will receive the most meaningful benefits to their overall business.
Samir Patel, Product Manager of SAP Conversational AI
Image Credit: Geralt / Pixabay