Artificial Intelligence (AI) is booming. And its growth is receiving attention around the world. You’ll see AI featured in news stories daily. Total investments for venture-capital-backed artificial intelligence companies in the US topped $1.9 billion in the first quarter of the year, the highest ever. Google, Amazon, Facebook and the tech giants are making big bets in this space. But what is AI really, and why is it important? Before diving into the buzzword version of AI that’s often treated as a cure-all (“big data” from a few years ago, anyone?), we have found it valuable to break down the key areas of AI in our conversations with our clients.
Three key areas of emerging technology
There are three areas that sit underneath AI that spawn multiple solutions in a variety of industries: Machine Learning (ML), Conversational Interfaces (CI), and Robotic Process Automation (RPA). Each very different and powerful in their own right, and in some cases each one individually can enable an AI solution or an orchestration of all three can enable an AI solution. Let’s break each technology down briefly.
Machine Learning (ML)
Machine Learning is a sophisticated family of predictive models that are integrated into a live data ecosystem. This could be as simple a model that predicts fraud and sits in a production database. It can also be more sophisticated deep-learning neural networks finding patterns in images and language such as identifying cancer from medical scanning images. This is an application of Natural Language Processing (NLP).
- Computer vision (CV) is a component of ML, which is a hot topic with the rise of image detection and self-driving cars.
- Deep learning (another component of ML) is also a popular topic. It is an analytical approach that has enabled many breakthroughs in modern computer vision and NLP. Deep learning thrives where solutions can be built over extremely large datasets with unstructured data (think large videos, music, raw voice, photos). Google has made a business of building solutions on large unstructured data, so it’s no surprise that they have released their leading deep learning development tool to the world – Tensorflow. Tensorflow enables developers to quickly build deep neural networks for their own applications. Google is continuing to make big investments in selling computing services around ML with their Tensorflow Processing Units (TPUs).
It is also important to note that ML and AI are often interchanged or confused in the marketplace. There are many conflicting definitions, but we find it easier to put ML under the AI umbrella and characterize it as the ‘brain’ in the AI ecosystem.
Conversational Interfaces (CI)
Conversational Interfaces, more colloquially referred to as ‘chat-bots’, have been covered extremely well by my colleague, John Sprunger. I would encourage you to check out his blog here. Conversational interfaces are technology frameworks that utilize NLP to understand topics and users’ intentions from unstructured voice or text inputs. While NLP acts as the brain, CI is much more than just NLP, as there are tools to help structure how the conversation flows. Today, most CIs are structured in a ‘choose-your-own-adventure’ style. The overall flow of the conversation is controlled by business rules. However, machine learning can be used to generate smart responses back to users (these are called generative CIs). The technology for generative CIs is not there yet: case in point, many still remember Microsoft’s failed experiment from their generative bot, Tay.
Robotic Process Automation (RPA)
Robotic process automation is the last key area under AI. RPA serves as the mover of information between front-end systems, and typically uses simple business rules to eliminate repetitive tasks that don’t require a human such as invoicing (which we would not characterize as AI). However, RPA is starting to work as an enabler of ML in technical architectures. You could use ML within an RPA process to drive process automation. For example, RPA may retrieve information from a clunky front-end system that has no database to obtain key elements on customers. This information can feed into a hosted ML model that predicts the probability of customer churn. RPA tools are also using Image Recognition to read forms and application screens. RPA helps facilitate the flow of information into the ML “brains” and moves the outcomes from the brain to other key systems or decision makers. For more information on RPA, you can read this whitepaper that discusses the impact of implementing RPA, common scenarios, changes and challenges.
How to get the most value from AI
AI is not a simple monolithic technology, but an ecosystem. It’s a usage pattern of new technological tools that have machine-learning baked into the core of how they work. ML serves as the brain, CI is the voice and ears, and RPA as the arms and legs enable valuable solutions. To ensure that AI is creating value, you must first have a clear analytics strategy that maps your business’ value creation story and overarching strategy. AI for AI’s sake will never be successful. If you have a clear strategy and/or use case, but don’t know how to activate AI, West Monroe can help. Reach out to us to continue the discussion.