Analytics is the future. It’s the present, too, but you already know that. Your company is gathering data (and has likely been doing so for years), and you’ve probably got a system or two to glean insights from that data to make smarter decisions. But just as you’re getting a handle on your analytics program, you start hearing about “augmented analytics,” and now you’re worried you need to adopt that too to stay competitive in an evolving world.
The good news is that augmented analytics will make your life a lot easier! Augmented analytics (according to Gartner, which would know), uses technologies “such as machine learning [ML] and AI to assist with data preparation, insight generation, and insight explanation to augment how people explore and analyze data in analytics and BI platforms.” Whatever you do and however you do it, augmented analytics serve up deeper intelligence from data with less heavy lifting. In this article, we’ll run through the ways augmented analytics will improve your analytics user experience and outcomes, no matter your level of technical skill.
Artificial intelligence combined with analytics enhances every application!Simplify analytics with AI
Data science and artificial intelligence: Enhancing every step in the BI process
Below you’ll see the traditional BI process: You start with a problem you want to use data and intelligence to try to solve, and you work your way through the steps toward hopefully doing that. Augmented analytics will improve every step of this venerable process. Data science and artificial intelligence go together for a wide range of reasons and can help users of all skill levels glean actionable intelligence from their data. Let’s look at how:
Augmented analytics evolves the cycle above by extending self-service BI and making analytics more accessible to users who are less directly involved with data science. AI elements incorporated into every step empower users of all skill levels to uncover the actionable intelligence they need to evolve their businesses. Platforms like Sisense lead the trend of increasing access to AI-powered analytics and infusing intelligence from these systems into user workflows, as well as apps and experiences.
From data preparation, with attendant data quality assessment, to connecting to datasets and performing the analysis itself, helpful AI elements, invisibly integrated into the platform, make analysis smoother and more intuitive. Traditionally, data analysis drew the boundary between coders and business users, but augmented analytics bridges this gap.
Tools of the AI trade
Augmented analytics is all about automating and improving as much of the analytics cycle as possible. AutoML’s aim is to automate the development of machine learning models. Every analytics user can benefit from this type of augmentation. Specifically, the familiar tasks that all ML projects have in common can now be standardized and automated. In broad strokes, these tasks include:
- Data preparation
- Feature engineering
- Model generation
- Model evaluation
As you can see, this gets complicated very quickly. Bringing these tools into the hands of financial analysts (called quants) and insurance actuaries quickly solves the problem of access and opens up the audience. A typical data science text outlining these methods is 1,000 pages of equations and algorithms. Today, much of the mathematics has been automated. The point is that these methods are standardized for many scopes, and noteworthy services already exist, such as Google Cloud AutoML. The benefits are harvestable in every field of human endeavor.
Code-driven analytics from data scientists
While there are low-code and no-code platforms in AutoML that greatly expand the field of AI users, it remains true that users with more coding ability can expect more benefits from AI. For example, a user with some basic insight into why data preparation — the first of the four main steps in ML mentioned above — is important to training an ML model will ensure greater accuracy in the forecasting outcomes generated by the model.
While an AI-enabled system is sometimes the target product a company seeks to build, AI itself can also benefit highly technical users as a development tool. Team members with data skills including SQL, Python, R, and other prototyping methodologies can work directly to enhance analytics modeling platforms like Sisense. More technical users can code to prepare data for advanced analysis, build more complex data models, create materialized views, do sentiment analysis, and more; the imagination is the only limit. In other words, data experts can dovetail their coding skills with AI functionality to produce more sophisticated and more accurate models. While AI is the tool under development, AI itself can also benefit highly technical users as a development tool. Team members with data skills including SQL, Python, R, and other prototyping methodologies can work directly to enhance analytics modeling platforms like Sisense. More technical users can code to prepare data for advanced analysis, build more complex data models, create materialized views, do sentiment analysis, and more; the imagination is the only limit. In other words, data experts can dovetail their coding skills with AI functionality to produce more sophisticated and more accurate models.
An ML user with statistics training and some basic knowledge of how deep learning neural networks operate can start with columns of revenue and annuity and generate additional features including mean, standard deviation, and kurtosis to enhance the model training data and thus improve prediction accuracy. The potential benefits are limitless. These methods have become so standardized that a new AI development paradigm has emerged that seeks to garner the development efficiency of DevOps: AIOps.
AIOps is about the application of DevOps methods to tackle challenges in the field of AI, a special branch of software implying a special development methodology. In particular, the DevOps practices hope to achieve operational efficacy for the development of services that must now be implemented in AI development.
The continuous development and release of applications as services according to the DevOps paradigm is now widespread and almost universally adopted, alongside the rapid proliferation of cloud computing. The same DevOps solutions to challenges that arise with the ever-increasing scale and complexity of efficiently building and maintaining software services are now demonstrating similar effectiveness while evolving as AIOps, with benefits for both engineers and end users.
Improved customer satisfaction is among many expected benefits of AIOps. For example, an intelligent customer service system can take independent action to resolve customer issues:
Ideally, an AI-based app or service can learn as it does its job, interacting with customers and recommending configuration adjustments to improve its own performance. Call centers already report that AI chatbots receive higher customer ratings than human agents!
The use of an intelligence system like this goes hand in hand with engineering: The AIOps development cycle will likewise include AI-based developer tools that communicate with the tools they create! Improved engineering productivity is therefore a concurrent benefit to be expected with AIOps. Increasingly, engineers and end users will be freed from tedious and repetitive tasks, and thus have more time to enhance the intelligence of the system as a whole.
Augmented analytics: Evolving the roles of data experts
As the roles of data scientists and business users shift and evolve thanks to augmented analytics, new questions arise. In light of reduced technical obstacles, should all business users assert a role in data intelligence and decision-making?
Traditionally, which team members did which tasks was tightly defined based on abilities: Coding ML algorithms was once the expert domain knowledge of data team members. But now that anyone can plug in a data discovery tool (particularly one using the model tournament concept), theoretically, almost anyone could evaluate a data model.
Why shouldn’t business members in all strata participate in building and deploying new data models to power their analytics needs? Issues of trust and credibility will play a role in deciding who can build and deploy what, regardless of how technology may empower us in the future.
Infusing actionable intelligence into workflows, products, and experiences is the future of analytics and AI.Explore actionable analytics
Inbar Shaham is a senior product manager at Sisense. She has 11 years of experience in product management, having worked for Clarizen, Takadu, and ICQ, among others.