In my previous article, I introduced the concept of augmented analytics that unites the best of machine intelligence and human intuition with the goal of speeding up the process of insight generation.
Today’s machine intelligence uses a variety of techniques to automatically generate analyses and insights in visual and narrative form, based on user selections, natural language search and advanced analysis types. Examples include charts and visualisations, narratives explaining key findings, insights into data relationships, and the generation of entire dashboards for further investigation.
Automation artificial intelligence (AI) speeds time-to-insight by automating a wide variety of user tasks, which includes data preparation (obtaining, combining data sets and transforming data) and creating analyses.
For example, when a user wants to analyse data from multiple sources, algorithms can determine the best ways to bring it together, profile the possible dimensions and measures, and suggest the right forms of visual representation and analysis.
AI supports powerful new ways to interact with data.
Now to the topic of natural language interaction. AI supports powerful new ways to interact with data. Users can now ask questions in natural language and the system understands the intent and context, analysing the data to generate the right responses. Whether this is done through search-based discovery or conversational interaction, AI delivers narrative answers and visual insights, boosting the innate human ability to question in different ways.
With an increasing number of businesses adopting advanced analytics, including advanced clustering, forecasting, prediction and modelling, it is often challenging to put this power into the hands of decision-makers.
Whether predictive calculations are performed by the analytics platform using automated machine learning, or in third-party data science tools, users need a way to interact so they can ask questions and factor powerful insights into decisions.
So, what must be considered when selecting an analytics platform built to maximise the value of AI? My advice is that when evaluating solutions, consider the following:
Does the solution have a purpose-built calculation engine? If the platform simply layers AI capabilities on top of a relational database, you’ll run into limitations. Instead, look for a solution that gives users the power to easily combine data, search and explore in any direction, and move at the speed of thought – with no pre-aggregated data or pre-defined queries.
Next, ask: Is the solution built on an open, extensible platform? Can it analyse both historical and changing data? In addition to evaluating historical data to provide a picture of what’s already happened, the system should facilitate the analysis of real-time data to trigger action as business events evolve. To do that, it will have to capture changes in data as they occur and deliver them to the analytics platform in real-time. Remember, AI will monitor data and evaluate it as it changes.
Also take note that it’s not enough to tack on a few AI capabilities. Companies will want the freedom to build and integrate with anything needed as their business, industry and marketplace evolve. Choose a platform that can be extended to handle new use cases, integrate in real-time with data-science platforms, and embed within operational apps and business workflows.
Then question if the solution context-aware? Does the system take a “one-size-fits-all” approach? If the only option for AI is search, there will quickly be users who need more in-depth visual analysis.
Look for a system that supports a full range of augmented capabilities, so the right experiences can be offered to the right people. And make sure insights are open and transparent. Otherwise, users will become distrustful, which in turn will compromise adoption and collaboration.
The system should be able to understand user context and/or intent when accessing data and materialising insights. If not, so-called “natural language” interactions won’t be natural – or relevant – at all.
Does the solution have AutoML and predictive analytics capabilities? Automated machine learning can enable users to easily generate models, make predictions, and test business scenarios using a simple, code-free experience.
So, what’s the bottom line on how to extract real value from AI in data analytics?
Typically, business analytics have been historical. For example, a company might examine which sales opportunities the business won and why. Insights can be gained from conducting further analysis on why it won the business.
With machine learning, the company can uncover the key drivers for wins and build a model to predict outcomes from forward-looking data. However, understand, this does not mean it can affect those outcomes – so, this is where explainable AI enters the picture.
Before you scream ‘not more jargon’, let me expand on this. Good old Wikipedia defines explainable AI as often overlapping with interpretable AI, or explainable machine learning as a field of research within AI that explores methods that provide humans with the capability of intellectual oversight over AI algorithms. The focus is on the reasoning behind the decisions or predictions made by the AI algorithms to make them more transparent.
By generating explainability data, it is possible to not only see what's likely to happen, but also which factors will drive a particular outcome. This last empowers the company to target action and investment.
Without machine learning, there is no way of revealing which sales opportunities the business actually won in the last quarter and what was the deal close rate. This type of information is veritable gold dust to the strategic decision-makers in any business.
Share