At After the flood, we are often asked to develop products that create predictive analytics, rather than passive displays of information or data due to the perceived value in AI services.
Back in 2009, we spoke at the O’Reilly Big Data conference in Santa Clara. Nine years later we spoke at the O’Reilly AI conference in London. The difference over time between the aspirations and capabilities of Big Data and AI sub-sectors is significant.
The Big Data Age focused on basic organising, storing and retrieving data; the Age of AI now leverages this data to produce predictive actions, suggestion, and experimentation. The key recommendations to adjust to this paradigm shift are:
- Focus equal effort on creating interfaces that learn information and give information from users in a transparent way.
- Analytics teams should equip themselves with AI and lean insight tools rather than passive dashboards.
- The transition of customer personas to live customer segments should include dynamic and seasonal behavioural insight.
We believe the trajectory of analytics to AI can be typified in the following way:
From static systems to learning systems:
Modern intelligence systems learn with more use and with more data. The early leaders like Netflix and Waze are those with massive datasets to train their intelligence on. These systems need training and face a design problem in creating ‘trainable moments’. Interfaces for AI need to solicit feedback on user behaviour for training purposes as opposed to simply providing information.
Product and design teams should focus equal effort on creating interfaces that learn from users, in a transparent way, as much as interfaces that give information to users.
“Interface Design is the design of conversations between the user and the data”—Marc Rettig, Distinguished Adjunct Professor of Practice, Carnegie Mellon Design
From analytics to experimentation culture:
Previously, data processing teams were set up to store and process data with timetabled releases of well-designed insights. This took place in industries that used slower decision-making loops as opposed to more modern, lean and instant decision-making.
Now, teams have the ability to run different experiments, ask questions and make predictions with their data, quickening decision-making, feedback loops, and allowing experimentation with quicker results.
This is a sound engagement tool for executives who are new to these concepts but it requires a level of translation and improvisation from data teams to manage their interest and questions.
Analytics teams should equip themselves with AI and lean insight tools rather than passive dashboards. Analytics teams should publicize their results and experiment cycles to encourage the business to collaborate. Leadership teams should incentivize all staff to understand and take part in the experimentation process with a named board member responsible for reporting on experiment culture and results monthly.
From customer personas to live customer segments:
In the past, customer personas were an amalgamation of qualitative and quantitative insight, held on file by analytics teams. They were shorthand for types of users that would be used to commission new research and understand usage results.
Now, a database of live and predictive data, learning new habits as they happen, allows experimentation for feedback on live services and a more granular cut of the customer data. This can be seasonal for instance, with experiments on the habits of Christmas shoppers or based on new online browsing behaviour such as a house search triggering advertising for brokers.
CMOs or Chief Customer Officers should modify their customer reporting to include dynamic and seasonal behavioural insights in addition to those rooted in the more static personas.
Our expertise and research with our clients suggest that companies will be influenced primarily by the design and outward facing attitude of the business to leverage AI systems.
From 2009 to 2018, static systems and dashboards have evolved to predictive, intelligent and learning systems. Most companies have traditional analytics system in place, but more and more understand the importance of an experimentation culture and predictive data to quicken and improve decision-making, ultimately leveraging company growth.
If a company is interested in engaging with AI systems, they need to create an atmosphere of engagement with board members to the executive and to the non-experts, allowing for buy-in and improvisation from data teams. In this case, companies should also work with companies that help to design for this new paradigm of experimentation culture.