Trends in Data and AI to align with
Looking ahead to 2023, I foresee a few trends that will play out in the data and AI space. If you are a leader of a data function, it might be worth aligning your product roadmaps to take advantage of these trends. Becoming better in using data is a journey, so I’m calling out different things to focus on, depending on the maturity of your organization in its use of data.
1. Reuse data across business use cases
Basic: It is now table stakes to be collecting business-specific data and ensuring that the data is of high-enough quality to support decision-making.
The trend now is towards data reuse across business use cases.
Good: You have broken down silos between business units (marketing, product, sales, fulfilment, etc.) as well as between functional roles (data analytics, data science) by getting all business data into an enterprise-wide data platform. This way, you are able to tie a customer journey or interaction as it moves through your business processes (e.g., from customer acquistion to purchase to shipping to retargeting).
Better: Your data platform is a data mesh, to ensure that data owners are responsible for data quality, and that data is available throughout the enterprise. However, you have also identified KPIs and common operations on data and captured them in an enterprise-wide metrics layer so that data processing steps are auditable, reusable, and maintainable.
Best: As the data footprint increases throughout you organization, you notice that your business workflow involves the same data pulled from different locations (logs, transactional databases, historical archives). Recognizing that the best data pipeline is one that is completely invisible — your data sources and databases and data warehouses are synced by automated ETL that provides a live, synchronized view. You have built a data platform that connects logs, OLTP, and OLAP seamlessly (ask your cloud provider how they are making this easier for you). Such a seamless connection enables you to reason logically about the data independent of the infrastructure. Ultimately, this makes your applications more resilient and developing them is faster.
2. Employ AI to extract insights from data
Basic: Your dashboards include predictions of key metrics and not just historical/backward-looking values.
Good: The dashboards are live (or the delay involved is immaterial to your business). This typically involves streaming ingest and data processing. Some of the insights are drawn from unstructured data (images, videos, natural language documents, etc.) using machine learning.
Better: The insights are embedded in the tools (websites or mobile apps) that decision makers use, and do not involve having them use dashboard tools. You are able to scale out the analytics and insights across even millions of users at a reasonable cost. The insights are drawn from all the relevant data, both structured and unstructured.
Best: The embedded visualization is interactive, allowing end-users to change the assumptions underlying the predictions and gauge the impact of different scenarios. It is also possible for the end-user to correct extracted data. Users get notifications/alerts when user-specified events happen.
3. Use generative AI as a productivity hack
Basic: You have autocomplete in search textboxes. The autocomplete is based on your product catalog and past history of user interactions.
Good: You have autocomplete and spelling/grammar correction in any text box into which a user can enter free-form text. You are using off-the-shelf tools to streamline business processes: for example, your marketing team is using generative AI to speed up the production of blog posts; your support team is using generative AI to respond to customer queries faster.
Better: Generative AI for text, music, and images is used as part of your content creation workflow. It may be used to generate summaries, fill in gaps, or even write entire articles or compositions. There is human oversight and editorial input throughout the process to ensure accuracy and maintain brand voice.
Best: The generative AI is trained/finetuned on your own data, and is used to streamline such activities as user reviews (suggesting topics that the reviewer should address next), support tickets (validating and asking relevant questions), and other user-generated content.
4. Optimize user experience with first-party data
Basic: You are doing A/B testing of user-facing websites and changing the user experience based on the tests.
Good: You have connected all the user journeys across all your sites and are able to successfully re-identify the majority of repeat visitors. You are able to provide a personalized user experience based on user profile and user intent, not just static things such as IP address.
Better: You provide contextual search within your website or app, and serve dynamic websites based on current user intent and activity.
Best: You have moved beyond A/B testing and towards fully dynamic webpages based on continuously trained propensity models.
Which of these things is your organization already doing? Which of these trends are on your 2023 roadmap? Comment below!