Why AI Data Readiness matters
Investments in AI development are increasing rapidly, and the pressure is on to ensure AI solutions demonstrate real value. Research with customers engaging in AI development indicates that AI Data Readiness is the primary predictor of AI success.
Defining AI Data Readiness
There are three components of AI Data Readiness:
Data should be available: Locating the necessary data, understanding who owns that data, and having a strategy to access that data is crucial to AI development planning.
Data should be trustworthy: Once the necessary data has been located among all of the data sources in the enterprise, the data itself must be interrogated. Knowing the aspects of the data's quality (completeness, timeliness, and consistency) is critical when deciding whether the data can be trusted.
Data should be usable for AI: Specifically, the data must be accessible to both the Agentic AI (or AI solution) using the data, as well as to the human end-user.
The components of AI Data Readiness need to be examined for every AI solution. Importantly, this will likely require collaboration with data owners and experts outside of your ServiceNow AI Platform team. To get discussions started with these collaborators, the section below outlines four key questions to understand your data's availability, trustworthiness, and usability.
How do I get started on AI Data Readiness?
Asking four basic questions will help you and your team to drill down on Data Readiness for any AI development project. Ask the following questions when collaborating with your development team, data owners, and data experts.
Do we know what data we have?
Answering this question will help you understand what data sources are relevant, where the metadata resides for data sets in those sources, and what types of data inventory you are able to use. Starting discussions with individuals your Enterprise Data Organization (if applicable) can often be a way to accelerate your knowledge about data that is available in the enterprise.
Who owns the data, and is it secure?
Once you have an inventory of the relevant data, collaborating with owners of and experts in that data is key to obtaining access to the data and understanding the format of that data. This question will also inform practices for keeping the data secure in transit and at rest, according to your organization's security policies and regulatory requirements.
Can we rely on the data we have?
The data owner and expert will be the main resources for quickly understanding the data quality. Specifically, you will want to focus on data completeness (frequency of null data values in relevant fields), timeliness (recency and refresh cycles of data), and consistency (measures of data quality over time; e.g., is the data quality stable, increasing, or decreasing?).
Can the right people and tools reach the data?
This question will address how the AI solution, and its end users, will access and interact with the data. This discussion will require collaboration between the admins and developers on your ServiceNow AI Platform team and the data owners. Ensuring security of the data, and configuring data access properly, will be key to answering this question.
Specific actions to foster AI Data Readiness
If you have AI solutions on your roadmap already, you can apply the above principles and questions to upcoming AI development projects. Alternatively, it is worthwhile to understand your AI Data Readiness for the areas you are most likely to pursue AI Development (e.g., the ITSM scope). To those ends, we suggest starting with some specific actions to exercise your "AI Data Readiness muscle."
Action 1: Audit your current data readiness
This action doesn't require understanding all of your data at once. For example, you can start with one internal ServiceNow table in the scope of interest (e.g., ITSM). For example, what is the data completeness of fields in the table that would be relevant to support case summarization?
Use this as a point of discussion with developers and admins on your ServiceNow AI Platform team. How should the team measure data completeness? By tackling one well-scoped dataset from a single source, you can start to build out standard processes for assessing data availability, trustworthiness, and usability for all future AI development efforts.
Action 2: Appoint a readiness lead or a task force
When something like AI Data Readiness is everyone's responsibility, it's not really anyone's responsibility. For this reason, it is a best practice to identify one person -- or a team -- that is responsible for recommending standard AI Data Readiness practices (having an executive sponsor can also add accountability, and perhaps some urgency, to this work).
Having a readiness lead or task force in place will improve the visibility of AI Data Readiness efforts, and help to foster networking with data-knowledgeable individuals outside of your ServiceNow AI Platform team. That networking will also be an investment in ongoing collaboration that will help to scale AI Data Readiness for future AI development.
Action 3: Require data readiness milestones in every AI initiative
It is essential to give development teams the time they need to ensure AI Data Readiness ahead of model training and testing and AI deployment. To this end, it is a best practice to review AI initiative project plans to guarantee that time is set aside at the beginning of the project to focus specifically on AI Data Readiness.
Require a milestone for data readiness activities, as well as a review of data readiness measures with the development team, data owner, and data steward. As data readiness practices mature and become more standardized, achieving a data readiness milestone should require fewer sprints (or less time) to achieve. Try to avoid a specific red flag for data readiness: "We'll worry about the data later."
AI Data Readiness is key to achieving value through AI solutions
To reiterate, the primary predictor of AI project success is data readiness. You can begin with simple steps today -- even if AI solutions haven't hit your roadmap yet -- to ensure future project success by engaging in the actions suggested above, and by using the four simple questions outlined to understand facets of AI Data Readiness.
