Strong data foundation
AI and its associated solutions have become easier to create and integrate in the 25 years since its introduction.
However, most AI solutions are still focused on specific areas, such as document processing, demand forecasting, and predictive maintenance. Compared with more complex, enterprise-scale solutions, these point solutions take relatively little time to conceptualize and build. This speeds time to value and makes funding easier to acquire. Also, because these one-off AI initiatives are much easier to accomplish than enterprise-wide projects, they have more consistent success rates.
So, what’s keeping broader, enterprise-wide AI projects from being successful? To a great degree, a deficiency in the underlying data foundation creates a barrier. Some of the challenges include:
A lack of relevant real data to accurately train models: Synthetic data can be used in certain scenarios, but it comes with its own challenges. Synthetic data may be time-consuming to create, and it may contain inaccuracies—yes, even synthetic data depends on the quality of the real data upon which it’s based.
A poor understanding of investment needs: The footprint of the business problem dictates the amount of investment that will be made in the AI solution. Any gaps in understanding the problem can lead to substandard ontologies or dictionaries, as well as an ill-defined technology base or misaligned skills.
Gaps in data security and privacy management: Insufficient attention to data protection can lead to legal and regulatory action.
Incompatibility between AI libraries: Unrestricted technology proliferation can create various incompatibility issues. This often happens when code isn’t designed following user-friendly, computer-aided software engineering principles. The situation can be made even worse when multiple vendors are providing similar and ever-evolving capabilities.
To scale their AI projects, organizations need to start laying down a stronger data foundation. This requires a full-spectrum approach that accounts for organizational, regulatory, technical, and ethical challenges. In addition, broader AI success depends heavily on constant investments in data literacy, because a strong data culture helps pave the path for successful adoption.
The following initiatives can help:
Drive modularity: Technology is ever-evolving, so businesses should focus on building plug-and-play components.
Bring out tacit knowledge: AI is not BI. It is not meant to help organizations understand themselves. Rather, its mandate is to develop a competitive advantage. To do so, AI solutions must unlock knowledge within the organization.
Focus on how the people-process aspects will evolve: When AI is introduced, businesses will need to re-balance activities between humans and machines. Yesterday’s process owners and SMEs will need to focus on training and fine-tuning the model rather than doing the job themselves.
Elevate the skill sets within the business and utilize all available knowledge: There is a vast difference in outcome when engaging a data scientist from the outside versus grooming business SMEs to play that role. Just hiring a bunch of data scientists will not achieve the intended result, because they won’t have the right contextual knowledge to create solutions that will work within an organization’s processes.
Find the right balance in the service delivery approach. A composite strategy that combines agile, innovative business solutions with a focus on the productivity of foundational initiatives is a must.
Get the fundamentals right: Assess the organization’s baseline in the following areas: AI applicability to business cases, AI infrastructure and tooling, organizational competency, strategy, and model management processes. Essentially, this includes the entire gamut of capabilities across people, processes, technology, data, and of course AI governance. This can always be done while delivering low-hanging fruits to the business.
Overall, organizations must stay focused on strengthening their data strategy, which is a core element for succeeding in the AI world. Building and leveraging a data foundation is key to developing a data-centric culture that’s equipped to effectively use the organization’s data assets. By creating a complete mosaic of technology and skills, based on a strong data foundation, businesses can accelerate broader success with AI—moving out of narrow project success into enterprise-wide transformation.