Why LOB Must Steer Successful AI Initiatives, Part One

Why a Reactive Business Team Approach Hinders AI Adoption

The first of this two-part series, discusses why LOB employees need to call the shots when considering an AI deployment and, by leading the charge, they can work smarter and achieve greater return from their AI investments. Part Two will discuss IT's role in helping them utilize AI to achieve better customer experiences as an example of this collaboration.  

Before determining the practical application of new enterprise technologies such as artificial intelligence (AI), organizations must first consider the recommendations from enterprise line-of-business (LOB) teams and the benefits they can practically gain from using them. LOB generally holds the budgets needed to fund these efforts, and only they can make the final decisions on the tools that can help their day-to-day jobs.

While knowledge management (KM) and data science (DS) teams are on top of the latest AI advancement and solutions, they are often disconnected from LOB, meaning they don’t fully grasp current business needs and processes and how those modern AI tools can help business users do their jobs more effectively. Consequently, LOB owners must actively drive and collaborate with IT/DS/KM professionals to identify appropriate AI use cases and tools from the start. If this collaboration does not provide tools the LOB teams can effectively use from the beginning of the projects, then progress will be slow, expensive, and run the risk of not being funded or successful.

Enterprises could save a lot of time and money if LOB led the charge and worked closely with the technical teams in providing their guidance throughout the development cycle. After all, LOB is the front line of any enterprise, and those employees understand firsthand which technologies work for them and which ones don't.

Examples of successful enterprise applications that were driven by LOB/users outside of the development team include Slack and Box. In both cases, IT suggested and implemented its preferred alternatives – generally from previously subscribed vendors such as Microsoft, SAP, and IBM – but it was the users who ultimately pressured IT to enable their preferred applications. Gartner Research reached this conclusion: The future of applications depends on the collaboration between IT and business units. The earlier that AI/IT development teams and line of business employees collaborate, the faster they can share knowledge and expertise to solve problems more quickly, reducing downtime, and delivering better business outcomes.

Why AI Projects Are Not Flourishing Quickly Enough

MIT and Boston Consulting Group have reported that only 11% of enterprises have realized substantial value from artificial intelligence initiatives. One underlying cause is that AI companies focus on selling tools to data science and machine learning (ML) teams instead of LOB. Data science and ML teams tend to choose tools that benefit their routine tasks and workflow, but building something useful for LOB requires a deep understanding of the specific use case and day-to-day routines, as well as close collaboration with the business users themselves. As a specific example, most existing AI approaches require a tremendous amount of data labeling from LOB to make their models work on their specific problems.

But there's a Catch-22 here: The process is so time-consuming that LOB stakeholders don’t want to do it, preventing AI's effective injection into the organization. This can be a strategic detriment to the enterprise. When possible to use effectively, AI-based tools are proven to help LOB leaders accomplish their business objectives more efficiently and accurately. They also help process data and discover patterns that would otherwise be time-consuming and difficult.

Data labeling aside, there are also ongoing efforts associated with tuning, optimizing, and retraining the models so they continue to produce desired outcomes. Business users need to be part of this process, partnering with the AI team to ensure the ongoing efficacy of their AI applications.

For example, natural-language processing (NLP), a branch of AI, is proving to be a de facto game-changer for organizations, especially when it comes to processing unstructured data, such as free-form text and documents. An NLP-enabled solution can:

• Understand all the nuances in human language intuitively, so users can compose their questions in their own words without having to learn keywords and syntax;
• Effectively process the entire knowledge repository composed of data in varying shapes and forms that contains the answers the user is seeking;
• Find the most relevant and comprehensive information for the user, based on its thorough understanding of the question, allowing users to utilize that information right away for better decision support; and
• Explain the reason behind the results to remove any speculation about resulting AI outputs, leaving users to feel confident about how to best leverage the AI output.

However, NLP is not being deployed broadly today. Why? Primarily, it is because to make each model work, business users must label and annotate a lot of data to train the model so it produces optimal output. This data labeling and annotation work is a time-consuming job that nobody enjoys. Business users often cannot afford to invest the time to label tens of thousands of company- and domain-specific terms. But, without the business user's participation in the development, the AI team will most likely fail to build an NLP solution that can deliver desired business results.

Recently, enterprises have deployed chatbots and intelligent virtual agents (IVAs) to try and leverage NLP attributes, but despite IT’s best efforts, they generally have fallen short in terms of finding customers the correct answers effectively. This is mainly due to interpretation and knowledge hiccups in conversations between bots and humans caused by the inability to fully understand unstructured text; few things can be more infuriating to customers.

Putting LOB in the Driver's Seat

Enterprise LOB staff members need to lead and participate in AI projects in their organization and provide continued guidance to IT/DS/KM throughout the solution procurement and development cycle. It is insufficient and inefficient to rely on these technical leads to find tools without clearly specifying how LOB can use them in their daily workflow and business applications to obtain expected business outcomes. LOB staff must drive this process because they are the ultimate beneficiaries and the ones who will actually fund the initiative.

If the tools being considered lack capability and/or cannot be implemented without a lot of time, resources, and money, LOB employees need to call the shots as to whether to wait until the right tool emerges or to lobby administrators to develop a solution in-house. By leading the charge, LOB can leverage more AI to work smarter and achieve greater return from their AI investments.

Read Part Two of this series here.