FacebookTwitterLinkedIn
  • Blog
April 24, 2019

Requirements for Enterprise-Grade Natural Language Generation

By: Claire Cunningham

It may sound obvious, but in order to effectively implement a new form of innovative technology across an enterprise, it is important for a business to fully understand and outline the requirements that will enable that specific type of technology to be successful. Often times, there is such a demand to be perceived as “on the cutting edge” of the competitive landscape that companies will hastily funnel a request to “figure out this new tech” to their IT department, not taking into account some key elements to ensure that this (likely significant) investment of time and money is beneficial for their employees and/or customers.

One such innovative technology that is receiving a lot of attention from large-scale corporations is Natural Language Generation (NLG), which is the concept of utilizing structured data as an input that ultimately produces human-sounding language as an output. Enterprise companies are leveraging NLG today to automate burdensome finance and compliance reports, as well as complementing or enhancing current data analytics tools. However, before embarking on an NLG-driven quest, it is key to understand two fundamental requirements for making this endeavor successful:

1. Securing buy-in and involvement from the C-suite all the way to the IT department

A huge factor in driving a successful NLG initiative is a reasonable level of involvement from an organization’s leaders, in terms of cross-functional partnership and business-side guidance. Considering that enterprises see artificial intelligence (AI) and process automation as one of their highest strategic priorities, only second to cost reduction, we can understand that these types of initiatives are mission-critical to business leaders. In order to achieve these priorities, it is crucial for executives to research and understand areas of their organization that would benefit from this type of technology by highlighting suboptimal processes based on employee feedback and incrementally exploring how AI and NLG can automate these processes. However, they cannot achieve these long-term objectives alone.

Stakeholders such as IT teams, data center managers, business analysts, and other automation influencers within the organization will also be required to make these solutions a success. Without the right data team to aid in the automation of sending data to a NLG system, the business analyst team to help inform the requirements of a respective solution, and some sort of development team to facilitate the final delivery of the natural language output, an enterprise risks investing in an idea without following through with an execution plan.

2. Identifying strong use cases with defined success metrics

Ensuring that your organization has successfully identified processes that can and should be automated is a major requirement for enterprise-grade NLG. The automation of these processes typically will result in operational efficiency by replacing manual efforts and/or enhancing engagement by transforming untapped data into new products or reports. Here are a few elements to consider that will help identify ideal NLG-driven opportunities and solutions:

How much time does my team spend creating data-driven reports? NLG will be highly valuable if it takes your team several hours, days, or even weeks to analyze and write reports explaining insights from data.

  1. Is my organization’s data structured? NLG leverages structured data within databases, spreadsheets, .csv files, or machine-readable formats like JSON.
  2. At what volume or frequency is the report issued? NLG will be highly valuable if the cadence of the report is frequently produced, customized based on audience type, and read by many people. Successful use cases range in volume from 200 stories per year to upwards of thousands.
  3. Are similar reports being drafted by individuals with varying levels of expertise? NLG enables a level of standardization and consistency that would not be possible with a team of dispersed human analysts that have differing levels of skill sets or writing styles.
  4. Do we have data that is being underutilized, and is there information in that data that can add value to your employees or customers? NLG will be highly valuable if your internal or client-facing dashboards are not being adopted due to a lack of analytical skill sets or lack of resources.
  5. Is there an opportunity to integrate language into new or existing data product offerings to enhance user engagement? NLG provides the opportunity to increase engagement in new or existing products by incorporating language that can lead to widespread adoption because the information is easier to consume.

It is equally important to note that, once you have highlighted strong business cases for NLG, it is key to decide on use-case specific success metrics and exactly how a new solution will be rolled out to the end user. Many enterprises struggle with developing a centralized governance structure for these types of initiatives, as some projects can become too siloed, do not have specific success criteria outlined, or lack organized training to utilize the new solution to its fullest potential. However, solving this obstacle may be as simple as getting all respective parties from the C-suite to the technology stakeholders in a room and answering the straightforward question: “What would a successful rollout of this solution look like?”.

The good news? If the two steps outlined above are followed, it is easy to replicate success across the enterprise, as you now have a playbook for operationalizing enterprise-grade innovative technology.