top of page
  • Writer's pictureJennifer Bittinger

Natural Language Generation (NLG) Empowers Work-In-Progress (WIP) Reports

The hope for most new technologies is for workers to focus on high value tasks while tech tools can complete somewhat mundane and repetitive tasks. The Array team is utilizing artificial intelligence, machine learning and natural language generation (NLG) to solve a broad range of workplace challenges for the construction industry with the aim of automatically completing low value tasks but also providing high value insight from data. NLG summarizes data, data tables and graphs into a textual summary, removing the risk of misinterpretation of data and speeding up the process of creating an accurate summation of what the data is conveying to an analyst or project team. Research has shown that textual summaries can be more effective than graphs and other visuals for decision support and that computer-generated texts (i.e. NLG) can be superior to human-written texts.

For this article, the primary topic will be focused on work-in-progress reports or what is commonly referred to as “WIP’s.” The WIP report is essentially a project schedule document that tracks active construction jobs of contractors and defines whether the project jobs are over-billed or under-billed.

First and foremost, the Array team believes that NLG can greatly empower Construction Project Managers which in turn empowers finance and accounting teams at their companies. NLG can help Project Managers (PM) explain the numbers and to provide further context to the projects. A cost-to-complete estimate is what many departments of the company want to know but it needs to be exact. The estimate determines where the job stands right now, which informs if the time to completion is on track and will be completed to requirements and specifications. Subsequently, PM’s must have accurate knowledge of real-time costs and the ability to explain spreadsheet numbers and models. NLG can quickly empower these PM’s by producing frequent reports and hyper-accurate reports because it explains the numbers and the state of a project with textual summaries and a story narrative of how the project is developing. It is common for an excel spreadsheet to indicate that a project may be 40% complete but a PM may look at the whole project situation and know that the reported 40% completion rate is not correct because the PM has more contextual knowledge of the whole project. Using artificial intelligence (AI) powering natural language generation to produce these textual summaries enables good communication and highly valuable conversations amongst various teams within construction companies and their financiers.

First, let’s review what WIP’s are and some best practices. WIP reports are essential for project managers, accounting, and finance teams to effectively manage and observe all key progress indicators on a project. Proper WIP’s will identify issues before they become bigger problems and monitor red flag issues. Effective WIP’s and their project managers collectively provide explanations that produce insight on how to deliver their projects on-time, on-budget and as-described.

Natural language generated summaries of excel data and financial numbers will make WIP’s more effective. Best case examples of WIP's will identify the state of over-billings and under-billing’s such as:

Over-billings: · Billings, more than actual progress · Customer is funding project costs · Adjustment artificially lowers income which in turn lowers taxes Under-billings: · Billings, less than or behind actual progress · Contractor is funding costs · Adjustment artificially inflates income which in turn increases taxes

Complicating matters is that work-in-progress reports (WIP’s) do not have a universal format. These reports vary from company to company to fit an organization’s specific set of needs. However, WIP’s generally include the contract amount, estimated costs, costs-to-date, the percent complete, billed revenue, earned revenue and over/under billings.

Here is an example of a work-in-progress (WIP) report and an artificial intelligence powered natural language generation (NLG) summary of that WIP report. As you can see, the NLG summary explains the report with context and insights on what the WIP report is saying. Having project managers, accountants and finance analysts receive these NLG summary reports, which empower WIP’s is invaluable.

How does an organization improve their WIP's and utilize a natural language generation platform for their business?

The Array team suggests that companies utilize a process that addresses key milestones to achieve. These milestones are focused on specific types of deliverables and a commitment to execute the process to empower WIP's and use artificial intelligence natural language generation tactics. Below is a graphic that outlines the three basic phases of implementing WIP's and NLG.

The first phase, “Document Planning”, is focused on determining the “problem to solve.” These problems may vary but a general theme may be information gaps and receiving information at specific times throughout the lifetime of a project. It is important to plan out what content and reports should be produced and how this new content will change the organization.

The second phase, “Microplanning”, is focused on identifying the types of language, industry terms and keyword indicators that must be used within the content that will be produced.

At this phase, companies should be discussing actual costs versus the budget, as well as the billing progress. You’ll also want to review the status of past change orders. This is also the opportunity to compare the project manager’s process in the calculating percent complete estimate with accounting’s calculation. In regards to WIP reports, the team should identify:

· Any under-billings, since these are typically undesirable

· Completion delays that don’t have accompanying change orders

· Overdue receivables

· Identify Unsigned Certificate to Occupy (CO’s) on completed work or unapproved CO’s

· Any substantial difference between the field’s and accounting’s percent complete

The third phase, “Realization”, is focused on converting these content and reporting requirements into real text. These text narratives will be produced during this phase and evaluated by project managers, accountants, and financial analysts. Within the NLG development phases, this phase will produce a CORPUS which is a body of written work that will be produced, tested, and optimized in a learning environment within a self-learning artificial intelligence platform.

At the end of this process, companies improving their WIP’s and utilize NLG to align the entire organization on the Cost-to-Complete at every stage of a project on all of their projects that will insure profitability, lower tax consequences and most importantly, will deliver a quality product to win repeat business.

Although we have focused primarily on how natural language generation (NLG) can improve and empower work-in-progress reports for construction and engineering companies, there are in fact many NLG applications for various other types of reports that address key performance indicators (KPI’s). Many of these KPI’s can be included in executive data dashboards that can be enhanced by natural language text summaries:


For more information on Array and NLG, please contact:

Jennifer Bittinger at


bottom of page