Generative AI
Project controls reporting

Project controls reporting with nPlan AutoReport vs. ChatGPT - what you need to know going in

nPlan AutoReport has been specially designed to meet the reporting needs of project professionals - the question is, can you achieve similar results with ChatGPT, the GOAT of chatbots? Dive in to find out!

Written by
Carlos Ledezma
Carlos is a Product Manager at nPlan. His work focuses primarily on the use of Generative AI functionality to super-charge the user experience of nPlan users. He also has extensive background in Machine Learning, which he uses to inform the technical aspect of the features that he works with.

Last week we unveiled a new product: nPlan AutoReport.

With AutoReport, project leaders, project controls folks and planners are able to create data-rich, concise and persuasive reports for capital projects in minutes. 

You may have heard of the technology that makes this remarkable feat possible: it’s called Generative AI. Generative (or Gen) AI has a history going back to the 1960s, but the tool that put the technology on the map is ChatGPT. 

Since its launch on November 30th, 2022, ChatGPT–the fastest-growing consumer software application in history–has become something of a household name. The latest version of ChatGPT, underpinned by OpenAI’s GPT-4o model, can process and generate text, images and audio, has ‘knowledge of the internet’ up to October 2023, and has set records in a number of Generative AI model benchmarks.

As such, it’s not surprising that we have already had a number of project professionals get in touch with us to ask how AutoReport differs from ChatGPT - effectively: ‘can’t I do everything that AutoReport enables me to do in ChatGPT already?’ 

Let’s take a look: 

Understanding capital project schedules

Regardless of which project you are working on, the most recent version of your schedule will contain valuable information about the accomplished and planned progress of your project. A host of project personnel including Planners, Risk Managers, and Project Managers all use the schedule as a way to understand whether their project is on track or not. 

AutoReport (and the Large Language Models servicing it) are able to read and understand project schedules; ChatGPT is not. And the same is true of ChatGPT’s strongest competitors such as Claude and Microsoft Copilot. 

In fact, AutoReport can absorb and reason about information from a number of different schedule formats, including Primavera P6 (.xer, .xml), Microsoft Project (.xml) and PowerProject (.pp) files - right out of the box. 

AutoReport can absorb and reason about information from a number of different schedule formats, including Primavera P6 (.xer, .xml), Microsoft Project (.xml) and PowerProject (.pp) files

Talking the (project controls) talk

As the name suggests, ChatGPT has been fine-tuned to provide information as part of a back and forth conversation. While this humanises interacting with the model, it makes getting information in a format suitable for project controls reporting an awkward business - as anyone who has repeatedly tweaked a prompt trying to get a specific output can attest. nPlan’s generative systems and user experience have been specifically designed for the project controls reporting use case, meaning that ‘prompt engineering’ is hardly required and reports can be pulled together that much faster (in hours - or minutes if a template is already in existence). 

Summarisation good, insights better

One task which ChatGPT excels at is reviewing large amounts of information and summarising it. But AutoReport goes one better - its generative system has been designed to understand all the documentation relevant to project controls and–not just summarise it–but extract insights from it. Crucially, this process is supported by a user interface which makes it easy to ‘provide’ documents to the model (and once your documents have been uploaded to the tool, you can reference them to generate multiple sections of a report without having to re-upload them at any time). 

AutoReport makes it possible to extract the most important information from sources provided by different departments like planning, risk and project controls, and create a report that accurately reflects the progress in different areas of a project - tasks which ChatGPT is simply not designed to facilitate.

Charting your progress 

Paying customers of ChatGPT can now create and manipulate basic tables, and generate simple charts (such as bar and pie charts) from uploaded datasets - as documented in this excellent article on ZDNet.

However, AutoReport’s functionality when it comes to graphs and charts is far superior to that of any other tool currently on the market. Firstly, our system is pre-loaded with all the rich forecast, schedule and risk register data from your project, so you can literally just start prompting, with no need to export data or upload it to a chat interface (with the associated data leakage concerns if you don’t have an enterprise plan with OpenAI). Secondly (and most significantly), we have developed a system that makes it easy for the user to create complex plots requiring lots of data querying.

Take a look at the plots below. The first and second plots show how you can create normal project controls plots with a simple prompt; the third one shows a completely novel plot, that includes thousands of data points, generated from a description in natural language 🤯

"An s-curve for the forecast of a [specified] milestone with a vertical line showing where the planned end date of that milestone is in that forecast"
"A tornado chart that shows the expected saving of the top 10 insights from nPlan ranked by potential days saved"
"A histogram showing the distribution of planned and forecasted P50 start dates for every activity in the project"

In conclusion 

ChatGPT, Claude, Microsoft Copilot - these are all useful Generative AI tools and represent phenomenal achievement; we use all of them from time to time here at nPlan! But their utility in  project controls reporting lags far behind that of AutoReport, because they have not been built or optimised for this specific use case. 

In fact, our aim with AutoReport is to deliver an experience that makes it seem like our users are not interacting with Generative AI models at all - instead, our technology has been designed to make the user feel that they are simply following a step-by-step process to create complex reports in collaboration with the top minds in their company.

I urge you to find out the truth of the claims in this blog for yourself; to book a demo or join our waitlist for a free two-week trial of AutoReport, talk to us.