Project planning software developer nPlan is reporting good progress towards its goal of predicting construction timelines more accurately through the use of machine learning/AI techniques.
Speaking to BIM+ at Digital Construction Week, Toby Buchanan, general manager at nPlan, said the two-year-old start-up has just begun collaborating on new projects with Network Rail and Shell Oil Company.
The information gathered will strengthen group data it has been assembling from working with 19 clients over the past year, including contractors Keir, Laing O’Rourke and Costain and clients such as HS2, and ultimately increase the accuracy of its predictive planning tool.
In recent months, nPlan has expanded its product focus, going beyond its initial web app to offer a set of APIs (application programme interfaces). These APIs do not require bespoke websites, making it easier for users to integrate with their existing reporting tools such as PowerBI and Tableau.
Toby Buchanan: much greater knowledge now
Commenting on the sector’s growing awareness of AI, Buchanan said: “It’s been interesting to see at Digital Construction Week that there’s much greater knowledge now than there was a year ago – people are far more aware of what it can do and how they can use it.
“There’s also a growing recognition within construction that we’re on a ‘burning platform’ and we need to move more quickly to adopt AI technologies. Even the old, slower companies are now recognising that.”
nPlan’s USP is that it learns in granular detail how construction projects have performed historically and converts that knowledge into algorithms that predict the outcome of future projects. As co-founder Dev Amratia (pictured above) puts it: “We learn what was planned to happen and what eventually happened. We’re an impartial source, with no bias or ‘eternal optimism’ which allows for much better decision making about how long a project will take.
“We have half a trillion dollars worth of projects behind us to provide the realism, which derisks the project, reduces waste and reputational damage and increases the confidence of investors.”
Buchanan adds: “Our patent pending algorithms can understand project schedules and can self organise to emulate human experience, at a scale and accuracy that has never been seen before. We have trained our algorithms with around 250,000 schedules, making it the largest library of schedules in the world.
“Each organisation that works with nPlan chooses to share its data in a secure and anonymous manner and for every new organisation that joins, each existing organisation gains the benefit of the collective intelligence.
“Our algorithm draws on many different attributes embedded in each single construction activity. For example, an activity named ‘pour concrete on level 15’ can be categorised based on the words in the activity name but also many other criteria. All these attributes have an impact on how likely an individual activity is to finish on time and, in turn, how likely the project is to finish on time.”