4 opportunities for AI to improve capital projects and construction.

There is a time when all the seemingly random experiences we have in our life and career come together, in one place, to (hopefully) make a difference in our world. That time has arrived for me and I’m writing this to share what I’ve learnt as the CEO of a 6-week old company. With great pride, I’d like to introduce you to nPlan, the UK’s latest AI startup. It has been 6 weeks since I formally started the company building process but it has probably been about 9 years since I informally started the process.

For context, I worked for Shell for 9 years, mostly as a Project Engineer on flagship (those worth >$500M) projects around the world.

I joined Entrepreneur First with a view that the way the world executes mega-projects continues to be deeply inefficient, and of course, I’m here to fix it. You’ve probably just rolled your eyes with that and I would have done the same when I began 6 weeks ago. In week 1, Matt Clifford told us that we absolutely had to bake company ambition in early or we’d forever lose the opportunity. The challenge with starting with near limitless ambition is in building laser-sharp focus on the “one thing” to fix, that if solved and scaled to the nth degree, can unlock tremendous value. This is where the value of customer discovery is absolutely critical to get right.

Building a business involves speaking to potential customers to find out what their problems are, to ensure we build what they want. It helps prevent us from building orbs that nobody wants except its’ creator. Understanding customers is challenging and I struggled to be effective initially, often feeling the need to just tell a potential customer what my idea is, in the hope for validation to escape the awkwardness of being an effective listener. Together with my super talented co-founder, Alan, we spoke to the industry of engineering and construction companies, to discover insight we can address. We met close to 40 people from CEOs to graduates interns and from MPs to academics. Here are four things we learnt and are now passionate to dramatically improve:

1. Projects are executed based on the collective brainpower of the project team and a handful of consultants. There isn’t a way to rapidly and seamlessly transfer knowledge within an organisation and as a result, repeated mistakes occur which costs the industry dearly. These costs often manifest themselves as liquidated damages and claims.
2. Project estimating and scheduling is often considered an art and not a quantitative science. These art pieces contain flaws in either logic or quantity and as a result, get disregarded much sooner than you’d want.
3. Risk is often allocated subjectively to schedules and estimates, which leads to inflated waste. Subjective risk often holds a large range, which means that specific items or targets are shrouded in uncertainty and are undermined because of a low credibility. If the project team doesn’t believe they have a realistic plan, it gets dismissed and overruns become increasingly likely.
4. Projects are all different and complex. The existing frameworks to manage size and complexity disconnects management from the work face, by creating layers. At each layer, information is filtered before proceeding upwards, which creates disputable awareness of how a project is actually doing. When things go off plan, two things happen, the first is that it takes forever (~30 days) before the management team learns about the deviation and the second is once discovered, they won’t know how best to react to ensure schedule/cost outcomes are maintained for the overall project. The tendency is to fight the fire, which inevitably allows 2nd or 3rd order critical path items to creep and bite back, resulting in a larger fire.

These insights have shaped us and we plan on discovering more about them, with the goal of eventually becoming experts in the field. What Alan and I are confident in, as are our first trial customers, is that if we can build a solution that results in a 10x improvement on the points above, it will be huge. Deep learning on project datasets is the most likely method of cracking this nut open. The good news is that we’ve come to this with the right tools, by asking the right questions. We’re excited about what lies ahead!

For my next post, I’ll begin to unwrap each of the points above and link how Deep Learning might work to unleash the value we’re all desperately waiting for.

If you’ve got any questions or thoughts about the above, get in touch on dev@nplan.io.