The industry’s current focus is on perfecting BIM workflows. But these are still very drawing-centric and mired in fallible historic working methods. Martyn Day speaks with Richard Harpham, VP of new start-up Slate Technologies on how a BIM-centric approach throttles construction
Richard Harpham has been an instrumental figure in the AEC software industry for as long as I can remember. We first met in London in 2002 when he showed me a new architectural design solution called Revit. Harpham then went on to join Autodesk and move to the USA to manage the introduction of Revit into North and South America. Subsequently, he led worldwide marketing for both AutoCAD and then all AEC products.
More recently, Harpham led the software commercialisation efforts for shooting star construction startup, Katerra.
Now, along with a team of Silicon Valley Alumni and renowned data scientists, he is part of a new technology company called Slate Technologies that’s currently beta testing new software with some of the largest construction companies in the world.
Martyn Day: A lot has happened since we first met to talk about Revit. How do you feel about where AEC technology is today?
Richard Harpham: Obviously, technology has brought many great things to architects, engineers, and contractors, not least of which is the emergence of new career opportunities directly attributable to technological advances. Also, there is no doubt that working conditions have been improved, and cleaner, more attractive career paths have created a more diverse working population.
However, it’s remarkable to me that most of the basic productivity, waste, rework, error, and cost issues remain largely the same as when Revit was first introduced. It’s also surprising that most of our information still lives in 2D documentation in an incredibly fragmented fashion. As an industry, we may have adopted 3D models, but we still seem to be slaves to 2D, all while leveraging decades old single-core software.
Martyn Day: I remember Revit before the BIM term was created. Why did the BIM idea blossom, and why do you think it became so central to the industry?
Richard Harpham: As soon as Revit was acquired, Autodesk internally wrestled to position it as something different to their existing CAD and 3D solutions. Fundamentally, AutoCAD, Architectural Desktop and Revit were all built to deliver coordinated documentation. Users found it easy to see how Revit greatly improved drawing coordination, but just being better at delivering drawings was a very limiting market position.
So, the Autodesk team created a new term, Building Information Modelling, specifically to better position Revit’s capability to host data. I remember there was debate about using the term Single Building Model (SBM), already widely used for 3D building models, but that term had heavy association with Graphisoft.
After a lot of marketing effort, the BIM term became adopted by some of the more rebellious technologists in large AE firms, with a key moment being a presentation of how BIM was used for the Freedom Tower in New York. Then, almost overnight, any serious architect technologist had to explain their BIM strategy to their firm’s partners, and the first BIM managers started to appear.
BIM was a competitive separator for forward thinking firms and was used extensively to promote their technical prowess at delivering modern design services. Remarkably, this BIM centricity has pervaded almost every aspect of AEC technology development since, both inside firms and in software developers. But, during the last few years, I’m seeing the generally accepted premise, that a ‘BIM’ should be the data core during building production, is now being challenged. Many now argue that it has become more of a millstone than an enabler.
Martyn Day: Do you think focusing on BIM may be the wrong approach?
Richard Harpham: I’d rather not say it’s wrong to utilise BIM, just that it needs to be positioned as one of many contextual sources that can be leveraged during building production. A single project BIM, where all project data must append to a single 3D model, almost never happens, and each profession ends up creating their own 3D model due to contractual risk or lack of trust. That, combined with the sheer file size issues of a centralised BIM, has us searching for software to solve collaboration challenges that do little to advance productivity.
This issue of BIM collaboration has created a ‘throttle’ in maintaining the pace of workflows during building production, which usually means information ‘falls out of 3D’ into 2D documentation to keep pace with the project.
Martyn Day: I don’t think you are saying much new there about today’s issues, but what are you now focusing at Slate on to address the bigger problems?
Richard Harpham: We’ve spent considerable time speaking to firms that are overwhelmed by their digital data, including BIM, finding surprisingly little of it is being used during critical decision processes.
We’ve also proven that simply digitising our legacy processes, as most current construction software does, is not providing the productivity returns we’d hoped for. We may be at an important watershed moment, where existing and trusted methods, based on a human’s cognitive abilities, may have reached their limit.
The team who formed Slate became increasingly interested in better understanding how our legacy processes and decision-making habits might hold the answers to improved software design. Numerous studies have proven that humans generally focus more when trying to avoid losses than find gains, sticking to the original plan whenever possible. This decision bias, when multiplied over weeks and months of decisions during building production, crushes most attempts to improve productivity.
One from the archives: the term BIM really took off in 2005 when Revit was used for New York’s Freedom Tower
Almost every software tool out there is trying to keep you on the original schedule or minimise change as it could increase costs. In AEC, we have inadvertently taught ourselves and our software that change is something to be avoided, as we can’t predict how good or bad a change might be.
We experience similar illogical human behaviour in everyday life, where someone might say, “I know there is a new road that is bound to be faster to the shops, but I’ve always driven this route, as I know the way and how long it takes.” Of course, most of us now trust the software in our car or smartphone that suggests ‘change opportunities’ all the time like ‘take the next exit for a faster route to avoid Traffic.’ GPS software is built to predict and imply positive change opportunities. You still get to decide whether to take the advice, but at least you now know.
Martyn Day: So, you are implying this is a process design problem, and we need to move away from trusted methods that have worked in the past?
Richard Harpham: Well, in almost every other human industrial endeavour, we are moving towards more agile context driven decision processes that see ‘change’ as an opportunity to improve, not an as an issue or event to avoid. Many professionals, tasked with digital transformation, are looking for solutions that help them improve governance of the very processes that discourage change occurring during building production.
When did we decide change was always bad? If you were relying on that GPS to steer you away from traffic, you’d be annoyed if it told you after the change opportunity. ‘I knew there was traffic ahead, but I kept you on the same route, because you always go that way.’
The new generation of software tools need to be a real-time assistant helping you make better decisions. Just as when starting a car journey along a GPS planned route, the moment you kick off a building project, all the context just changed. So, just like a GPS, your digital building decision assistant needs to search for patterns that can predict positive change opportunities, presenting them at the right time to improve outcomes.
Martyn Day: So, this is what we can expect from your new company, Slate. From the sound of it, you’re proposing yet another AI/ML solution?
Richard Harpham: Yes, we are all getting a little fatigued with those terms. These are early days in even understanding how best to employ ML and AI. Yes, we have several BIM-focused, ML/ AI driven image capture for comparing progress on site to sequenced models. But, for those solutions to be successful, we must wait for the model versions to catch up, to check that the construction is keeping up with the schedule. It makes my head hurt to think of how much additional effort and cost they could introduce if we try to do it with the current BIM tools.
At Slate we started by first studying why decisions are made the way they are during building production, so we can understand why seemingly avoidable issues occur repeatedly. We clearly found that many of these repeated inefficiencies result from human behaviour, rather than digital inefficiency. So, before we train our ‘machines’, we’re making sure we better understand what machines need to do well to overcome what humans do badly.
Let me give you just one example. A prize-winning study from Christopher Chabris and Daniel Simons deals with the effects of cognitive illusions. They studied two fascinating human behaviours: Inattentional Blindness and Change Blindness. These have evolved over millennia to help people deal with the overwhelming amount of data our eyes are presented with, leading to very narrow cognitive focus.
Unless it’s something like a tiger attacking you from your side view or something moves so slowly it is not a threat, your brain just won’t see it. While this helped us not get distracted in the jungle, this auto-filtering can blind us to potentially valuable decision context during our working day.
Computational machines, such as we are building at Slate, do not need to filter visual information in the same way. Once trained, they will be able to see, count, relate, introspect everything presented to them in their focal range, then predict potential issues or opportunities without bias, without missing anything.
In this way, Slate’s goal is to reveal valuable insights that a building professional would not ordinarily be aware of, helping either avoid issues or take advantage of opportunities to improve outcomes.
Martyn Day: It sounds like it’s going to take a lot of data, and in a process that has 28 firms not sharing data that well (or happily) how can these AI/ML solutions get a clear picture of the overall project? Or is this only for discrete processes?
Richard Harpham: While we do have a data sharing and access problem, we don’t have a digital data shortage problem. Recently, data-specialist Splunk Inc. released a study stating that as much as 55% of potential data in an organisation is considered “dark”, that is, data that is unknown, undiscovered, unquantified, underutilised or completely untapped.
They coined the term ‘Dark Data’, to describe data held in siloes and software, scattered across a firm’s ecosystem along with their vendors. While working with several General Contractors, we’ve discovered a massive existing resource that might provide huge value in delivering decision context. One common example comes to mind. Nearly every firm I’ve talked to creates a form of ‘lessons learned’ document after a project, but none I’ve met have leveraged these as a resource for future work in any systematic way.
While Slate is not just for discrete decisions, we’ve started by trying to understand what individual personas in the building process might benefit from for decision assistance. We’ve seen firms build ‘Mission Control’ approaches to centralise decision making, but this has proven expensive and hard to implement.
At Slate, we are attempting to introduce something much easier to adopt by individuals using their mobile devices, focused initially on augmenting decisions before, during and after their scheduled tasks. We’re also evolving easy to implement cross firm data integrations, intersecting multiple data streams to reveal valuable opportunities that otherwise might never have been found fast enough to impact outcomes.
Martyn Day: What can we expect from Slate’s commercial solutions?
Richard Harpham: Early this year, expect our ‘mobile-first’ solution that can provide immediate value to decision makers during the pre-construction and construction phases, giving individuals immediate decision context during their everyday tasks. Slate then leverages its proprietary dynamic scheduling capabilities to ensure resulting change decisions can immediately update the overall schedule as well as the order of individuals’ tasks.
Soon to follow, Slate’s increasing number of integrations with subcontractors and material supplier’s software and systems, will create data insights, valuable to the Executive suite as well as the individual executing tasks.
I firmly believe we are at the start of a significant shift in how we deliver buildings, where a new set of ‘machines’ works hand in hand with humans, to support the industry in its long overdue increase in productivity and profit.