Machine Learning
Time management as an opportunity for staff wellness and task automation is the hardest problem to consider when delegation is presented and negotiated by old contractual methods based on deadlines. This obstacle offers potential for machine learning to identify the strengths and weaknesses of how management distributes duties. The technology also offers insights into profit margins and actual staff value. It is a slippery slope to audit a company and identify that the amount of overtime needed to complete projects is due to not automating set up process, but also not recognizing the difference between projected time to model 3D geometry versus the actual time that it takes to create the 3D geometry including revisions.
Two case studies to take into consideration are mining of the 3D BIM Model regarding how staff make design changes and revisions, and the overall organization process for the 3D BIM Model set up. In the first example you can mine how the production staff thinks about modeling complex geometry and recording it for patterns so one can recreate simple tasks such programing on square footage, sun path for energy analysis, adjacency of building to transportation, material usage and, more fundamentally, the amount of time required to complete modeling elements.
The latter example is something we currently employ in the industry; however, this is tasked to industry insiders that use plugins to automate certain processes such as printing, coordinating clashes between trades, linking files. There should be a way to mine the everyday use of these tools and workflows so that through relevant data across multiple projects, machine learning can specify the order of operations and where one sees more efficiency in set up process.
The dissonance of both examples is that there is no set standard as to what software should be used; what steps to take when setting up coordination between the different building trades. Machine learning can be leveraged to identify patterns in successful projects by reverse engineering how the 3D model was produced and giving the models an audit grade. This grade can be mapped against inefficient projects. The process of staff making changes would be more of a micro investigation into how an office works that then could be relayed to the client in a dashboard format similar to a health chart. Further extrapolation could reveal better contract negotiation options where currency is exchanged through smart contracts based on the live virtual models and the actual production of 3D elements versus the 2D drawings that we currently submit in the field.
The push forward would be to start combining the human and virtual aspects with machine learning to make design more innovative as well as reduce time clocks and deadlines while producing more profit and efficiency.
Natural Language Processing: CRM, PDF Translation, BIM Execution Plans, API Narratives
Customer Management Relations
Natural Language Processing (NLP) can translate and synthesize the spoken word; what if one were to combine machine learning with recordings of multiple sales calls. An organization would be able to identify key phrases and words that are used to close deals. The crux of this process would be to differentiate between vocabulary and charisma. There is a strange magic in sales calls that requires investigation as to whether the pitch or the gusto of the conversation wins the deal. This can be further extracted to understand the cold call and when it’s an appropriate time to hand off to a more specialized manager. Hand off time and analysis of conversation would be an asset in training and value of cold caller. One would have to get past the big brother syndrome for the future of sentiment.
PDF Translation
In the construction industry we still use 2D drawings that have matured into a digitized vector or raster format known as the PDF. These drawings are composed of lines, hatches, and geometry to explain space, mechanical systems, structural systems and so on. Each trade has an unwritten standard. While the NLP is used for words, it should not limit its understanding of drawings. If one considers Japanese symbols as drawings one can use the same logic in deconstructing the language that is used to create drawings for buildings. The documents are vast in saturation. Once this conceptual form of analysis is deployed, we can establish a global drawing standard based off work that has been built versus standards established by institutions. Furthermore, we could identify redundancies in overproducing information to cover liability. Essentially this task is given to Quality Control leaders of the industry to define if a drawing is apropos. The value is in democratization but also an actual standard that reduces mistakes.
BIM Implementation Contracts
In AECO we spend a vast amount of time sieving through BIM Execution Plans and auditing them for offices. These are a set of measures usually written for a client by an owner’s representative that all trades must try and follow from software usage to 3D model Level of Development. A fundamental problem is identifying the workflow of an office and how to make it conform to the client’s plan from design to operations. The ideal situation would require all parties involved to have their own BIM Execution Plan, respective to trade. In this scenario one could use NLP to help identify similarities and major differences in the global contract that is owned and revised by the client to match workflows of those hired to do the work. It would also underline risk factors and cost benefits in a more distilled format. Often those tasked with modeling and constructing the project do not read these documents as they tend to be lengthy and technical. Another way NLP could be used is to distill the important portions for each trade. The value here is quality control on the owner’s side as well as participation by the trades to a unified workflow.
API Translation into Narrative
The ultimate use of NLP would be in translating the creation of geometric 3D models into a narrative format. For example, we could use NLP to translate API information of Autodesk software such as Revit into a narrative based conceptual story that project directors or design stake holders can understand. This would bridge the gap between the production effort but also explain what is going on under the hood. GraphQL is doing this to a degree for development of tools in a query. Translating API’s into a formal language that anyone can use gives one a direct advantage when talking to clients as it becomes a story and internally prescribes what can be expected.
AI Implementation
As AI becomes a more mainstream tool in AECO, new roles will be created with the opportunity for growth in management and the rise of new departments. Because the industry is latent in emerging technologies, this would be early adoption, both opportunities to upskill and reskill will apply. This is a growth opportunity to create new services that also influence internal management applications such as CRM, sales and consulting. In an AECO company the leap to explore would mean great differentiation in the future market.
Time savings will be the major shift of establishing leadership in the industry, which would allow for more research and development. Buildings are the perfect combination of tools and humans. Currently machines play the role of creating the content that is designed and orchestrated by people. We would be offsetting the coordination and error involved in creating a document set and constructing the building. AI machines will create new innovations in materials, schedules, space planning, and sustainability. Humans will have better quality control, more time to reflect on the function of the designs and maintain cost savings. Essentially, we would be introducing better tools to humans that would be autonomous with direction. The role of people would be to standardize how we use AI in the AECO industry.
