Decision makers in construction do not have access to the right information at the right time. This causes a lot of inefficiencies, lack of productivity, delayed projects and cost overruns in the $10T construction industry.
Gryps leverages robotic process automation (RPA) and artificial intelligence to create digital robots and augment them with machine learning to solve legacy problems in construction. We will improve the construction industry in a meaningful way through empowering stakeholders with the power of quality information.
Gryps solution: connecting to information silos
Automatically get the information from the source: We call this the “Ingest” process. Construction data is fragmented and this means that mission critical information resides in multiple sources since each company needs to maintain their system of record. For example, the general contractor might use a Procore or Prolog product but the architect might use Sharepoint or Newforma products while the facility owner uses eBuilder or PMWeb products.
Challenges with APIs: The common approach is to use Application Programming Interfaces (APIs) to ingest the data and documents into data warehouses and storage systems. It is very common to see analytics solutions offered on top of data warehouses and data lakes. Some modern ones use Big Data Analytics techniques to leverage large amounts of data or they use streaming techniques to provide near-real-time insights.
While these approaches are very interesting and useful, in our experience, they do not work well for the construction industry because:
- Limited API Functionality: Some of the applications used in construction either do not have complete APIs that support all functions or do not have APIs at all
- Fragmentation: Each system is owned and maintained by a different company and IT organization.
- Managing Permission Levels: Some stakeholders cannot get access to the APIs because the system does not support granulated access provisions on the API level or it is a burden for the IT organization to provide. Even in cases where APIs are supported, teams are given access to the user interface but not to the APIs.
These issues prohibit most construction projects from using existing out of the box integration Platform as a Service (iPaaS) options. This means teams:
- Access old data
- Log into multiple systems to search and find the latest documents
- Lose access to data if a contractor is terminated
- Duplicate data entries
- Spend significant time verbally updating each other for reporting purposes
Gryps solution: providing people access to the right data
Focusing on Documents: The most important information in construction is in documents. There are millions of different types of documents across the construction industry. For example, contracts, invoices, payment requisitions, drawings, insurance certificates, equipment manuals and permits. These are either documents to start with or data exported to documents when transferred between companies. It is the simplest, most universal method of sharing information.
Based on 25 years of building technology solutions across the construction industry and studies on customer data, on average around 80%-90% of documents in construction are in PDF format. Currently around 50% of these PDF documents are scanned PDFs. They are typically stored in many diverse storage silos which makes it impossible to find easily.
Collecting Document Context: The context around the documents is equally important. For example, knowing if a document is an approved, a draft, or a rejected contract and who created it makes a big difference.
Understanding Documents through Machine Learning: Gryps uses document understanding AI techniques to digitize, classify, and extract information from the documents. These results are then combined with the document context information ingested from the source system to provide comprehensive knowledge mining capabilities to the user.
Gryps digitizes these documents using construction-improved OCR technology. This is an important step in improving model accuracy.
The challenge with machine learning models in our industry is to achieve high levels of accuracy with the limited amount of annotation available in the construction domain. In recent years, transformer-based transfer learning approaches have achieved significantly improved performance over the traditional methods (see BERT). Gryps leverages these techniques to improve the accuracy of the models. This means that these models are pre-trained on a much larger corpus of data from other domains and then fine-tuned for the construction vertical.
Our solution:
- Ingests documents from diverse silos and then enables users to analyze the data and deliver real-time actionable analytics not available without Gryps.
- Orchestrates the digital robots so that the complex tasks are done to help our customers be more efficient, effective and increase ROI.
- Uses RPA and AI techniques to create rule-based digital robots and augment them with Artificial Intelligence in order to solve long unsolved problems in construction.
We are on a mission to improve the construction industry in a meaningful way through empowering stakeholders with the power of quality information.
This is just the beginning…
Our plans include evolving our digital robotic services from rule-based to intelligent robots. We plan to augment the enterprise data with open source data from the web. These are small parts of our vision to empower people working in the construction industry.
Are you a Brilliant engineer who is looking for exciting deep technical challenges that will modernize the gigantic construction industry and create exponential value? We are looking for smart and passionate people to join our team. People who want to build cutting edge robotic process automation tools, leverage advanced computer vision to analyze millions of documents and deliver artificial intelligence that will help the construction industry escape the dark ages.
We are hiring and we look forward to hearing from you.
Are you a real estate owner or work at a construction company with hundreds of thousands of documents stored in diverse software silos? We would love to speak with you!
