Digital & AI: Revolutionizing Infrastructure with Digital and AI
By: Heidi Kim, Manik Jain, Patrick Roche, Amir Ganaba, Sheila Ho, and Alex Vickers
With the historical $1T of federal funding now available through the Infrastructure Investment and Jobs Act (IIJA), the Inflation Reduction Act (IRA), and the Creating Helpful Incentives to Produce Semiconductors and Science Act (CHIPS), the U.S. infrastructure sector is faced with both an incredible opportunity to build for the future, and a set of challenges in meeting this capacity. With the sector already facing strain from climate change, legacy ways of working, and a shortage of qualified workers, the demand for sustainable infrastructure that is efficiently delivered and maintained has never been more pronounced.
To deliver greater efficiency and outcomes, other industries have not only unlocked the benefits from analytics, AI, and automation, but are already beginning to tap into possibilities brought to life by Generative AI. However, the public sector and construction industries significantly lag other industries, struggling to adopt new technologies due to the time required to update legacy technologies and effort required for transformative change of organizations entrenched in long-standing, risk-averse practices.
The ninth article in our series, Bridging the Gap: Propelling America's Public Infrastructure Forward, examines how to harness digital and AI solutions to drive outcomes across the infrastructure lifecycle.
To stay ahead, it is crucial for state leaders to recognize the immense potential of AI and its integration into infrastructure. Recognizing the starting point for digital and AI for most public sector and construction organizations is significantly less mature than for other industries, we recommend four key points for states to focus on initially:
Identifying and targeting use cases across the infrastructure lifecycle
As a first step in understanding how to unlock the power of digital and AI, states should start with identifying use cases across the value chain for infrastructure:
Development & Planning: Digital and AI enables governments to improve long-range planning by providing data-driven optimization across the portfolio of infrastructure investments, improving cost and schedule forecasting, and more accurately modeling risks by scenario. For example, tools such as creating a digital twin, or virtual replica, of a city or infrastructure networks can enable states to optimize capital investments in a virtual world before applying them to the real one.
A few additional examples of use cases of Digital and AI in development & planning are:
In one example, the Netherlands implemented digital and AI tools to optimize their water systems based on projected demand and climate risk. This allowed them to make decisions based on short-term needs while accounting for long-term changes that were predicted from current data trends.
Design & Engineering: Digital and AI can be implemented in design and engineering to optimize design and mitigate risk. For example, it can redesign elements in infrastructure to combat the largest climate risks and/or make design decisions (i.e., more expensive, more durable materials) that increase cost upfront but lower lifecycle O&M costs due to lower maintenance. It can also be used to facilitate cross-discipline coordination in design by catching design clashes and mitigating them early on. AI can even be integrated with GIS through AI-powered drone mapping for safe and affordable land surveying.
Measures like these can substantially improve the overall lifespan and health of infrastructure while decreasing costs. Additional uses include:
In a recent article, the Harvard Business Review discusses the importance of Building Information Modeling (BIM) in construction. BIM can convert conventional blueprints into digital visuals, enabling the management team to execute a project twice: initially through a virtual platform to identify tie-ups, then subsequently constructing the actual project based on lessons learned. For example, a contractor used BIM models for Oman’s Muscat airport to help coordination across disciplines (i.e., architectural, structural), adjusting resources to align with shifting schedules, and identify clashes, resulting in a $7M savings and zero change orders.
Construction: AI-driven analytics help optimize planning and coordination by providing real-time transparency on staff, resources, risks, and progress, and by optimizing processes (e.g., identifying processes that can be done concurrently). Prior to project start, AI can also be utilized in resourcing to optimize procurement (e.g., calculating trade-offs between material quality and cost).
Finally, digital and AI can be used during execution in several ways...
In one use case, Mortenson used digital and AI to optimize scheduling during the construction of the U.S. Bank Stadium in Minneapolis, allowing the owners to continue to receive revenue by starting construction of the new stadium before demolishing the old stadium.
Operations and Maintenance (O&M): Digital and AI will be critical in the shift from costly reactive maintenance to data-driven predictive maintenance, improved demand management, and coordination and streamlining of the field force (for which many processes are still managed via pen and paper).
Some of the uses of digital and AI in O&M are:
For example, the Cuomo Bridge utilizes over 300 sensors for predictive maintenance. These sensors give operators data on the health of various components of the bridge (i.e., cables, concrete) and inform operators of maintenance that needs to be performed. This saves states substantial resources and costs.
Collecting the data and building the data infrastructure
Identification of use cases represents the first step and critically, helps states prioritize where to begin with data collection efforts. States should first take stock of existing data sources, identify gaps, and begin to build a robust set of required data and supporting infrastructure. There are three components to consider:
Data collection: A robust data collection process is vital for training the foundational model for selected use cases. To start developing and collecting data, states can consider:
Recommended by LinkedIn
Data standardization and open platform access: In order to utilize and maintain the data across agencies, states must implement standards to ensure data formatting is consistent (i.e., same formatting and language) and compatible with different agency systems. States can consider creating open-access platforms to allow government agencies, educational and research institutions and private sector access. By allowing open-source contribution and use of data platforms, states can accelerate the pace of development for AI solutions.
Data automation and security: Automation of data collection and processing is a critical step, and the effort required to do so is often overlooked. For example, many DOTs have utilized LiDAR (remote sensing technology to collect accurate and detailed 3D data about the quality and status of objects and surfaces), but ultimately become outdated due lack of automated data flows. Automation of data uptake and ingestion will allow states to use data while it’s still relevant.
It is imperative to create a cyber-security framework to ensure digital standards include aspects of data security and privacy while detecting and monitoring threats.
Building and maintaining AI foundational models to provide actionable analysis
The foundational model or “AI” ingests inputs from a series of data (e.g, sensor inputs, past projects, and data on climate shifts) and recognizes patterns to identify optimal solutions across use cases. Using a combination of human guidance and machine learning models will result in better and more consistent predictions from data sets, with machine learning continuing to refine after initial human input.
To create these models or AI, states must understand their internal capabilities and how they must evolve. Some questions that will help guide states into understanding what is needed for their foundational model are:
How will the foundational model “AI” be built?
States will need to form an unbiased assessment of internal capabilities along a variety of dimensions (i.e., data collection, machine learning). States can develop a framework to understand what parts of the system they can develop and what parts of the system they may need to outsource.
What capabilities do states have to store and share data?
Foundational models take a wide variety of data from the whole lifecycle of infrastructure to optimize each step within the lifecycle. Agencies must have mechanisms (i.e., platforms, standards) to share data and capabilities across agencies to ensure that information and capabilities are not being siloed.
What technology infrastructure is needed?
States must understand the gaps between their current technology and the technology they need to sustain the data and the foundational model (i.e., cloud computing, scalable processing capabilities).
How will this be funded?
The foundational model, along with data collection, often requires significant financial investments for technology procurement, training, and ongoing maintenance. Some states may face budget limitations that restrict their ability to allocate funds to such initiatives. States should seek to allocate dedicated funding for digital and AI initiatives, including creating funding programs, grants, public-private partnerships, and incentives to support pilot projects and research.
Developing the foundation for Responsible AI (RAI) and long-lasting organizational change
While the previous sections represent important first steps in understanding how to leverage AI and analytics, the biggest enabler and challenge lies in changing how the organization works. As states begin to experiment with AI and scale up capabilities, it is critical to embed Responsible AI (RAI) principles throughout the organization to prevent key watchouts (e.g., overconfidence based on imperfect information provided by GenAI models, legal issues from unauthorized use of data)
Responsible AI is the process of developing and operating artificial intelligence systems that align with organizational purpose and ethical values. By implementing RAI strategically, states can resolve complex ethical questions around AI deployments and investments, accelerate innovation, and realize increased value from AI itself. Critically, RAI strategy spans beyond just data, technology and tools – it requires a set of principles, supported by explicit governance structures, processes and changes to culture.
What next?
By implementing digital and AI in infrastructure, states can propel these systems forward by predicting changes and implementing solutions through data-driven analysis. To enable a digital and AI-driven ecosystem, states should start by:
Be on the lookout for the next article in our series, Bridging the Gap: Propelling America's Public Infrastructure Forward, which will cover best practices in infrastructure operations and maintenance.
Get in Touch
Heidi Kim (heidi.kim@bcgdv.com)
Manik Jain (jain.manik@bcg.com)
Patrick Roche (roche.patrick@bcg.com)
Amir Ganaba (ganaba.amir@bcg.com)
Sheila Ho (ho.sheila@bcg.com)
Alex Vickers (vickers.alex@bcg.com)
It's all about digging out insights for decision making from the vast (and increasing) amounts of data we are collecting.