AI won’t fix construction delays until jobsites fix data

AI in construction remains a promise until jobsite data is structured and centralized

Top line: Artificial intelligence will not solve construction delays or budget overruns without structured, centralized data. The industry must first adopt digital workflows and standardized data capture so AI can turn jobsite activity into reliable, actionable insight.

Why this matters now

Most construction sites still operate with a patchwork of tools: updates shared via messaging apps, paper checklists and cloud folders. Those tools and folders are often days out of date. With information fragmented and delayed, AI feels disconnected from the realities of daily work. Until data is current, complete and centralized, AI remains a theoretical promise rather than a practical asset.

Clear, measurable problem

20% of projects run late. 80% of projects go over budget. Those outcomes are directly linked to data gaps: missing, inconsistent or stale inputs prevent planning engines and predictive models from offering timely, accurate answers.

How other industries made AI useful

A decade ago agriculture was seasonal, equipment‑heavy and driven by gut instinct. Farmers began sampling soil, logging yields, and flying drones. Once structured data points accumulated in agriculture, AI models informed growers when to irrigate or harvest. Structured inputs turned raw field data into precise, daily guidance for farmers. Construction can follow the same path as agriculture with structured data.

Prescriptions that move AI from theory to practice

The key to unlocking AI’s potential in construction lies in utilizing a single platform that standardizes processes, empowers team collaboration, and captures and organizes data in one place. Before AI can assist in construction, the right digital tool must be in place. Before AI can elevate jobsite performance, teams must first bring their workflows together in one digital environment.

Practical guidance:

• Work from the same plans, in the same place — choose the right software, align on standardized workflows, and ensure the whole team is trained and confident using it.
• Connect site and office teams using shared systems that sync in real time so everyone is working with the latest updates in a single project space.
• Centralize project data in the cloud to ensure seamless access, live collaboration and easy analysis regardless of team location.
• Start small — identify one workflow, digitize and structure it, then measure the gain before scaling.

How data must be captured

How data is captured matters as much as where it’s stored; data must be structured from the start for AI to use it. AI can only work with what it understands; site activity needs to be documented in consistent, clear and connected ways.

Examples of structured capture:

• Photos and videos pinned to digital floor plans, automatically tagged with timestamps and location for accurate context.
• Predefined digital forms and templates that guide on‑site teams to consistently collect and report the right information.
• Centralized document management with workflows, approvals and records stored in one place for easy access and tracking.

Not all data is created equal. Photos and videos captured at the right moment, from the right place, with the right context are among the most valuable records. When visual records are tied to digital floor plans, timestamped and tagged by location and trade, they become the basis for smart decision‑making rather than just records. Forward‑thinking teams are integrating visual capture into daily routines to build consistent, project‑wide visual histories.

Product examples show the mechanics: PlanRadar’s SiteView feature uses a helmet‑mounted camera to record site walks in 360° and automatically map imagery onto floor plans via an AI‑powered platform. That process creates a complete, structured visual log of how a site evolved over time. Structured visual datasets will fuel AI tools that detect deviations, spot patterns and forecast delays before they become costly. Structured visual documentation moves projects from reactive fixes to predictive, data‑driven delivery.

Real projects where structured data changed outcomes

Zachry Construction adopted ALICE Technologies’ generative‑scheduling engine. Years before rollout, Zachry standardized how takeoffs, crew rates and weather delays were logged across projects. Because inputs share one format, ALICE can ingest them and generate hundreds of buildable sequences in minutes. What once required a full‑day workshop can now be a live what‑if session where planners adjust a resource or constraint and the schedule updates in real time. Early trials surfaced sequences that trimmed the critical path by up to two weeks and highlighted risks before ground was broken. Without consistent quantities, rates and calendars, even the best AI tools are ineffective; with them, AI becomes a fast, tireless, fact‑driven planning partner.

For years Skanska USA logged safety observations, toolbox talks and near‑miss reports into siloed PDFs and spreadsheets. Those valuable records rarely surfaced beyond monthly meetings. Skanska created a central safety warehouse and standardized hazard categories, severity ratings and jobsite metadata for every record. With a clean data model, analysts produced reliable dashboards and trend maps rather than static summaries. Skanska launched Safety Sidekick, a generative AI chatbot trained solely on the firm’s incident history and best‑practice library. Field supervisors can pose plain‑language questions and receive instant answers with inspection excerpts and toolbox‑talk scripts. Skanska reports a 40% reduction in time spent preparing toolbox talks and increased day‑to‑day engagement with site safety protocols nationwide.

A global data‑center developer consolidated schedule exports from 15 concurrent builds into a single activity taxonomy inside Foresight. Weekly P6 APIs replaced annual uploads, consolidating scattered files into a single clean dataset. Foresight’s AI forecast engine delivered probabilistic finish dates and early‑warning delay signals. Executives switched to monthly meetings, and reporting costs dropped 90% from $2 million to $200,000. Proactive resequencing avoided roughly $100 million in delay costs. Structured data enabled the results; AI made them visible.

Startups and steady steps

Getting started with AI doesn’t require overhauling an entire tech stack; it requires structuring existing work to make it searchable, analyzable and automatable. Identify one workflow that frustrates the team, digitize and structure it, proving AI’s potential incrementally. Take one process. Digitize it. Structure it. That’s how to prepare for AI.

Regional market signals

Micron Technology filed a detailed site plan with the town of Clay, New York for a planned chipmaking complex. The site‑plan filing is 336 pages. The filing offers the clearest look yet at the sprawling complex Micron intends to build in the upstate New York town. The filing starts a town review Micron will need to start construction of the fabrication plants. Town Supervisor Damian Ulatowski called the filing a critical step and said the town gets to see it. Micron had previously provided a site layout, but the site plan application provides the clearest picture of the complex. The application contains dozens of images showing building appearances from the road, proposed trees and shrubs, sign size and other visual aspects. Submission of the site plan will launch a town review to ensure the plan conforms to building and zoning codes. The town planning board will review and vote on the site plan. The town board will vote on rezoning the last three parcels acquired for the project last year. The site plan submission is the latest in a series of major advances for the project in recent months. In June, the Onondaga County Industrial Development Agency released a 20,000‑page environmental impact report on the Micron project. Micron is in the final stages of obtaining the first three permits it will need to start clearing the land in November. Micron has hired Gilbane Co. to start the estimated $1 billion job of clearing and leveling the site. The entire 1,400‑acre site is owned by OCIDA, which will eventually sell the land to Micron. The site plan application focuses heavily on the first fab. The first fab will be built at the northwest corner of the site along Caughdenoy Road, just south of the electrical transmission lines that define the northern boundary. Each fab will cover 28 acres of land. Each fab will stand about as high as Syracuse University’s JMA Wireless Dome. Micron says it plans to spend up to $100 billion over the next 20 years building four fabs in succession. The project is backed by up to $25 billion in taxpayer subsidies. The project will help Micron bring about 40% of its chip production back to the United States. Nearly all of Micron’s chips are currently produced in factories in Asia. Clay hired a new planning commissioner, Brian Bender, who started the job on a Tuesday. The town will likely have to hire four or five code enforcement officers initially to assist Bender with inspecting construction once it starts. The town signed a reimbursement agreement with Micron in July to cover Clay’s expenses in reviewing the project. The site‑plan application was reviewed earlier in the month and moved forward. Micron will not need site‑plan approval to begin cutting trees and leveling the first 700 acres of the site this November. The site plan must be approved before any buildings are constructed. Construction on the first fab is scheduled to begin in late 2025. If all four fabs are built, Micron’s campus would be the largest private development in New York state history and one of the biggest private developments in the nation. Each fab would contain 600,000 square feet of cleanroom space, 290,000 square feet of cleanroom support space, and 250,000 square feet of administrative space. Each pair of fabs will be supported by 470,000 square feet of central utility buildings, 200,000 square feet of warehouse space, and 200,000 square feet of product testing space housed in separate buildings. The site plan includes ancillary on‑site electrical substation buildings, water and wastewater treatment and storage, and industrial gas storage.

Labor and recovery signals in a major U.S. city

New York City’s construction industry remains smaller than it was before the COVID‑19 pandemic. In 2024 the city averaged 143,100 construction jobs. That is an 11% decrease from 161,300 construction jobs in 2019. New York State remained 4% below its 2019 construction employment levels, making it the second‑slowest recovery of any state after West Virginia. The main factor cited in the sluggish hiring is lagging demand for non‑residential construction. Residential spending surpassed 2019 levels by 2023, reaching $22.8 billion. Nonresidential spending was $22.2 billion in 2023, 3% lower than in 2019. Nonresidential projects were projected to have fallen further in 2024 by another $572 million. Jobs in the remodeling sector recovered better in New York than any other subsector within the construction industry. The slow recovery disproportionately harmed immigrants, who held 61% of the city’s construction jobs as of 2023. Construction was the fifth highest‑paying sector in the city as of 2020 with an average salary of $95,800. The number of construction‑related employees at the NYC Department of Buildings dropped from 662 in March 2021 to 519 in March 2024. NYC was taking 70% longer to approve permits in June 2024 compared to June 2021.

Industry restructuring on the East Coast

FlatironDragados announced a spinoff of its New York and New Jersey subsidiaries to create a civil construction firm focused on the region. Three longtime area companies, Schiavone Construction, John P. Picone, and E.E. Cruz, will form a transportation and water infrastructure specialty firm called SPC Construction. SPC Construction will target work on the New York metro region’s tunnels, bridges, roadways and water systems. SPC Construction will operate as a subsidiary of FlatironDragados. SPC Construction will be led by Managing Director Jesus Diez. Joe Sheehan will assume the role of chief operating officer of SPC. Neil Wegman will serve as SPC’s CFO. Financial terms of the deal were not disclosed. The deal is targeted for completion by the end of the second quarter. The press release did not address redundancies or layoffs resulting from the combination.

Bottom line

Structured data is the backbone of any AI workflow. Project teams that replace ad hoc updates and siloed records with standardized capture, centralized storage and shared workflows will be the ones that convert AI from a promise into a daily tool for safer, faster and less costly delivery.