AI in Construction: How GCC Contractors Automate Delivery

Most of what gets written about AI in construction is written by people who have never chased a payment certificate or sat through a Thursday progress meeting with an unhappy client representative. This guide takes a different angle. It looks at what AI automation actually does on a GCC project today, which tools are real, where the sales pitch outruns the reality, and how a contractor should adopt this technology without burning money on a pilot that dies after one project.

## Where the Hours Actually Go on a GCC Project

Before asking what AI can do, it is worth being honest about where delivery teams lose time. On a typical GCC contract, the pattern is familiar:

• Progress measurement: A planner walks the site, estimates percent complete by area, and updates the Primavera P6 programme by hand. The client's engineer often disagrees with those percentages, and the argument resurfaces at every payment application.
• Document control: Drawings, RFIs, transmittals, and material submittals flow between client, consultant, main contractor, and subcontractors. Every party keeps its own register, and the registers drift apart. Someone builds from a superseded revision, and the cost of that mistake lands on the contractor.
• Payment applications: The subcontractor submits a payment application, the main contractor's QS assesses it, the consultant certifies (or cuts) it, and the client pays weeks later. Each handover involves re-entering quantities into a different spreadsheet.
• Reporting: Weekly and monthly reports are assembled by copying data out of P6, Excel, and site photo folders into PowerPoint, often the night before the meeting.

None of this is engineering. It is data handling, and that is exactly the category of work automation is good at.

## What AI Automation Realistically Does Today

Strip away the marketing and today's practical applications on site fall into four buckets:

• Automated progress tracking: Site imagery captured on a fixed walking route is compared against the BIM model or programme, producing an objective record of what was actually built and when. The real value is not the dashboard; it is having evidence when progress is disputed at certification.
• Estimating and tender support: AI-assisted takeoff and document analysis speed up quantity extraction from drawings and help estimators find the scope gaps and contradictions buried in tender documents. The estimator still owns the number.
• Document control and correspondence: Automation can route submittals, flag overdue RFI responses, check that a transmittal references the current revision, and keep a single register that all parties draw from, which is where common data environments such as Oracle Aconex and Procore already dominate GCC projects.
• Reporting: The weekly report can be assembled by software from live data instead of by a document controller at midnight. This is the least glamorous application and usually the fastest payback.

## The Tools That Are Real

Four products come up constantly in this space, and all four are real, deployed systems rather than vaporware:

• Buildots: Helmet-mounted 360 cameras capture the site during routine walks; AI compares the footage against the model and schedule to report installation status trade by trade.
• OpenSpace: 360 photo capture mapped to floor plans, creating a navigable visual record of the site over time. Widely used for progress documentation and dispute evidence.
• Doxel: Uses imagery and lidar to quantify installed work against the model and budget, aimed at owners and contractors who want earned-value style tracking without manual measurement.
• Smartvid.io: Applied AI to site photos and video to flag safety risks such as missing PPE. It later rebranded as Newmetrix and was acquired by Oracle, a useful reminder that this market consolidates quickly and tool selection should account for vendor longevity.

Note what these tools have in common: they automate observation and comparison. None of them manages a project. The project director's job survives contact with all of them.

## One Concrete Workflow: The Weekly Progress Report

Consider a contractor running a mid-size building project who wants to stop burning two staff-days a week on reporting. A realistic automated workflow looks like this:

• Capture: A site engineer walks a fixed route twice a week wearing a 360 camera (the Buildots or OpenSpace pattern). No extra site labour; the walk was happening anyway.
• Processing: The capture platform maps imagery to floor plans and, where a model exists, updates installation status per zone and trade.
• Consolidation: An automation layer pulls that progress data, the current P6 milestone status, the RFI and submittal registers from the common data environment, and the week's HSE statistics into one structured dataset.
• Output: A dashboard for internal management, and an automatically generated PDF progress report in the client's required format, ready for the project manager to review, adjust the narrative, and issue.

The human still writes the commentary and owns the message to the client. What disappears is the copying, formatting, and chasing. This kind of glue-layer automation between existing systems is precisely the work covered in our business automation services, and it matters more than any single tool purchase.

## Where AI in Construction Is Oversold

Anyone who has delivered projects in the region should keep a healthy scepticism about the following claims:

• "AI will optimise your schedule automatically." Schedule logic reflects commercial strategy, procurement reality, and claim positioning. Software can flag logic errors and test scenarios; it cannot know that a sequence exists to protect an extension of time claim.
• "Full automation of document control." The failure mode of document control is rarely technical. It breaks because the consultant's team, the contractor, and subcontractors each operate under different contractual obligations and have different incentives about what gets formally recorded. A tool cannot fix a party that benefits from ambiguity.
• "AI-generated quantities you can bid on." Automated takeoff accelerates measurement, but on a lump-sum tender the estimator carries the risk, and no serious contractor signs a bid on unverified machine output.
• "Works out of the box." Progress-tracking AI needs a decent model, consistent capture discipline, and someone on site who owns the process. Projects without that discipline generate impressive dashboards of wrong data.

## A Pragmatic Adoption Path for a Contractor

A typical scenario: a GCC contractor with several live projects wants to modernise but has been burned by software that nobody used after month three. A sensible sequence:

• Start with reporting, not prediction. Automate the weekly report and the registers first. The payback is immediate, the risk is near zero, and it builds the data discipline everything else depends on.
• Pilot progress capture on one project. Pick a site with a motivated project manager, run it for a full quarter, and measure something specific, such as time to certify progress or hours spent on reporting, before and after.
• Fix the data plumbing. Most value is lost in the gaps between P6, the ERP, the document control platform, and Excel. Integration work is unglamorous and decisive.
• Only then consider the advanced layer. Predictive analytics and AI scheduling assistants are worth evaluating once clean, trusted data exists. Doing it in the reverse order is how pilots die.
• Assign ownership. Every tool that succeeded on our projects had a named owner on site. Every tool that failed was "everyone's responsibility."

Regional context matters here too. Contract structures, approval cultures, and client expectations in the Gulf differ from the US market these tools were built for, which is why we maintain a dedicated GCC practice focused on making automation work inside those realities.

## The Bottom Line

AI in construction is real, but it is not magic. Today it reliably automates observation, comparison, document flow, and reporting, the exact areas where GCC delivery teams bleed hours. It does not replace judgement about sequence, commercial strategy, or people. Contractors who start with the boring automations, prove value on one project, and build data discipline will compound an advantage. Contractors who buy dashboards first will fund another generation of shelfware.