Helping Qatar Become the Best Place in the World to Build

When we began working with the State of Qatar on how they intake, review, and approve building permits, their ambition was clear: become the best place in the world to build. That meant getting from a building proposal to an approved plan quickly, with confidence that every plan met national and local codes and regulations. At that time, Qatar's permitting was taking about 40 days on average, which put Qatar ahead of most countries, but that wasn't good enough to match the country's ambition.

Anyone who has applied for a permit is familiar with where the delays come from. An architect or a contractor submits a plan, a reviewer finds a problem, and the application goes back for corrections. The next review finds something else, and it gets sent back again. Meanwhile, the land sits unused, construction waits, and financing costs accumulate.

Qatar asked Brain Co. to shorten the permitting process while keeping reviews accurate. We sent our engineering team to work on site with the permit reviewers at the Ministry of Municipality and across Qatar’s municipalities, learning how they assessed applications and testing the system with them. As a team, we keep coming back to what this project taught us. Partly because of the engineering rigor it required, but mostly because of what it took to earn the reviewers’ trust.

A high bar and reasons to doubt

The reviewers were skeptical from day one, and honestly, they were right to be. They had spent years learning to read building plans and apply the codes correctly. Their work required judging how rooms, floors, and plot boundaries related to one another, then accounting for exceptions that only come with experience. Approving a plan that does not meet standards has very real legal, economics, and human safety consequences.

Qatar had also tried this before. Another Silicon Valley AI vendor had attempted the work and failed to deliver. We were walking into a project with people who had already heard the AI pitch once and watched it fail. They needed evidence that we could make it work, and we needed to deliver it quickly.

We knew right away the assignment would demand more than getting a model to read a blueprint. We had to hit reviewer-grade accuracy, learn how each of the eight municipalities apply their rules in practice, and build a solution that applicants and reviewers would trust deeply to do their jobs. We took on all of those challenges and dealt with them one by one.

Learning the work on site

We started with one municipality, studying local blueprints and sitting with reviewers to learn what they checked, how they reached decisions, and where applications got stuck. Our team went on to visit all eight municipalities. Sometimes that meant driving more than three hours to talk to one reviewer. If that person could explain a rule or an edge case we hadn’t accounted for, the trip was worth making.

We kept the contact direct and conversations private. We had a WhatsApp group with every reviewer in Qatar, so they could send us questions and problems live as they reviewed applications. We asked for feedback constantly, made revisions, and told them exactly what we had fixed. Those conversations helped us work through differences between the written guidelines and how rules were applied in practice.

They also informed our priorities. Reviewers told us that about half their work involved modifications to existing buildings, not just new construction. Those applications needed different treatment, because an existing room could fall outside current requirements without being part of the proposed renovation. We moved modification applications up the product roadmap because the product must account for the work on reviewers’ desks.

Alongside that work, our engineers trained models to identify objects such as buildings, entrances, and windows, on plans more accurately. Reported failures became test cases, so we could check each revision against problems we had already seen. We went through tens of iterations, with the engineering and feedback from users informing each other throughout development.

Two sides of the same application

During the discovery process, it became clear to us that efficiency depends on both ends of the application process. Reviewers needed tools to assess applications and make decisions faster, while applicants needed clearer guidance to submit compliant plans to avoid unnecessary back-and-forth. Although both groups worked from the same plans, they needed different experiences: an applicant wants to know what to fix before submitting, while a reviewer needs to inspect the application, assess the evidence, and decide whether it meets the requirements. So we built two connected portals around those jobs.

The applicant portal first checks whether the blueprint, application form, and municipality’s land records agree with each other. If the plot dimensions differ from public records or a building’s height doesn’t match a label, it flags the discrepancy. Then it checks the plan against building requirements, showing the applicant where each violation is, which rule applies, and what needs correction. This gives them a chance to address the problems before submission.

Reviewers work through plans page by page, so their portal follows that process. They see violations that need attention and explanations from applicants requesting an exception first, since that’s where reviewers spend most of their time. They can inspect the evidence behind a result and decide how to proceed. Checks that passed are also available for additional reviews. Section heads have a separate view for final sign-off.

As the two portals became part of the daily workflow, we kept working with applicants and reviewers to make the system more accurate. Each round of feedback gave us another case to test and another detail to get it right.

Finding what three reviewers missed

Reviewers could see the system improving over time, but they still needed to know how much they could trust the results. One case gave them a reason to look at the system differently.

We ran an application through the system that had already passed three reviewers and received a permit. The AI found issues all three of them had missed.

The reviewers and senior stakeholders were in disbelief when we showed them the results. The reviewers went back through the plans and checked every fact the system cited. As they verified the issues, they began to see how well it had pieced together information across the application to catch violations they had missed. We’d shared evaluation results before, but seeing the system catch errors in an application they had approved made those results real.

We saw people move from skepticism to using the product in their work, then to advocating for others to use it. That progression meant a lot to the Brain Co. team. Reviewers who had questioned whether AI could do the job were now helping us take it to other municipalities.

Applicants saw the benefit too. “I check the AI as the basis for my review,” said a consultant who uses the applicant portal. Another consultant reported: “My application rejection rate dropped by more than 40% after using this portal.” His experience spoke to a broader need we heard from architectural firms: catch errors before submission and spend less time sending plans back and forth. Some firms even asked for the portal to be made mandatory by the state so their teams could use it every time.

A national rollout, customized for each municipality

We deployed the system across all of Qatar’s municipalities at once, then visited each one to gather feedback and adapt the product to its local rules and review process. The system is now live nationwide for single-family housing applications. One of Qatar’s main goals was to help reviewers process more applications, faster. We compared the AI’s decisions with the reviewers’ decisions to find the cases it handles reliably, and we began automating those. Automation is the reason the accuracy bar is so high: reviewers have to be able to trust the system with routine cases so they can spend more of their time on the complex applications that need their judgment. Our team reports that the AI’s review accuracy is 97%, which is above the human average.

The permitting process can now take under two hours, compared to more than 40 days previously. We estimate this tremendous time saving contributes approximately $375 million a year to Qatar's economy, from less money spent holding land idle and an earlier start to construction.

We’re now extending the system to more residential building types, including apartments and compounds, with commercial reviews planned after that. Each expansion builds on the AI agents we’ve trained to read plans, flag code violations, and learn from errors, as well as the relationships we’ve built across the municipalities.

Build until people can depend on it

What stays with me about Qatar's project is how much of the work happened beyond engineering itself. Yes, we needed engineers who could solve hard technical problems, but we also needed those who were willing to travel literally thousands of miles to meet a reviewer, work through a disputed rule, or reorder the product roadmap when users needed something sooner.

At Brain Co., we believe in building products that customers find useful and come to depend on. That belief guides where we spend our time, what we build next, and how we respond when something isn’t working. In Qatar, we paired our engineering expertise with the grit to work through difficult cases and the willingness to let users’ needs set our priorities. We stayed close to the reviewers and applicants and took responsibility for getting the system to work for them. That’s what sets us apart from anyone else.

We’re now expanding our product for multi-family homes and commercial infrastructure to support all construction across Qatar. The goal is the same one we started with: help make Qatar the best place in the world to build. We’re also bringing what we’ve learned in permitting to the U.S., so more people can spend less time waiting and get on with building.