Reducing risk for real projects in the real world.

Sound Transit

The largest transit expansion in the US uses TwinKnowledge to open safer stations faster — increasing safety-review efficiency by 13×.

Problem: too many reviews, too little staff

Sound Transit, the metro Seattle public transit authority, is overseeing the nation's most ambitious transit expansion — over $100B and 120 miles over the next three decades. As-built conditions need approval before going into operation, and Sound Transit's five-person Design Technology team provides a safety confidence score to the Chief Engineer for critical "go / no-go" decisions on opening each section for public use. Reviewing nearly 4,500 sheets of as-built drawings biweekly, the team turned to TwinKnowledge to drastically increase review efficiency — and the Chief Engineer's confidence.

Opportunity: automate everything but the critical "complies or doesn't" decision

Working closely with the Design Technology team, TwinKnowledge identified several bottlenecks in their workflow where AI could dramatically increase efficiency:

• For each review item, information had to be found within — and mapped between — three different sources: drawings, PDF documents, and spreadsheets, some thousands of pages long.
• Those documents were scattered across the project's SharePoint — Sound Transit's hub, contractors' hubs, and subcontractors' hubs — and had to be located and retrieved before any review could start.
• Each contractor and subcontractor recorded information differently, making it hard to standardize the review.

TwinKnowledge saw an opportunity to automate the work behind these three issues and remove it from the team's manual process entirely: scouring SharePoint for the right files, finding and mapping relevant information, and standardizing review across inconsistent records.

Solution & value: pre-packaged review sets, 1,300% more efficient, errors caught

Deploying TwinKnowledge's AI-augmented QA/QC workflow, Requirement Validation, Sound Transit increased average review speed from one item every 26 minutes to one item every 2 minutes — a 1,300% increase. By integrating directly with SharePoint and using AI to find, index, and standardize information across folders, documents, and stakeholders, Requirement Validation delivers the team pre-packaged review sets for each item. Freed from first finding and collecting all the necessary information, reviewers focus on what matters — safety compliance — leading to safer stations that open sooner.

Get in touch about bringing TK Assistants onto your project

SHoP Architects

One of the most innovative architecture practices in the world looks to TwinKnowledge to help them innovate.

Problem: finding needles in haystacks

Named Fast Company's Most Innovative Architecture Firm in the World, SHoP Architects know a thing or two about good innovation. When TwinKnowledge presented its AI Assistants to SHoP's senior executives, they quickly recognized the potential to reduce the 75% of time project managers spend searching for answers to RFIs during construction administration (CA).

Opportunity: SHoP's strengths meet TK's AI Assistants

Working closely with SHoP's CA team on the Fashion Institute of Technology (FIT) expansion, TwinKnowledge found that most of their RFI-response pain came from three cases:

• Extracting exact answers — numbers, specifications, detail information — from within thousands of pages.
• Needing answers that spanned both the specifications and the drawing set.
• Uncovering the story of how a design component changed throughout the project.

Having joined the project mid-CA, these problems were amplified for the FIT team. TwinKnowledge saw an opportunity to pair the construction-information understanding of its TK Assistants with the design judgment of SHoP's CA team — automating the information search and retrieval that was costing SHoP so much time.

Solution & value: answers in seconds, 3× the quality control

Using its proprietary AI technology, TwinKnowledge unified the entire history of FIT's project information — drawing sets, specifications, and change documents — into a single knowledge base, without copying or moving the originals. With TK Assistants staffed on the ongoing project, SHoP's architects get answers in seconds where it once took hours. Preliminary results show RFI-response time falling from 75% to 25% of total CA time, letting architects spend more time on proactive on-site QA/QC — a win-win for SHoP's CA team and the FIT project at large.

Get in touch about bringing TK Assistants onto your project

National Homebuilder

One of America's largest homebuilders and most admired companies turns to TwinKnowledge to standardize QA/QC with AI.

Problem: standardizing projects across contractors

One of the nation's largest homebuilders has been on Fortune's World's Most Admired Companies list ten years running — for good reason. With many housing projects underway across the country at once, they wanted to improve how they carry learnings forward and standardize designs from project to project, creating operational efficiency and QA/QC practices at scale.

Opportunity: setting the table for contract-document alignment

In collaboration with their Director of Quality Assurance, TwinKnowledge was tasked with using TK's AI Agents to identify discrepancies in contract documents during the initial project phases — an area the homebuilder identified as the source of 75% of their project issues. Through weeks of studying their review workflows, TK found the issue stemmed not only from the difficulty of matching corresponding information across two dependent contract documents (e.g., the Scope of Work and the appropriate drawing details), but also from the fact that their sources of truth for bidding documents were scattered across two departments, multiple documents, and multiple systems of record.

TK saw an opportunity to unify all contract-document information within a single knowledge base, so corresponding information could be compared side-by-side to ensure alignment across every document. By expanding document-alignment capability at the very start of projects, the homebuilder could eliminate up to hundreds of downstream RFIs caused by design misalignment.

Solution & value: custom queries for instant side-by-side review

With its proprietary AI technology as the foundation, TwinKnowledge is building a user experience tailored to construction-document review that lets this homebuilder instantly compare any design side-by-side across scope information, project specifications, and drawing submittals.

Check back in the coming weeks for the numbers behind how TK is transforming operational efficiency for one of America's largest and most respected homebuilders.

Get in touch about bringing TK Agents onto your project

AWS: AI-Powered Construction Document Analysis

AWS features TwinKnowledge in its Physical AI blog for pairing computer vision and large language models to review construction drawings at scale.

A pioneering approach to construction-document AI

In its Physical AI blog, AWS highlights how TwinKnowledge combines computer vision and large language models to analyze construction drawings and specifications — surfacing the information trapped across sprawling document sets and catching the misalignments that drive costly downstream rework.

"TwinKnowledge is a pioneering AEC AI company that's reimagining how the industry designs and reviews construction drawings at scale."

Scaling industry-transforming AI on AWS

The collaboration shows how a purpose-built document-processing pipeline — running on AWS infrastructure — can transform how AEC teams work with drawings, specs, and submittals.

"…can transform document processing workflows in the AEC industry and help scale industry-transforming AI."

Read the full story on the AWS Physical AI blog