AI Review
AI Review automatically checks 2D technical drawings and 3D CAD models against internal design rules and ISO GPS or ASME Y14 standards, annotating deviations in the file and issuing a report. Built for engineering teams in manufacturing.
What is AI Review?
AI Review is a software platform that automates the review of 2D technical drawings and 3D CAD models. It is published by T&S Engineering Intelligence AB, a Swedish company based in Gävle, and presented on a bilingual Swedish and English website. The product grew out of IndX, the open innovation initiative run by Linköping Science Park, which the three founders joined in 2020; the first pilots ran with Väderstad and Siemens Energy. The team comes from software, mathematics, engineering and design, and the publisher states that the platform now handles thousands of engineering drawings every week for major manufacturers across Europe. Technically, the tool combines machine learning, deep neural networks and computer vision. The models are strictly predictive, never generative: they identify objects, symbols, dimensions and design information in a drawing, but cannot create or alter anything in it. What they read is then compared against the applicable standards, ISO GPS, ISO 1101:2017 and ASME Y14.5, and against the company's own internal design rules. The output is twofold: deviations annotated directly in the drawing or model, and a detailed report explaining each finding. Rules do not stay locked with the publisher. Through AI Review Studio, a customer trained on the module creates and edits its own review rules, which is how Alfa Laval and LiU Formula Student use it. Deployment is flexible. The platform runs in the cloud, with hosting providers in Sweden and Europe, or on-premise inside the customer's own IT infrastructure. It integrates with existing PLM and business systems or runs as a stand-alone web application, and is reached through a REST API on a customer-specific entry point plus browser-based administration interfaces. Native import covers PDF and STEP, export is PDF, and other formats go through third-party converters. The publisher positions the tool as a complement to the checks built into CAD software, aimed at advanced criteria and company-specific rules rather than as a replacement. Named references include Siemens Energy (Turbine department, more than 40 engineers, over 30 attributes checked automatically), Väderstad AB since 2021, Alfa Laval and LiU Formula Student, alongside two anonymised cases in food production and manufacturing.
What it does
- Automatically detect deviations in 2D technical drawings and 3D CAD models
- Check geometric symbols and tolerances (GD&T) against ISO GPS and ASME Y14
- Verify compliance with the company's own internal design rules
- Annotate each deviation directly in the drawing or the model
- Produce an explanatory report of the non-conformities found
- Create and edit your own review rules in AI Review Studio
- Plug the review step into a PLM or ERP system, or run it as a stand-alone web application
When to use AI Review / When not to
A quick filter to help you decide if AI Review is the right fit.
When to use AI Review
- Design offices handling high drawing volumes, where manual review cannot keep up: Alfa Laval targets 10,000 drawings a week through the tool.
- Manufacturers working to ISO GPS, ISO 1101:2017 or ASME Y14.5, or to a proprietary internal standard such as the RPS rule set used by one anonymised customer.
- Organisations whose final drawing approval depends on a handful of senior reviewers and has become a bottleneck: at LiU Formula Student one reviewer served 77 designers.
- Engineering teams that need an on-premise deployment because their drawings are confidential, or that want the review step wired into an existing PLM or ERP workflow.
- Teams training junior designers, since the report explains the cause of each deviation, a teaching benefit highlighted by both LiU Formula Student and Väderstad.
When not to use AI Review
- Anyone looking for a generative assistant: the models are strictly predictive computer vision and neither create nor modify anything in a drawing.
- Engineers expecting the tool to replace human judgement. Design intent, manufacturability and critical decisions stay human, and the publisher positions AI Review as a complement to the checks built into CAD software, not a replacement for them.
- Teams that want a self-service product: there is no free plan, no online trial and no published price, and every engagement starts with a sales demo.
- Projects that need an immediate go-live, since each deployment requires the AI reviewer to be trained on the customer's own drawings first.
- Users working outside PDF and STEP for import, or expecting a mobile app or a browser extension. Neither exists, and other formats require third-party converters.
How to use AI Review
A typical end-to-end flow, from setup to results.
- Request a demo through the Book a demo form on the home page, or write to sales@tsei.se; the walkthrough of the system is presented as free.
- Discuss deployment with the publisher: cloud hosting with professional providers in Sweden and Europe, or on-premise inside your own IT infrastructure.
- Provide a representative set of your drawings so the AI reviewer can be trained, since every instance is tailored to the customer.
- Agree which standards and internal design rules the reviewer must enforce (ISO GPS, ISO 1101:2017, ASME Y14.5, company rule sets).
- Connect the platform: REST API on your dedicated entry point, integration with your PLM or ERP, or stand-alone use as a web application.
- Submit a drawing that is ready for review, either manually from the browser or automatically from your workflow.
- Read the output: deviations annotated in the drawing or model, plus a report explaining the cause of each one.
- Correct the flagged items and resubmit, using the review as the step before a drawing goes to the final human reviewer, as LiU Formula Student does.
- Automate the loop at scale where it pays: at Alfa Laval, every automatically generated drawing is sent straight to AI Review.
- Take the AI Review Studio training, then write and edit your own review rules; training material, example exercises and explanatory texts are provided.
Pros & Cons
Pros
- Measured gains reported by users: LiU Formula Student credits AI Review with catching around 40 percentage points more errors and saving roughly 33 hours of review per 1,000 drawings.
- Proven at industrial scale: at Siemens Energy more than 30 attributes and criteria are checked automatically for over 40 engineers in the Turbine department, and Väderstad reports fewer review iterations.
- Teaching value, because the report explains the cause of each deviation; LiU Formula Student describes the tool as not only a reviewer but also a teacher.
- Rules stay with the customer: through AI Review Studio a trained team writes and edits its own checks without going back to the publisher.
- A real choice between cloud and on-premise deployment, which suits confidential drawings, backed by hosting in Sweden and Europe, a European publisher and ISO/IEC 27001 certification.
- The models are predictive only and cannot modify the customer's drawings.
- Published service commitments (99.3% availability target, 48-hour RTO) and named reference customers with quotes attributed to identified people, which is uncommon at this level of maturity.
Cons
- No public pricing and no pricing page at all, so budgeting is impossible without going through the sales process.
- No free plan and no online trial; the only free item is the commercial demo.
- Minimal legal documentation: a privacy policy dated June 2023 with no mention of the GDPR, no retention period, no sub-processor list and no data processing agreement.
- No support, legal or data protection email address is published. The only contact is sales@tsei.se, which sits on the publisher's domain rather than on aireview.se.
- No public API documentation: the REST API sits behind a customer-specific entry point, so integration effort cannot be assessed before contact.
- Getting started is not instant, since the reviewer must first be trained on the customer's drawings, and native formats are limited to PDF and STEP for import and PDF for export, with no mobile app and no browser extension.
- Many FAQ answers redirect to a sales conversation rather than documenting anything (PLM integrations, cloud versus on-premise, differences from CAD built-in checkers), and the publisher is a small structure, with three employees declared in 2024 according to third-party registers, serving major industrial customers.
Pricing & Plans
No pricing is published anywhere on the AI Review website. A full enumeration of both page sitemaps, six Swedish pages and two English pages, returns no pricing page, no plan, no starting price, no currency and no billing unit. The commercial model is contact-sales: the prospect books a demo, presented as a free walkthrough of the system, and a quotation follows from that conversation. There is no free plan and no self-service trial, and the demo is the only free item advertised. Commercial enquiries go to sales@tsei.se, an address on the publisher's own domain.
- AI Review does not expose a plan grid of any kind
- and access to any commercial terms goes through sales@tsei.se or the Book a demo form.
- cloud with professional hosting providers in Sweden and Europe
- or on-premise inside the customer's IT infrastructure.
- AI Review Studio is presented as a module that customers use
- with Alfa Laval teams trained on it and LiU Formula Student using it
- and no indication that it is billed separately.
Data, GDPR & hosting
A consolidated view of how AI Review handles your data.
GDPR overview
AI Review's published documentation never mentions the GDPR. A full-text search across the site (privacy policy, legal index, acceptable use policy, service specification) returns no reference to the regulation, and none to the Swedish term dataskydd outside the cookie banner configuration. No legal basis, no rights of access, erasure or portability, no data protection officer and no Article 27 representative are stated anywhere. No data processing agreement is published or announced, no sub-processors are listed, and no retention period is given. The policy is dated 16 June 2023, posted and effective the same day. Several factual elements nevertheless point the right way: the publisher is established in Sweden, therefore inside the EU; the service is hosted in Sweden and Europe or on-premise; ISO/IEC 27001 certification is claimed. The publisher invokes the EU AI Act, not the GDPR.
Who owns the data?
AI Review's homepage FAQ states that the customer retains full ownership of its data. Each AI reviewer is custom-built and trained on that customer's own drawings, and the publisher states that models trained on one customer's data are not shared with other clients. T&S Engineering Intelligence claims ISO/IEC 27001 certification, verifiable through iafcertsearch.org. These commitments appear only as FAQ answers: the privacy policy, effective 16 June 2023, covers nothing but personal account data (name, email, password, company, location, billing details and server logs) and is entirely silent on the technical drawings themselves, on who may access them and for how long.
Reuse rights
The privacy policy of 16 June 2023 distinguishes two streams. Account data, namely name, email address, password, company and location, is collected at registration; billing data covers the billing address and payment card, and T&S Engineering Intelligence keeps records of billings and transactions. The servers also keep log files of access to and usage of the Services. The stated purposes are billing, notifications, support and improvement of the service, communication about the account, and legal obligations; cookies serve authentication, plan selection and anonymous statistics. Technical drawings follow a separate logic: they are used to train an AI reviewer dedicated to that customer, which the publisher says is not shared with other clients. No published clause grants or restricts the customer's own reuse of the review output, and none defines what the customer may do with it, so that has to be settled contractually. The publisher classifies its models as minimal risk under the EU AI Act (Regulation 2024/1689).
Data retention & training
Hosting summary
The service specification states that AI Review is hosted with professional hosting providers in Sweden and Europe, or on-premise at the customer's own IT infrastructure. The choice is discussed with the publisher and presented as a customer decision rather than a fixed architecture. Availability is targeted at 99.3%, aligned with the hosting providers' own commitments; deployments are described as zero-downtime, there is no recurring maintenance window, and the recovery time objective after a disaster is 48 hours. Strict security controls are claimed whether the platform is deployed in the cloud or on-premise, under an ISO/IEC 27001 certification the publisher says can be verified on iafcertsearch.org. One distinction matters. Those statements describe the production service. The public website is a separate matter: aireview.se resolves to 46.16.236.16, operated by Oderland Webbhotell AB in Gothenburg, Sweden, which is the WordPress showcase site and not the platform where drawings are processed. No jurisdiction clause, no named data centre and no sub-processor list is published, so the exact hosting arrangement has to be confirmed with the publisher.
Things to keep in mind
Risks and trade-offs to weigh before adopting AI Review.
- You cannot budget before engaging commercially: no price is published anywhere, so any comparison with alternatives happens only after a sales conversation.
- The privacy policy dates from June 2023 and is silent on the GDPR, on retention periods and on sub-processors, and no data processing agreement is published or announced. All of that has to be secured contractually.
- Your drawings train the model. The publisher says the resulting reviewer is not shared with other clients, but the exact scope of that training, and whether it can be reversed or the model deleted, is not documented anywhere; clarify it in the contract.
- No support, legal or data protection contact is published. The only address is commercial, sales@tsei.se, and it sits on the publisher's domain, tsei.se, not on aireview.se.
- The FAQ answers on PLM and ERP integrations are statements of feasibility that end in an invitation to get in touch, not documented connectors. Treat Siemens NX, Teamcenter and SAP as claimed capabilities until proven in your own environment.
- The publisher is a small structure, with three employees declared in 2024 according to third-party registers, for a tool placed at a critical point in an engineering workflow. Check business continuity before a critical deployment.
- The tool does not replace engineering judgement: design intent, manufacturability and critical decisions stay human, and a team that stops reading drawings closely because the AI checked them loses more than it gains. Note also that the aireview.se domain dates from March 2023 while the company has existed since 2019, so the domain age is not the company's age.
Setup & Integrations
Technical difficulty
Moderate to high, and never self-service. The path starts with a sales demo, then the AI reviewer is trained on the customer's own drawings: one anonymised manufacturer is described as still having the tool tailored to its drawings before launch. Stand-alone use as a web application is the lightest option; connecting the REST API on a customer-specific entry point, or wiring the review into a PLM/ERP or drawing generation system as Alfa Laval did, needs a technical team. AI Review Studio requires training, which one user found straightforward. No public API documentation exists to size the effort beforehand.
Deployment
Integrations
Behind AI Review
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What does AI Review actually do?
Which standards does it check against?
What kind of AI is behind it?
Can it run on our own servers?
Who owns the data, and is our model shared with anyone?
Does it replace the checks built into our CAD software?
Which file formats are supported?
What service level is committed?
Is there a free trial or a published price?
Can we write our own review rules?
Should you pick AI Review?
AI Review is a deliberately vertical product: it reviews technical drawings and CAD models in manufacturing, and pretends to do nothing else. That focus is its main strength. The models are predictive rather than generative, so they read a drawing, compare it against ISO GPS, ISO 1101:2017, ASME Y14.5 and the company's own rules, annotate the deviations and explain them, but never rewrite the file. Maturity is easier to judge here than with most tools, because the publisher names its references and attributes its figures. Siemens Energy, a partner since 2020, has over 30 attributes checked automatically for more than 40 engineers; Väderstad has been a customer since 2021 and reports fewer review iterations; Alfa Laval is aiming at 10,000 drawings a week; LiU Formula Student measured roughly 33 hours saved per 1,000 drawings. Two differentiators stand out: AI Review Studio, which lets the customer write and edit its own rules without going back through the publisher, and a genuine on-premise option, which matters when the drawings themselves are the intellectual property at stake. A European publisher, hosting in Sweden and Europe and ISO/IEC 27001 certification reinforce that argument. The weak point is everything around the product rather than the product itself. There is no published price, no trial, no free plan, no API documentation and no support or legal contact address; the privacy policy dates from June 2023 and says nothing about the GDPR, retention or sub-processors. All of that has to be settled contractually, after a sales conversation. This is a tool for an organisation with a real volume of drawings and a standard to enforce, not for an individual engineer. If that describes you, the demo is worth the call; if you need to compare prices before speaking to anyone, you cannot.
- Choosing a selection results in a full page refresh.
- Opens in a new window.