# AI to Read Optical Drawings for Quoting: ISO 10110 Prints and Lens Callouts > Source: https://netray.co/resources/ai-to-read-optical-drawings-for-quoting > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: Knowledge Base ## Short answer How AI extracts figure, scratch-dig, radius, centration and coating call-outs from optical drawings so estimators review a draft quote instead of retyping. ## Category ai-automation ## Content **Intro:** AI can read optical drawings, including ISO 10110-style prints and customer lens drawings, and extract substrate, radii, thickness, figure, scratch-dig, centration, clear aperture and coating call-outs into a structured list with page references. An estimator then confirms the extraction and reviews a draft quote built from similar past jobs. It does not replace judgment about whether a tolerance is achievable, but it removes the retyping and searching that consume the first hours of every RFQ. This guide explains what it does well and where it needs human review. **Sections:** #### What an Optical Print Contains **Body:** Optics prints mix graphics and tables. A lens drawing under ISO 10110 uses a table of surface data and a coded indication for material, stress birefringence, bubbles and inclusions, surface form, centering, surface imperfections, and coating. Many customer prints are less formal: a PDF with a title block, notes such as scratch-dig 40-20 per MIL-PRF-13830B, a coating specification reference and tolerances scattered across views. The estimator has to find all of these, interpret them and key them into a quote. **Bullets:** - Material and grade, including any index or homogeneity requirement - Radius, thickness, diameter, clear aperture and their tolerances - Figure, power and irregularity, with test wavelength - Scratch-dig, centration or wedge, bevel and edge notes - Coating spec and environmental or durability notes #### Extraction With Source Citations **Body:** A useful extraction tool shows where each value came from, so the estimator can check it in seconds. It should flag conflicts, such as a note that says 60-40 while the table says 20-10, and missing data such as an unspecified test wavelength. Scanned and hand-annotated prints are harder, so the tool should report low confidence rather than guess. Treat the output as a checklist the estimator signs off, not as a final answer. #### Matching Against Your Own History **Body:** Extraction becomes valuable when it connects to your ERP. With the call-outs structured, the system can find past jobs with the same material, similar size and the same tolerance class, then show actual hours by operation, yield, coating cost and quoted versus actual margin. The estimator sees what this kind of part really cost, not just a standard cost. That is the difference between a tool that saves typing and one that improves quote accuracy. #### Keeping Controlled Prints Private **Body:** Optics prints for defense, space and laser customers may be export-controlled or carry confidentiality terms. Sending them to a public AI service is a risk. Run extraction on open-weight models inside your own network, apply the same access rules as your document store and log what was processed. The customer remains responsible for its own export and security obligations, and the deployment should make those easier to meet, not harder. #### When It Is Not Worth It **Body:** If you receive only a few RFQs a week, or most are repeats of parts you already know, the setup effort may not pay back. The best fit is a shop with a steady flow of varied custom prints, a backlog of quotes waiting for an estimator and a decent amount of historical job data. Without usable history, the tool can still extract call-outs but the cost matching will be thin. **Cta:** Send Netray a few anonymized optical prints and past job records and we will show what extraction and history matching look like on your own data, on-prem. **Faq:** #### Can AI read ISO 10110 drawings? **Answer:** Modern document and vision-language models can read many ISO 10110 style prints, including tables and coded indications, but accuracy varies with print quality. The practical approach is extraction with source citations and confidence flags, followed by estimator confirmation. Scanned or hand-marked drawings need more review. #### Will AI decide if a tolerance is achievable? **Answer:** It should not. It can show how often you have held a similar tolerance, with yield and hours from history, but feasibility depends on your equipment, material lot and process capability. That judgment stays with the estimator and process engineer. #### Is it safe to put customer optical prints into an AI tool? **Answer:** Only if the tool runs within your controlled environment and respects the confidentiality and export terms attached to the print. Public AI services are generally not appropriate for controlled or NDA-covered drawings. An on-prem deployment with access controls and audit logs is the safer pattern. ## Related Slugs - ai-to-read-aerospace-drawings-for-quoting - how-to-quote-custom-optical-components - itar-compliant-ai-tools-manufacturing - machine-shop-quoting-software ## Search terms AI read optical drawings, ISO 10110 drawing extraction, AI quoting optics, automate optics RFQ, lens drawing OCR quoting, optical print review software --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json