.png)
Frameworks, core principles and top case studies for SaaS pricing, learnt and refined over 28+ years of SaaS-monetization experience.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
What Are the Pricing Models for AI Vision Screening in Optometry
AI vision screening is moving from a promising add-on to a real operating decision for optometry practices. The buyer is no longer asking only whether an algorithm can identify referable diabetic retinopathy or flag retinal risk. The harder question is commercial: what should a practice pay for, what should a vendor charge for, and how can both sides avoid a pricing model that discourages screening volume?
The stakes are practical. A one-location practice needs a predictable monthly cost and a workflow that works with one camera and a small technician team. A regional optometry group needs a contract that can expand across sites without creating invoice disputes. In both cases, the product often sits in a regulated workflow where the AI output is limited to a specific clinical indication, not a general eye-health judgment. FDA records classify diabetic-retinopathy detection software as a prescription Class II retinal diagnostic software device, designed to evaluate ophthalmic images for screening.
Monetizely's position is clear: AI vision screening in optometry should be priced primarily per completed, analyzable screening, backed by an annual minimum commitment and a modest site platform fee. Hardware should be financed or leased separately when needed. Per-seat pricing belongs in clinician-review and analytics products, not autonomous screening. Payment based on referrals, diagnoses, or reimbursement collections should be rejected.
Five linked commercial choices prevent pricing from disrupting care delivery
A price is the end of a sequence, not the first decision. Monetizely's 5-Step Pricing Framework starts with business goals and segments, then moves through packaging, the pricing metric, price points, and operational execution. The sequence matters because a practice cannot sensibly choose a rate before deciding whether it is buying a narrow diabetic-retinopathy screening tool, a broader retinal-risk platform, or a clinician workflow product. The approach is developed further in Monetizing Agentic AI.
For optometry, the five steps lead to a specific commercial architecture:
The core insight is simple: an optometry practice buys an answer for a patient encounter. It does not buy tokens, inference time, or a clinician's ability to sign in.
Before choosing a meter, leaders should distinguish among the products already emerging in the market.
| Product type | Named example | What the buyer receives | Most defensible primary meter |
|---|---|---|---|
| Autonomous disease screening | AEYE-DS; EyeArt | An AI result for a defined screening indication | Completed, analyzable screening |
| Retinal imaging AI with local software | Optain Assure Plus | Image grading across several stated conditions and risks | Site or device license, with analysis volume bands |
| Clinician workflow and image-data platform | RetinAI Discovery for Clinics | Data aggregation, image review, collaboration, and AI insights | Named clinician or reviewer seat, plus modules |
| Enterprise imaging network | Large multi-site deployment | Integration, governance, rollout support, and pooled volume | Annual network agreement with committed screen volume |
AEYE-DS describes autonomous diabetic-retinopathy screening that can run with indicated portable or tabletop cameras, while EyeArt's FDA summary describes image capture, server-based analysis, and outputs that include detected, negative, or ungradable results. Optain's January 2024 product specification describes local software that grades retinal images for diabetic retinopathy, AMD, glaucoma, and cardiovascular risk, whereas RetinAI positions Discovery for Clinics as a broader workflow and patient-management platform for optometrists and ophthalmologists.
The table shows why one rate card cannot serve every ophthalmic AI product. Autonomous screening should be sold as a clinical test workflow; analytics platforms can reasonably use seats and modules.
The Agentic Monetization Spectrum, or AMS, tests an AI product on three dimensions: zero-human ability, operational domain, and output/cost ratio. Zero-human ability asks how much of the work remains with a person. Operational domain asks whether the product performs a narrow task, a workflow, or work across functions. Output/cost ratio asks whether the value produced grows roughly with AI cost or far faster than it.
Those dimensions point pricing toward different anchors. A low-autonomy assistant can stay on a per-seat model because the clinician remains the visible source of work. A broad, highly autonomous system may justify an output or outcome meter. AI vision screening sits between those poles: it produces a defined clinical output autonomously, but staff still capture images, manage exceptions, explain findings, and coordinate referrals.
| AMS dimension | AI vision screening score | Evidence from the category | Pricing implication |
|---|---|---|---|
| Zero-human ability | Medium to large | AEYE-DS states that no specialist over-read is required for its defined diabetic-retinopathy use case, but staff still acquire images and clinicians manage care. (aeyehealth.com) | Move beyond clinician seats, but do not claim payment for the entire patient outcome |
| Operational domain | Small | FDA-cleared diabetic-retinopathy tools are indicated for a defined retinal screening task, not comprehensive ocular diagnosis. EyeArt outputs disease status or an ungradable result. (accessdata.fda.gov) | Use a narrow unit: a completed bilateral screening |
| Output/cost ratio | Linear to inflecting | Each additional screening requires image capture, processing, data handling, and support, while its near-term economic value is bounded by a single clinical workflow. | Use a per-screen meter with commitments, not a pure flat fee or a referral bounty |
AMS does not support a seat model for autonomous screening, and it does not support charging for a clinical outcome that depends on follow-up care. It supports charging for the completed screening output.
That distinction matters in a regulated setting. Eyenuk's EyeArt received FDA clearance in August 2020, and AEYE-DS received a subsequent 510(k) clearance on April 23, 2024. Neither clearance makes the software a replacement for the full scope of eye care. A product should therefore be priced for the job it actually performs: analyzing a defined image set and returning a defined result within its cleared indication.
Many vendors will present familiar software price structures. The buyer's task is not to admire the variety. It is to identify which meter reinforces the desired patient workflow and remains workable as volume grows.
A per-screen model works because it matches the event that creates both clinical and economic value: a patient receives an interpretable AI screening result. The vendor can meter that event cleanly. The practice can tie cost to volume. Finance can forecast spending from diabetic-patient visits rather than guessing how often clinicians will log in.
An unlimited subscription can look attractive, especially for a practice trying to remove friction from adoption. AEYE Health has described a subscription that bundles camera access and unlimited exams, and later described a monthly subscription model with unlimited exams. That approach can accelerate early use, but it also creates a problem: a 20-screen-per-month practice and a 300-screen-per-month practice can consume radically different value while paying the same amount.
The sensible answer is not to abandon predictability. It is to create predictability through an annual commitment. For example, a vendor can include 1,200 completed screens per site each year, bill quarterly, and charge a stated overage rate after the commitment is consumed. That structure gives the practice a budget while preserving a direct link between price and clinical throughput.
The table makes the choice plain: a completed screen is the main meter; a platform fee and annual minimum make that meter easy for buyers to budget.
Seat pricing belongs to review work, not automated screening work
The software market offers a useful contrast. RetinAI announced in December 2022 that its Discovery CORE research platform would move from a two-month trial to a €179-per-month, per-user subscription. That is logical for a collaborative image-analysis environment where clinicians and researchers repeatedly use a platform to review data, compare scans, annotate findings, and coordinate work.
The same logic does not hold for autonomous screening. A five-doctor optometry group may have one technician who operates the camera, two optometrists who review the chart, and several front-desk users who schedule follow-up. Which of those people should carry the AI seat? The question itself signals a broken meter.
Modern AI SaaS companies are reaching similar limits. As of September 7, 2026, Cursor lists a $40-per-user-per-month team plan but also uses included and on-demand model usage for variable AI costs. GitHub Copilot Business similarly lists $19 per granted seat per month and includes a monthly pool of AI credits. Those companies can preserve seats because developers remain the organizing unit of value. They add usage controls because compute varies.
Autonomous vision screening reverses that relationship. The patient encounter is the organizing unit of value, while the clinician is part of the workflow around it. A vendor that prices screening software per optometrist will either overcharge low-volume practices or leave revenue on the table at high-volume sites.
Hardware should be financed as equipment because it solves a separate adoption barrier
Camera availability is often the real constraint. A practice with a compatible fundus camera may need software integration, training, and support. Another practice may need the camera, maintenance coverage, connectivity, and AI workflow at once. Treating both customers as if they are purchasing the same product creates needless commercial confusion.
FDA documentation makes the equipment link concrete. EyeArt's cleared configuration specified particular cameras and required macula- and disc-centered images, while AEYE-DS identifies use with the Topcon NW400 and Optomed Aurora cameras. The camera is not merely a peripheral. It is required infrastructure for the screening workflow.
A practical offer should therefore separate three charges:
- One-time implementation: EHR or practice-management integration, workflow setup, staff training, and launch support.
- Equipment payment: Purchase, lease, or rental for the retinal camera and any associated maintenance.
- Software and analysis: Site platform fee plus committed and overage completed screens.
This separation helps buyers compare alternatives honestly. It also protects the vendor from hiding expensive equipment support inside a low advertised software rate. A low-volume practice can lease the hardware and begin with a smaller annual screen commitment. A larger group can purchase cameras outright and negotiate a pooled volume commitment across locations.
Equipment can be bundled for procurement convenience, but it should remain visible in the economics. The camera removes a capital barrier; the AI analysis produces the billable clinical output.
A five-location optometry group does not need five separate commercial models. It needs one standard definition of a completed screen, site-level workflow controls, and the ability to pool volume across locations.
CMS's 2026 Quality Rating System technical specifications include CPT 92229 as an autonomous eye exam pathway for diabetic-retinopathy measurement. That recognition makes standardized documentation and result delivery commercially important. It does not guarantee coverage, payment, or appropriate use in every payer contract or care setting.
The following structure keeps the commercial model stable while scaling administration.
| Customer segment | Contract architecture | Primary meter | Commercial objective |
|---|---|---|---|
| Single-site optometry practice | One site fee, low annual screen commitment, optional camera lease | Completed screen | Reduce adoption hurdle |
| Two to 20 location group | Per-site platform fee, pooled annual commitment, standard overages | Completed screen across the network | Encourage rollout without penalizing uneven site volume |
| Large health system or retail network | Enterprise agreement, implementation statement of work, minimum annual volume, site rollout schedule | Completed screen with network commitment | Support integration and governance at scale |
| Ophthalmology or retina review group | Named reviewer seats plus data-storage or module fees | Active clinician or reviewer seat | Monetize recurring professional workflow |
The unit should not change merely because the buyer gets larger. What changes is the level of commitment, integration support, pooling, and governance.
A practice should calculate what it will actually pay over three years, not compare only monthly subscription headlines. The model below shows why low-volume sites need a platform fee that stays modest and why high-volume sites need a declining effective cost per screen.
The economics improve with adoption because the site fee supports software access, support, compliance work, and integration, while the analysis charge scales with use. A vendor can use volume bands to lower the per-screen rate above agreed thresholds, but the invoice should always show the number of completed screens, included commitment, overage screens, and any equipment payment separately.
The model confirms the commercial purpose of the annual minimum: it lowers the effective price as screening becomes routine without forcing a low-volume practice into an unlimited plan it cannot justify.
A rate card fails when the invoice cannot be reconciled to the patient workflow. The operational rules should be specific enough that a technician, practice manager, and finance leader reach the same count.
The contract should define a billable screen as a bilateral exam that returns an analyzable result within the product's cleared workflow. It should not count a failed upload, duplicate patient record, technical outage, or image set that the system marks ungradable because of capture quality. EyeArt's FDA documentation explicitly recognizes ungradable outputs, which makes this distinction more than a commercial nicety.
A buyer should require the following before signing:
- A monthly export that lists completed, positive, negative, ungradable, duplicated, and reversed screens.
- No-charge retakes caused by software, transmission, or system-quality failures.
- A stated treatment for staff capture errors, with a reasonable retake allowance.
- Separate line items for implementation, equipment, software access, and per-screen analysis.
- A true-up schedule that avoids surprise monthly bills after a seasonal volume spike.
Those rules also create better incentives. Charging for every image upload encourages waste. Charging only for a valid, completed screening encourages the vendor to improve image-quality guidance and the practice to improve capture technique.
AI vision screening will not scale in optometry because vendors invent a more elaborate rate card. It will scale when the price reinforces routine screening at the point of care.
The strongest offer is a site platform fee plus an annual commitment to completed, analyzable screens, with transparent overages and optional camera financing. The primary meter remains the completed screen. The platform fee pays for the standing capability. The annual commitment gives the vendor predictable revenue and gives the practice a predictable budget.
The model avoids three costly errors at once: treating AI like a staff productivity tool, treating a clinical result like a referral bounty, and hiding equipment economics inside software pricing.
- Set a screening-volume target before selecting a vendor. Base it on eligible patient visits, not on the number of optometrists or cameras in the practice.
- Choose a vendor whose billable event matches a bilateral, analyzable clinical result. Do not accept billing based only on uploads, clicks, or user accounts.
- Compare three-year costs using one consistent volume forecast. Include platform fees, implementation, hardware, support, committed volume, overages, and expected site additions.
- Create separate buying paths for autonomous screening and clinician-review software. A per-screen purchase and a per-seat analytics purchase solve different problems and should not be forced into one agreement.
- Use pricing to drive adoption behavior. Give staff a workflow that supports retakes, makes results visible in the patient record, and does not make each additional appropriate screen feel like an unplanned expense.
Footnotes
- https://www.amazon.com/Monetizing-Agentic-AI-Handbook-Transformation/dp/B0H7Z13VKJ/
- U.S. Food and Drug Administration, “Product Classification: Diabetic Retinopathy Detection Device,” last updated August 3, 2026. (accessdata.fda.gov)
- U.S. Food and Drug Administration, EyeArt 510(k) summary and AEYE-DS 510(k) record, August 3, 2020 and April 23, 2024. (accessdata.fda.gov)
- Centers for Medicare & Medicaid Services, 2026 Quality Rating System Measure Technical Specifications, 2026. (cms.gov)
- AEYE Health, product and implementation materials describing autonomous screening, camera compatibility, and subscription structure, accessed September 7, 2026. (aeyehealth.com)