OrthoTriage AI wants to give general dentists an objective complexity score before they commit to a case.
Go spend twenty minutes in the Invisalign Doctor Community Facebook group. You'll find it almost immediately: a GP posting a photo of a patient's teeth, asking their peers "would you treat this?" Sometimes there are fifty comments. Sometimes they disagree wildly. The GP walks away more confused than when they started, treats the case anyway, and eighteen months later they're issuing a refund or calling in a specialist to clean up the mess.
This is not a rare edge case. It's a daily occurrence across tens of thousands of general dentists who got Invisalign certified because aligner therapy has become a meaningful revenue stream, then discovered that "certified" and "qualified for every case" are not the same thing.
Align Technology has trained over 220,000 doctors globally on Invisalign. A significant chunk of those are GPs, not orthodontists. The training covers the mechanics of the product. It does not reliably teach clinicians where their personal case complexity ceiling is, or how to recognize when a patient's skeletal pattern or crowding severity pushes a case beyond GP territory.
The result is predictable. GPs take on borderline cases based on gut feel, peer opinion, or optimism. Specialists receive referrals that are either trivially simple (lost revenue for the GP, wasted appointment for the specialist) or dangerously complex (the GP didn't recognize what they were looking at). There's no standardized framework that exists between "I think I can handle this" and "I know I can't handle this."
Clinical triage tools in orthodontics do exist, but they're not designed for this specific decision point. ABO grading rubrics are for board exams, not pre-case screening. ClinCheck is post-acceptance. Dental Monitoring is for ongoing case tracking. The pre-commitment moment, where a GP looks at photos and decides whether to take the case at all, is genuinely unoccupied.
Two things converged. AI-assisted image analysis got cheap and capable enough that you can extract meaningful orthodontic complexity signals from clinical photos without proprietary hardware. And GP adoption of orthodontic procedures accelerated fast enough that the training gap became visible at scale.
The Facebook group posts are the proof-of-demand signal here. GPs are already doing manual crowdsourced triage. They're uploading photos and asking strangers for clinical opinions, which is both a liability nightmare and evidence that demand is completely real. Nobody invented this behavior. It emerged organically because the alternative, a reliable tool that doesn't exist, would be preferable.
If you assume 25-30% of Align's trained doctors are actively offering aligners, you're looking at roughly 60,000 US GPs. At $49/month, that's a $35M annual revenue ceiling for the US alone. Add the EU and ANZ markets, add the GPs using Spark or uLab instead of Invisalign, and the number climbs meaningfully. This is not a billion-dollar TAM. It's a focused, reachable, defensible niche, and for a solo founder or small team, that's actually better than chasing an abstract massive market.
Break-even is 42 paying customers at the solo plan price. That's achievable in a matter of weeks through community channels if the product works.
Honestly, the competitive landscape here is surprisingly sparse for a problem this obvious. No YC-backed company has touched this specific workflow. Dental Monitoring has distribution and clinical credibility, but their product is post-case monitoring, not pre-case screening. CephX does AI-assisted cephalometric analysis, but that requires radiographs and is aimed at orthodontists who already have the clinical training to interpret results.
The most credible competitive threat is Align Technology itself. They own approximately 80% of the GP aligner market, they see every case that goes through their portal, and they have the data and distribution to bundle a complexity score into their existing GP onboarding flow without breaking a sweat. Their 2024 iTero integration announcements should make any founder in this space nervous. If Align decides this feature is worth building, they can commoditize the core value proposition for the majority of the target market overnight.
The brand-agnostic angle is the real wedge. Align cannot credibly offer neutral complexity scoring that might tell a GP to use a competitor's aligners. The roughly 40% of GPs who use Spark, uLab, or other systems have no incumbent solving this for them. That's the defensible entry point.
The validation test described here is smart and worth doing before writing a single line of code. Post in the Invisalign Doctor Community, collect case photo submissions via Typeform, pay an orthodontist consultant $50/hour to score twenty cases, and email results within 24 hours. If you can't get ten GPs to submit real cases in two weeks, the problem isn't painful enough to pay for.
Assuming validation holds, the MVP is genuinely buildable by one developer in six to eight weeks:
The ML layer comes later, at week twelve or so, once you've accumulated two hundred or more labeled cases through the clinician feedback loop. That feedback loop is the whole game. Every GP who submits outcome data (treated successfully, referred out, treatment abandoned) is labeling training data. After five thousand cases, the model's accuracy creates a performance gap that a new entrant without that data cannot close quickly. That's the actual moat.
I want to be honest about the FDA situation because it's the issue most indie hackers will wave away and then get badly surprised by later. Clinical decision support tools that influence diagnosis or treatment selection often require 510(k) clearance as Class II Software as a Medical Device. The FDA's 2024 AI/ML guidance meaningfully narrowed the "clinical reference tool" framing that startups use to avoid this pathway. Framing the output as educational rather than diagnostic is legally fragile, not legally safe.
The regulatory cost, $250K to $1M+ and eighteen to thirty-six months, would likely kill a bootstrapped company before it reaches product-market fit. The practical path is to launch under enforcement discretion with aggressive disclaimer language (not a substitute for clinical judgment, for reference only), consult the FDA pre-submission program early, and plan the 510(k) pathway for after you have Series A capital. This is not a solved problem. It's a managed risk, and you need to manage it consciously.
There's another risk that's harder to quantify. The GPs who most need this tool, recent Invisalign certifications who feel anxious about case selection, are also the most likely to stop using it if it consistently tells them to refer cases out. Referrals feel like lost revenue. A product whose core job is preventing bad cases will sometimes conflict with the user's financial incentive to take those cases. Retention dynamics get complicated fast.
And then there's liability. If OrthoTriage scores a case as GP-safe and the outcome is poor, the plaintiff's attorney will subpoena the tool's output. That exposure requires careful contract language and probably professional liability insurance that a bootstrapped product cannot easily afford early on.
The validation test is week one. The community channel (Invisalign Doctor Community Facebook group, dental study clubs) is week two. If ten GPs submit real cases and three express willingness to pay $49/month, you have enough signal to build.
Keep the FDA risk front of mind from day one. Talk to a regulatory consultant before launch, not after. Document every disclaimer decision you make and why.
Lose sleep about Align's roadmap. Watch their product announcements. Build the brand-agnostic positioning into every piece of marketing copy so that when Align eventually adds complexity scoring, you have a story that still makes sense for the 40% of the market they can't own.
The outcome dataset is the long game. Every case submission, every clinician feedback response, every referred-and-confirmed outcome is a data point that your competitors cannot buy. That's worth optimizing for even when it slows down other things.
This idea is genuinely worth exploring, with eyes open to the regulatory trap and the Align dependency risk. The demand is proven, the technology is accessible, the gap is real. What's uncertain is whether you can build a sustainable business before the incumbents get bored enough to look in this direction.