Build vs Buy: AI Planning for Implant Makers
Build vs buy surgical planning software OEM: honest engineering, regulatory and maintenance trade-offs for implant makers, and when to license a vendor-neutral CT-to-plan engine.
Key takeaways
If you make orthopedic implants and want to offer modern digital planning, you face a classic build versus buy decision. Building an in-house AI planning platform means hiring a data-science and medical-imaging team, sourcing and curating annotated CT data, running validation studies, and carrying the software as a regulated product for its whole life: not one project, but a permanent engineering and quality obligation. Buying or licensing a vendor-neutral planning engine lets you offer CT-to-plan planning tied to your implant catalog without building a robot or a data-science org, so your team stays focused on implants and distribution. The honest trade-off: building maximizes control and long-run ownership but is slow, expensive and risky; licensing trades some control for speed and focus. Salnus is built for the license and co-development path: an implant-agnostic, CT-based planning engine (segmentation, CPAK phenotyping, alignment options and sizing support), currently Research Use Only (RUO), not a cleared device. This post lays out the drivers so you can decide honestly.
Why the question comes up now
Surgeons increasingly expect a 3D, CT-based plan, and the robotic platforms have set that expectation. If you sell a strong implant but have no robot and no planning software, you are competing against companies that bundle planning as part of the sale. You do not need a robot to answer that, but you do need a credible digital planning story. That leaves two routes: build one, or license one. For the broader landscape of who offers what, see our guide on which manufacturers offer AI knee planning.
The honest cost of building
Building looks attractive because you keep everything in-house. Be clear-eyed about what "everything" includes.
- A specialized team. Medical-imaging AI is not general software. You need people who can do CT segmentation, geometric modeling, and clinical measurement, plus quality and regulatory staff. That is a standing payroll, not a one-time hire.
- Data. Models are only as good as the annotated imaging behind them. Sourcing, de-identifying and expertly annotating CT data is slow, expensive and never truly finished.
- Validation. A planning claim has to be validated against expert ground truth, with reproducibility and generalization across scanners and populations. See our note on validation metrics that actually matter.
- Regulatory carry. As Software as a Medical Device (SaMD), a planning product needs a quality system, documentation and a clearance pathway, and it stays a regulated product every day it exists.
- Maintenance forever. Model drift, new scanners, OS and browser changes, security patches, and post-market surveillance do not stop after launch. The build cost is the smaller half; the maintenance tail is the larger half.
None of this is a reason not to build. It is a reason to price the whole life of the product, not the demo.
The honest trade-off of licensing
Licensing a vendor-neutral engine removes most of that standing burden, and adds its own considerations.
What you gain: speed to a working offering, a team that already lives in CT-to-plan work, and focus. Your engineers keep building implants and instruments instead of becoming a medical-imaging AI shop. Because the engine is implant-agnostic, it plans onto your catalog without forcing you to also own the planning science. For how that neutrality works, see implant-agnostic surgical planning.
What you weigh: you depend on a partner, so the contract matters. The things to define up front are IP ownership, field-of-use and exclusivity, data handling and residency, roadmap and support commitments, and what happens if either party walks away. A licensing deal is only as good as those terms, and they should be negotiated, not assumed.
A middle path: co-development. Build versus buy is not binary. A co-development or white-label engagement lets a manufacturer shape the planning workflow around its catalog and clinical philosophy while a specialist carries the imaging AI and validation. You get influence without standing up a full data-science org.
Where Salnus fits
Salnus is deliberately positioned as the license and co-development option, not a robot vendor and not a competitor for your implant relationship. The model is straightforward and manufacturer-pays: Salnus provides the vendor-neutral, CT-based planning software (segmentation, CPAK phenotyping, alignment options and sizing support); you keep your implant relationship and channel; and IP is contract-defined rather than assumed by default. Because Salnus is implant-agnostic, the same engine works across catalogs, so a manufacturer can offer modern planning without building a robot or a data-science team.
Two honest boundaries. First, Salnus is currently Research Use Only (RUO), not a cleared device, so an engagement is framed as a pilot, co-development or licensing collaboration, not the purchase of a finished cleared product. Second, the point of vendor-neutral software is that it complements rather than replaces surgical judgment; for how it sits next to robotic systems, see robotic versus software TKA planning.
A simple way to decide
Ask three questions. Is planning core to your strategy, or a feature you need to offer? If it is a feature, licensing usually wins. Do you have, or want to permanently fund, a medical-imaging AI and regulatory team? If not, building underestimates the real cost. How much control over the science do you need versus speed to market? More control favors build or co-development; more speed favors license. For a technical view of what the planning engine actually does, see how CT 3D preoperative planning works, and for how the tools compare, our 2026 orthopedic 3D planning software comparison.
Bottom line
Building an AI planning platform gives you the most control and the most long-run ownership, but it is a permanent engineering, data and regulatory commitment, and the maintenance tail outlasts the build. Licensing or co-developing a vendor-neutral engine trades some control for speed and focus, and lets you offer modern CT-based planning tied to your catalog without building a robot or a data-science team. If you are an implant maker weighing that decision, Salnus is built to be the license or co-development partner, currently Research Use Only. To explore a pilot or co-development engagement, see what Salnus offers manufacturers.
Reviewed by the Salnus biomedical engineering team.