We Are Not an Accessibility Company. We Are a Technology Company.

CLERC is not an accessibility company.

We want to say that clearly and early, because the assumption follows us into every conversation. CLERC is Deaf-led. The team is Deaf. We work with sign language data. Therefore, people conclude, we must be building an accessibility tool: an app, a translator, something that helps hearing and Deaf people communicate.

We are not.

CLERC builds sign language AI infrastructure. Our customers are AI research labs, sign language recognition teams, translation engine builders, computer vision groups, and academic labs working on multimodal models. We sell structured sign language data. That is the product. That is the business. The accessibility outcomes that will eventually follow are downstream of that.

This is a deliberate, assumed choice. It is also the result of a lesson paid for with years of work inside the field. Here is why we drew the line where we did.

What years inside accessibility taught us

Before CLERC, we spent years building inside the accessibility space. We know the terrain, the people, the grants, the pilots, the pitch decks. We know what works, and we know what quietly breaks three quarters into the project.

The pattern is always the same. You build a tool aimed at Deaf users. You raise a little money on the premise of impact. You try to turn the tool into a product. And then the business model fractures in exactly three places.

First, the market is smaller than it looks. Sign language is not monolithic. ASL, LSF, BSL, JSL, and dozens of regional variants are each effectively a separate product surface. A single accessibility tool cannot honestly serve all of them without either flattening the differences or fragmenting into a dozen half-finished localizations.

Second, the buyer is not the user. Individual Deaf users will rarely pay a software subscription for an accessibility tool. The buyer is the institution: the school, the hospital, the government agency, the corporate HR team. Those procurement cycles are long, the RFPs painful, and the willingness to pay has a hard ceiling set by public budgets.

Third, accessibility is plural. Deaf is not the only constituency. Blind, low-vision, cognitive, dyslexic, motor, all live under the same policy umbrella and the same procurement line items. An accessibility company that wants to be financially serious is pulled toward covering all of them, diluting focus until nothing is excellent. A company that stays narrow on one constituency struggles to reach scale. Both failure modes are well documented inside the field.

This is not a critique of the people working in accessibility. Many are doing important, beautiful work. It is a critique of the business model. As a for-profit company trying to build something durable, CLERC no longer believes that accessibility, framed as a product category, is a viable standalone thesis for sign language at today's cost structures and today's willingness to pay.

What we actually sell

CLERC sells the raw material that every serious sign language AI system needs and that almost none of them has: a high-quality corpus, built Deaf-led, with native signers and Deaf annotators, structured from day one.

That is not an accessibility product. That is infrastructure.

The sign language corpora that already exist were built for linguistics: documentation, phonology, lexicography. They were never designed to train a modern vision-language model. Their annotation schemas, their framing, their licensing, their sampling strategies all reflect a different purpose. CLERC's dataset is purpose-built for ML training. That gap is the reason we exist.

Our customers are not end users. They are the teams whose models fail silently because the training data underneath them is inconsistent, incomplete, or annotated by people who cannot read sign language fluently. They are the research labs trying to publish on SLR benchmarks that do not yet exist. They are the avatar companies whose 3D characters sign nonsense because no one ever gave them clean per-frame annotations. They are the translation engine builders who keep hitting the same ceiling because they have a data problem dressed up as a model problem.

For those buyers, the willingness to pay is a different order of magnitude. A research lab licensing a structured corpus is a six-figure conversation, not a sixty-euro subscription. The unit economics close. The market is narrow, technical, and global. The product gets better every time the corpus grows, because the underlying asset compounds.

This is why CLERC chose tech over accessibility. Not because accessibility does not matter. Because this is the layer where the business actually works.

Why a Deaf team has to build it

If the product is structured sign language data, then data quality is everything. And data quality, in sign language, is not a process problem. It is an identity problem.

A hearing team annotating sign language makes judgment calls it is not equipped to make. Gloss boundaries are ambiguous. Regional variants get flattened or mislabeled. Facial expressions that carry grammatical meaning get tagged as emotion. Register shifts get erased. The error rate compounds silently, and by the time a customer trains a model on it, the damage is already baked in and invisible.

CLERC is Deaf. The annotators are Deaf. The signers are native. This is not a values statement. It is a technical requirement for honest data.

That requirement happens to align with something we believe matters independently: Deaf people should be the ones building the infrastructure that hearing AI companies are about to train their systems on. Whoever builds the data layer writes the grammar of every downstream model. We would rather that grammar be written by the community it describes than by people who learned sign language from a textbook.

But we want to be precise. We are not asking customers to buy from CLERC because we are Deaf. We are telling them that a Deaf-built corpus has measurably lower annotation error than a hearing-built one, and that difference will show up in their model accuracy the moment they train on it. The ethics and the engineering point the same way. That is the useful kind of alignment.

The long-term dividend

Here is the part that looks paradoxical and isn't.

By positioning CLERC as a tech company rather than an accessibility company, we can create a different kind of value, one that ultimately complements, rather than competes with, the work accessibility actors are already doing.

If the data layer is strong, downstream systems get built that were not previously possible. Better SLR. Better translation. Better avatars. Better education tools. Better interpretation augmentation. Better content creation pipelines. None of these are products CLERC will ship directly. Many of them will be built by accessibility companies, educational platforms, public-sector teams, and research groups whose mission is exactly the one CLERC has chosen not to pursue. Those actors need a trustworthy data foundation to build on. Providing that foundation is how we contribute.

Over time, this shifts where economic gravity sits in the sign language AI ecosystem. Deaf entrepreneurs, Deaf artists, Deaf researchers, Deaf freelancers gain the ability to build on top of a foundation that speaks their language natively. That is structurally different from a situation where every tool is built by hearing teams and then marketed back to the community.

This is the opportunity dividend, and it comes from building tech, not from competing with accessibility. The two layers reinforce each other. CLERC's job is to make the bottom one honest.

What this means in practice

In practice, this shapes everything. We don't chase accessibility RFPs. We don't pitch disability-services buyers. We don't benchmark against apps. We don't call ourselves a translation tool, because we aren't one. We sit in front of CTOs and research directors, not accessibility procurement officers. We benchmark against data infrastructure companies and corpus providers. We price on value delivered to technical teams, not on public-sector willingness to pay.

This is a narrower market. It is also a market that pays.

---

CLERC is building the sign language AI infrastructure the field needs. We are building it from inside the Deaf community. We are building it as a real business, with real customers, real revenue, and real unit economics, because that is the only version of this that lasts long enough to matter.

Everything else follows from the data.

Follow @CLERC to track the data layer we're building.