Could AI Replace Bilingual Teachers in American Classrooms? A Houston Pilot Has Answers

Students at Alpha School’s K–8 campus, located in a renovated office building south of Zilker Park in Austin, have already completed many hours of arithmetic, reading, science, and history by 1 p.m. on a typical school day. Before lunch, the AI-powered, self-paced training that Alpha refers to as “2 Hour Learning” was completed. The children spend the remainder of the day working on leadership workshops, financial literacy exercises, and passion projects with what Alpha refers to as “guides”—well-paid adults who are specifically not required to have teaching credentials and whose duties include social development, emotional support, and motivation rather than academic instruction.

For the 2026–2027 academic year, the school will expand to Houston’s The Woodlands, where K–8 seats will cost $40,000 annually. The question that bilingual educators in the same metro area are keeping an eye on is whether any iteration of this model can address the one issue that Houston’s school system has been unable to resolve through traditional hiring: the ongoing lack of competent bilingual instructors.

In 2023, there were over 16,000 open teaching positions in Texas, with bilingual certification being one of the hardest to fill. Using a combination of multilingual paraprofessionals, temporary waivers, and technology supplements, Houston ISD, which serves over 194,000 kids and has one of the biggest populations of English language learners in the nation, successfully managed this issue. From the standpoint of a district administration, it is easy to see the appeal of an AI model that could hypothetically handle Spanish-English vocabulary training without having a bilingual-certified instructor.

In a supplemental capacity, Alpha has been testing its 2 Hour Learning platform with 1,500 public school children in more than 50 districts. This is done in parallel with academic enrichment rather than in place of teachers. Before believing the headline claims of “2x learning speed” and “top 1% nationwide test scores,” one should carefully examine the data, assuming it exists at all, as it is primarily internal and has not been independently confirmed by researchers.

It is worthwhile to look more closely than broadly at the differences between what AI currently performs in bilingual classrooms and what a bilingual instructor does. According to a Walton Family Foundation and Gallup poll, six out of ten American teachers used an AI tool during the 2024–2025 school year. The majority of these applications were in administrative or lesson-planning contexts, which reduced the time cost of tasks that don’t require human judgment.

For ELL students who need to ask a clarifying question in Spanish while a subject is being taught in English, AI translation capabilities and vocabulary scaffolding tools are actually helpful, and they are rapidly getting better. They don’t read the room, though. This type of attunement involves cultural familiarity, language-based emotional intelligence, and the accumulated knowledge of a specific student over months of classroom relationship.

For example, a bilingual teacher may notice that a particular child has become quiet not because they are confused about vocabulary but rather because something uncomfortable happened at home before school. No AI system in use today comes close to that capability, and the truthful ones in the industry don’t make such claims.

As the Alpha model has grown, its legitimacy has suffered certain institutional setbacks. The charter application was rejected by Pennsylvania’s Department of Education due to the AI educational model’s “untested” status. South Carolina, Utah, North Carolina, and Arkansas all saw decreases. In 2024, Arizona became the first state to authorize a public version. In late 2025, the comptroller of Texas temporarily prohibited Alpha and other Cognia-accredited schools from taking part in the state’s new voucher scheme. The examination is justified.

The bilingual teacher shortage in Houston’s public schools cannot be solved by a school model that charges $40,000 annually, serves mostly wealthy families who can afford the tuition, and claims to produce top 1-2% national academic results through an AI platform that has no independent verification of those claims. At most, it is a thought-provoking exercise that takes place in a different reality than the one in which the majority of Houston’s ELL students really learn

The charter rejections in several states, the cautious 50-district supplemental piloting, and the Alpha expansion into The Woodlands all give the impression that the education technology sector and the real working conditions of bilingual public school teachers are still talking past each other. Over the next ten years, AI will transform the work that bilingual educators conduct in the classroom.

Some of it has already changed. Routine translation activities, vocabulary drills, and customized practice exercises are examples of duties that could benefit from AI support in order to free up teacher time for more difficult human work. However, the more labor-intensive aspect of teaching a language is more difficult. Additionally, it doesn’t seem like that part is on the replacement timetable as of yet.