It feels like AI is everywhere in education right now. Generally, we’re optimistic that it can help improve outcomes for multilingual learners and their teachers, but we’re cautious of its overuse. Too much of it overpromises, oversimplifies, or quietly moves decisions out of a teacher’s hands without anyone noticing. We built Flashlight360 on the premise that AI should work like a good assistant, handling the busywork and offering a starting point, while every decision about students stays with the teacher who knows them best.

This article lays out how we use AI, the principles we won’t compromise, and introduces Tia, the teacher intelligence assistant that you’ll be seeing more of soon. Here’s where we stand.

Your Students Aren’t Talking to Chatbots

We’ve had many conversations with teachers about the trends they are seeing with AI. When teachers tell us they’re uneasy about AI, they usually mean generative AI: ChatGPT and tools like it that produce content on their own. Those worries are fair. Generative AI can invent facts, and student feedback is no place for guesswork.

AI is a broader category than the chatbots in the headlines, though. It includes technology schools have relied on for years: machine learning models that suggest a score against a rubric, speech-to-text that turns a recording into a transcript, and text-to-speech that reads feedback aloud. That is the AI inside Flashlight360 today. Narrow tools, each with one job, and a qualified person checking the work.

So here are two commitments you can hold us to: First, every piece of feedback a student receives in Flashlight360 starts with a person. Our team writes it, in English, by hand. There are no chatbots in the app, and nothing composes feedback on its own. Second, we don’t wire open-ended generative AI tools like ChatGPT into the application. The outside services we do use are the narrow specialists named above: turning speech into text, reading text aloud, or translating feedback, and each one works under an agreement that bars it from storing your students’ data or using it for anything beyond that single request.

Down the road, generative AI may earn a place in our tools for district administrators, for jobs like assembling program reports and answering questions about the data you already own. If it does, the same rules apply: your sensitive data will stay on our servers, and it won’t interface with students.

Your Data Is a Responsibility We Take Seriously

This is the piece we want to be unmistakably clear about. Whether it’s a student’s writing sample, a teacher’s account, or an administrator’s district-level data, that information exists for one reason: to help you teach, support, and lead. It is never sold. It is never shared with outside parties without your express consent. It is never used to train AI models outside of Flashlight360. Any model we build learns inside our walls, from our own teams’ work. And yes, we meet FERPA and COPPA standards, the federal laws that protect student records and children’s privacy.

We treat your data as something entrusted to us, and we build our product with that responsibility in mind. Our engineering team continuously assesses, tests, and adapts our systems against security vulnerabilities as an ongoing part of how we operate. Protecting student, teacher, and administrator data is a standing commitment we hold ourselves to every day, and we take real pride in it.

Teachers Make the Final Call

Nothing in Flashlight360 finalizes a decision about a student without a teacher’s review. AI offers a suggestion. The conclusion belongs to the teacher, because only a teacher has the context, the relationship, and the judgment that a rubric or a model was never designed to hold. The goal is for AI to reduce repetitive and time-consuming work so teachers can spend more time on instruction and support.

This is also where our Flashlight360 Scoring Team comes in. They’re real people who review student speaking and writing every day. AI gives them a faster starting point, so they can extend that same careful, human review to more students without stretching what any one person can realistically do in a day. Every piece of written feedback a student receives is always crafted by a person on our team. 

We Tell You What’s Happening

We don’t build things quietly and let you discover them later. When AI is involved in a feature, in-app support and our Partnership Success Coaches will help you understand exactly what it does, what it doesn’t do, and where its limits are. We release AI abilities in small steps. Every AI assist is carefully released and monitored; most notably, its impact is shaped by your feedback. We work in the open, and we want your feedback on the decisions that matter most for you and your students. 

You’ve likely already interfaced with two existing features in the app that run on this kind of narrow AI.

Transcriptions make student speaking and writing submissions easier to review on the page when a busy classroom makes listening hard. This also enables teachers to print more readily, whether that’s for a parent-teacher conference or as evidence of a student’s progress over time.

Read It To Me lets students hear their feedback read aloud, sped up, slowed down, or switched between English and Spanish, giving multilingual learners scaffolding for understanding their feedback.

Transcriptions and Read It To Me both run on Deepgram, a speech technology tool we use specifically for this work. Before anything reaches Deepgram, we strip out all personally identifiable information, so a request is just the audio or text itself, never who it belongs to, which keeps us FERPA compliant. And our agreement with Deepgram is strict: they can’t store what we send or use it for anything beyond the single transcription or read-aloud request in front of them.

Spanish feedback works the same way. DeepL, a dedicated translation service, operates under the same kind of agreement: nothing we send can be stored or used beyond that single translation request.

Beyond these narrowAI features, and in the spirit of openness and transparency, we have an exciting announcement!

Meet Tia

When a student submits a Teacher-Ready Assignment, you’ll see something new next to their work: a suggested score recommended by Tia.

Tia stands for Teacher Intelligence Assistant. But say the name out loud. In Spanish, tía means aunt, the trusted adult who shows up, pitches in, and cares about your kids almost as much as you do. In some cultures, tía is even what children call a beloved teacher.

So what is Tia, exactly? Tia is the machine learning inside Flashlight360 that suggests scores on student work. Its training happened close to home. For 4 years, our internal team of teacher professionals have reviewed hundreds of thousands of speaking and writing submissions from multilingual learners across the country. That careful human work is what taught Tia which performance features matter, specifically for a K-12 multilingual population.

Where You’ll See Tia

Tia-suggested scores are its first visible job, and a score only becomes official once you say so. When a student submits a Teacher-Ready Assignment, Tia suggests a score using the same rubric a teacher would use to assess the five speaking and three writing indicators. That suggestion is never final. It’s visible, clearly labeled with how confident it is, and entirely yours to confirm, adjust, or set aside. Students never see Tia’s suggested score.

Where Tia Is Headed

Tia will grow the way it started: one careful step at a time, with teachers reviewing every score delivered to students. Next, we’re working on helping Tia connect score detail across assignments and benchmarks, so you can see growth over time without building that picture by hand. Further out, we expect Tia will suggest what a student might work on next and point you to resources that fit. The rule stays the same the whole way: Tia suggests, you decide.

What This Means for You

For teachers, it means Tia gets you to a starting point faster, without ever taking the final word away from you. For administrators, it means faster turnaround without compromising on quality. For students, it means feedback that’s easier to access and easier to understand. For everyone, it means the same thing is true today as it always has been: a person is accountable for a student’s score, a person is accountable for what happens to your data, and we don’t take either of those responsibilities lightly.

If your district has specific questions about data handling, retention, or security practices, we’d love to put you directly in touch with a Flashlight360 team member. Your questions and feedback deserve a real conversation.

Common Questions

Who (or what) is Tia? Tia is our Teacher Intelligence Assistant: the machine learning inside Flashlight360 that suggests scores on student work. It is a trusted helper for your classroom, with you in charge. 

Does Flashlight360 use generative AI, like ChatGPT? There are no chatbots in the app and nothing that creates content on its own. The AI at work is machine learning for suggested scores, speech-to-text and text-to-speech for transcripts and read-aloud, and machine translation for Spanish feedback. If generative AI joins our tools for administrators in the future, your sensitive data will stay on our servers, and it won’t interface with students.

Does Tia grade my students? No. Tia offers a suggested score. A teacher reviews and confirms every score before it’s official.

What happens if Tia gets something wrong? A teacher catches it during review. That’s exactly why the review step exists. 

Is my students’ data safe? Yes. We meet FERPA and COPPA standards. It’s used only to support scoring and instruction, never sold, never shared for unrelated purposes, never used to train outside AI models, and it sits inside systems our engineering team actively monitor and harden against vulnerabilities. If you want more detail on how this applies to your district, the Flashlight360 team is available. 

What powers the Transcription and Read It To Me features? Deepgram, a speech technology tool we chose for this work. Spanish translations of feedback come from DeepL, a dedicated translation service, under the same restrictions. If you have questions about how these features work, we’re happy to answer them.

Have a question this article doesn’t answer? Reach out to your Flashlight360 team directly. We’re glad you’ve met Tia. We think you two are going to get along great.