Right now, the discussion about AI in K-12 education is everywhere—dominating headlines, inspiring hope, and raising tough questions. From breakthrough innovations to concerns about equity and ethics, the conversation is loud and ongoing. And complicated.

So we want to cut through the noise and show some of the real ways AI is improving communities. 

At TalkingPoints, we believe the most important question isn’t ‘How do you use AI?’—it’s: ‘How can we use AI to do the most good for students, families, and educators—especially in communities that have long been under-resourced?’

As we celebrate 10 years of impact, we’re launching a new blog series to explore that question from our perspective as an AI-native organization. Over five parts, we’ll spotlight how we’ve used AI to break down barriers and unlock the superpower of families. We’ll share what we’ve learned along the way and what’s next as we build toward a brighter future in education.

We’re excited to share our reflections, tools, and lessons learned—and to hear from you. What questions are you asking about AI? What possibilities are you exploring? Let’s shape the future together.

Have thoughts, hopes, or worries about AI in education? Send us a note—we’re listening.

What’s next: Exploring AI and the future of school–family collaboration

Written by Aram Gugusian,

Head of Engineering at TalkingPoints

When I helped build the first prototype of TalkingPoints, I wasn’t just writing code. I was building something my own family needed.

Growing up as an Armenian student in Uruguay, I saw how language and culture shaped – and often limited – the relationship between families and schools. That experience still drives how I think about technology today: not as an end in itself, but as infrastructure for connection.

For more than a decade, TalkingPoints has stayed focused on one thing: helping educators and families work better together so students can thrive. AI has been central to that work – not as a buzzword, but as a practical tool we’ve applied carefully, iteratively, and at scale.

How is TalkingPoints using AI in K–12 school–family communication?

We were using AI in school communication long before it was fashionable.

On the product side, our most foundational use of AI has been translation. TalkingPoints supports two-way communication in over 150 languages. That didn’t happen by prompting a model and hoping for the best.

We built and managed our own custom translation models. Early on, there was no off-the-shelf solution that worked for the kinds of messages schools send: nuanced, contextual, often sensitive. We invested heavily in human translators to generate high-quality data, then used that work as a feedback loop to train and continuously improve our models. It was slow. It was expensive. It worked.

That investment put us in a position where our translations outperform general-purpose models for education-specific use cases – even as larger companies train on vastly bigger datasets. Quality comes from curation, not just scale. 

On top of translation, we now use AI to:

• Classify message topics and intentions
• Turn unstructured text (like absence explanations) into structured data
• Surface patterns in attendance, engagement, and communication
• Provide educators with feedback and suggestions while they write messages

Work that once took teams of people months to complete – manually reading, tagging, and analyzing messages – now happens in minutes. The difference isn’t incremental. It’s like going from horses to cars.

How does TalkingPoints keep AI in edtech engineering fast—but safe—for K–12?

We also use AI heavily inside our engineering organization.

Engineers use AI as an assistant to write code, generate tests, identify bugs, retrieve documentation, and navigate a large and complex codebase. We’ve even built custom internal tools. For example, we’re layering AI-powered workflows on top of off-the-shelf project management tools, which weren’t designed for how we actually work.

This has changed our velocity. Things that would have taken multiple engineers months can now be built by a small team in weeks.

But we’re careful. AI works best when developers know what “good” looks like. We don’t let it run unsupervised. Everything we build has to operate reliably at scale, across millions of messages and users. Evaluation matters as much as generation.

“Evaluation matters as much as generation.” -Aram Gugusian, TalkingPoints Head of Engineering

AI can help with evaluation, but only if you already have strong data. Models don’t replace judgment; they amplify whatever you feed them.

Cost, scale, and reality checks

AI is powerful — and still expensive.

For organizations like ours, large-scale analysis of our data can cost more than traditional infrastructure. That’s why we don’t run system-wide analysis indiscriminately; we work with carefully selected subsets. Big companies are building their own hardware to address this. 

Over time, costs will come down. But right now, scarcity is real.

There’s also hype (some of it will fade), but AI is already driving real economic change, and it’s not going away. The question isn’t whether to use it, it’s how to use it responsibly.

Our approach is simple: read constantly, test relentlessly, share learnings internally, and embed what works into our actual processes. Every other Friday, engineers share what they’ve tried – not just demos, but real outcomes.

“AI is already driving real economic change, and it’s not going away. The question isn’t whether to use it, it’s how to use it responsibly.” -Aram Gugusian, TalkingPoints Head of Engineering

What’s next for AI at TalkingPoints—and what does it mean for district leaders?

Our platform already goes far beyond translation. We help educators write stronger messages, nudge families with relevant learning activities, and give district leaders real-time insight into what’s happening across their schools.

Through our Solutions Incubator, we’re exploring what comes next, including agentic and retrieval-based systems, without rushing deployment.

Some examples:

• Using AI agents to help partners map their data into our schemas faster, reducing back-and-forth and enabling more self-serve integrations
• Letting users retrieve and act on data through natural language (e.g., “Show me students with chronic absenteeism due to health issues and message their families”)
• Automating classification, retrieval, and action – always with humans in the loop

We’re not chasing novelty. We’re solving very specific problems that reduce friction for educators and families. For K–12 district leaders, these systems mean clearer visibility into patterns like chronic absenteeism and more targeted support for schools.

Staying grounded

Three principles guide everything we build:

• Trust before automation — AI should strengthen human relationships, not replace them.
• Simplicity at scale — Tools must work on real devices, in real classrooms, for real families.
• Usefulness over novelty — If it doesn’t improve outcomes or save time, we don’t ship it.

The future we’re building isn’t about more features. It’s about clearer communication, better coordination, and systems that help teachers and families focus on what matters most: student success.

The technology is here. The infrastructure is in place. And we’re just getting started.

See TalkingPoints in action – Watch this 90 second video

Celebrating 10 years of partnership and impact

For a decade, TalkingPoints has supported students, families, and educators through the power of effective family-school partnerships. As an education technology nonprofit, our award-winning communication and family engagement platform has improved outcomes for districts and students across the country.

We connect 9 million+ educators, students, and family members annually and have facilitated more than one billion conversations—building trust, fostering relationships, and fueling student success. Named by Common Sense Education as “the best overall family communication platform for teachers and schools,” TalkingPoints drives measurable gains in attendance and academic achievement, backed by rigorous, causal research.

Join us in building a future where every child has the support they need to thrive.