Navigating the AI Influence

I’ve started to write this article many times, wanting to show comprehensively how exactly our company uses AI capabilities, so it might be helpful for others to look through our scenarios and maybe take something as an example to follow.

The thing is, every time I began writing, we were right in the middle of introducing something new in the field of AI. So it always felt like I should wait a little longer, so the article wouldn’t become outdated the moment it was published. But today is no different—we’re developing brand-new functionality that simply wasn’t possible until these recent AI-centred times. And I assume tomorrow won’t be different either—we’ll again be developing something extra.

So, I decided to change my approach. Rather than trying to share something fully accomplished, I’ll treat it as a journey. Let’s say these thoughts are just today’s snapshot of how we’re leveraging AI in our company at this moment, and there will be another one later. And another one, hopefully.

Let’s begin with who we are, because it’s important to clarify both what we are and what we aren’t, before we dive into exploring our AI maturity.

Who We Are

A short overview of what we at Watchers.io do in a nutshell: we provide an easily embeddable social layer for digital platforms, so any platform can instantly become social by integrating us. Their users get the opportunity to interact with each other, share, learn, discuss the platform’s content and products, help one another, and much more.

The main reason a platform owner would want to gather a community inside their product rather than starting a group or channel on external social media is simple: when users build a community inside your platform, they stay on your platform. Their behaviour evolves from basic chatting to deep, platform-connected scenarios. This bonding leads not only to better retention and longer sessions, but also to genuine loyalty and a higher volume of transactions.

So, as you can see, our product was not initially AI-first. Our value could be (and has been) delivered even before the current AI revolution, since we’ve always had human-to-human interactions at the heart of our solution.

Nevertheless, today we are AI-powered in our product, and—dare I say—AI-first in the way our business operates.

AI-Powered Product

Below, you can see the list of our main AI-powered products and features: what they do, which technologies we use, and what value they provide for our clients.

AI-Optimised Moderation

No community engagement is possible in a toxic environment. And when I say “toxic,” I’m not speaking only about inappropriate language, harassment, or emotional aggression, but also about scamming and spamming attempts.

We started building our moderation system right from the beginning of product development, but we strengthen it with AI every single day. Our goal is to keep community spaces safe in real time, healthy, and trusted for both end users and platform owners.

Value: a safe, trusted space free from scams, spam, and hatred.

Client impact: critical (if content moderation is poor, the community solution will most likely be removed from the platform).

AI technologies used: proprietary natural language processing models (based on DistilBERT, BERT, RoBERTa), OpenAI API, Gemini API, and Kimi for certain languages.

Development effort: high. Achieving 95%+ precision across major languages requires continuous work from AI engineers and data scientists. Reaching a trusted level for even ~10 languages demands roughly a year, and models need to be fine-tuned for each industry domain.

AI Translation

This is a capability that simply wasn’t possible before AI, at least not at the level of quality required. The value of allowing users who speak different languages to exist in the same social space is hard to overestimate. Thanks to AI translation, it no longer matters what language a person speaks; they can still stay connected with like-minded individuals.

Translation can be applied to both outgoing and incoming messages; we chose the second approach.

Value: enabling interaction between all users, regardless of language.

Client impact: significant for platforms with a global reach.

AI Technologies used: Google Translate, Gemini, plus our proprietary service for caching, orchestration, analytics, and billing.

Development effort: low, if you simply call Google Translate, but this quickly leads to unpredictable and massive costs without proper orchestration. Medium if you build a structured, optimised service around it.

AI Agents

Everyone is talking about agents now; they exist for literally everything. So I’ll tell you about one agent that’s less common.

A few months ago, we introduced the AI Sports Coach, an intelligent agent designed to provide end users of our sports-platform clients with expert-level sports insights. It’s like having your own media source/Google/ChatGPT right inside the host platform, so the user never needs to leave to search for information. They can interact with the AI Coach in both private and public chats.

To ensure trusted results, we built an agent that not only retrieves information but also validates it, while our own services maintain context around the user’s query using both short and long memory.

Value: users get sports-related knowledge without leaving the platform.

Client impact: significant for sports platforms.

AI technologies used: Gemini and OpenAI models plus proprietary services.

Development effort: medium.

Achievement Management

This could be implemented without AI, but with AI, everything becomes much livelier.

We connect to the client’s achievement/gamification/loyalty system (if they have one) via API, and we also maintain our own internal achievements based on triggers in community activity. Right now, we’re transforming our achievement system into an AI-powered one: the model analyses the user profile, another model analyses context, another generates a non-generic congratulatory message, and yet another creates the accompanying image. Every achievement becomes unique.

Value: lively, non-boring, non-repetitive achievements.

Client impact: great for engagement and user stickiness.

AI technologies used: Gemini for understanding the context, OpenAI for image generation.

Development effort: medium

AI Semantic Analysis

This is something we started long ago as a feature, but now we’re rebuilding it into a product.

Our clients have access to a gigantic amount of valuable information willingly shared by end users in integrated public chats. On average, around half of all interactions and communications inside these chats are directly connected to the host platform’s content, products, or overall strategies, making this an exceptionally valuable layer of insight for platform owners. This first-party data helps determine who drives engagement, who is at risk of leaving, which topics create positive or negative sentiment, and countless other insights, like the product and content feedback for the platform owner, which our social layer is integrated with.

This data is far more valuable than scanning external social networks:

• Users rarely discuss your platform on random Facebook pages.
• You can identify only a small portion of your users on external platforms—even in your own Discord or Telegram channels.
• But inside your integrated social layer, you own the data. You just need a powerful tool to understand it.

We vectorise all text information and apply cascades of machine learning algorithms for classification, context recognition, semantic analysis, sentiment analysis, summarisation, etc. Creating a “user fingerprint” and “event fingerprint” is a tremendous amount of work.

Next year, we’ll offer clients a newly built version of this product to help their businesses grow.

Value: actionable insights from user conversations.

Client Impact: significant for medium and large platforms with many users and high chat activity.

AI technologies used: to be disclosed after market launch.

Development effort: high.

Core Principle

To finalise the part about our AI-powered products (we think they are AI-powered, not AI-first), I’ll mention our most important principle—the one I stated at the beginning:

We keep human connections at the heart of our solution.

We sincerely believe that AI tools should empower these connections, not replace them with human-to-AI communication. Do platform owners ask us about this type of replacement, being fed with the market hype?

Yes, they do. But we stand by what creates strategic value for both end users and platform owners—and this persistence absolutely pays off.

AI-First Business

This part will be short.

All our internal routine processes are already AI-first, freeing up precious time for speaking with clients (which we don’t delegate to AI—except for the optional first layer of support), and for inventing new ideas and building strategic concepts.

What we have AI-centred:

• Routine coding (mostly with Claude and Gemini)
• Testing and automated testing
• Prototyping (Claude Code and Google AI Studio)
• Design (Figma Make, OpenAI, Nano Banana)
• Analytics (an internal assistant built on OpenAI, transforming tasks into SQL/Python queries)

Of course, every piece of AI-assisted work is thoroughly double-checked by humans.

What will we have in a month or two? Honestly, I don’t know, and that’s great. We live in an amazing world where we move fast with the flow of the river. We can control our movement, navigation, and focus. Still, the speed of the flow is enormous, and we definitely don’t know what we’ll see around the next bend.

Let’s stay curious yet critical-thinking, hardworking yet open-minded, persistent yet resilient—and we will see what’s next.

Drive your business with AI-powered solutions

If you want to grow your metrics and scale with the power of communities and the help of AI tools, book a demo call with the Watchers product team. We will help you to integrate all the needed tools with the highly developed AI systems.