DEV Community: Antti Kaipila

Scaling SEO Across Languages, Countries and Services: The Real Challenges of Multi-Territory Architecture

Antti Kaipila — Wed, 12 Nov 2025 23:58:57 +0000

Supporting one language is easy.
Supporting one country is manageable.

Supporting multiple countries, multiple languages, and many services — while keeping Google happy — is an engineering problem that almost no tutorial really covers.

This post walks through the real-world challenges of building a multi-lingual, multi-territory, multi-service website that actually ranks, using concrete examples from:

Finland
Sweden
Estonia
Latvia
Lithuania

We’ll cover:

Slugs
Canonicals
Alternate tags (hreflang)
Shared languages across different countries
Authority compounding under one domain
Correct anchor text strategy across regions

1. Why Multi-Territory SEO Is a Different Beast

Most “multi-language SEO” content assumes:

One domain, a few languages, same market.

But if you operate across multiple countries, you don’t just have languages — you have:

Local legislation
Local pricing
Local expectations
Localized search terms
Localized service names
Localized slugs
Shared languages across borders (like RU + EN in the Baltics)

This results in a combinatorial explosion of nearly identical pages.

If you don’t structure them correctly, Google:

picks the wrong canonical
shows the wrong language to users
collapses pages together
splits your authority

A perfect example: home cleaning, which exists in different countries with different names and languages.

2. Slug Strategy: Fully Localized, Not Just Translated

It’s easy to accidentally use English slugs everywhere, like:

/fi-FI/services/home-cleaner
/sv-SE/services/home-cleaner

This destroys local relevance.

Correct approach:

Finland: https://nuuduu.com/fi-FI/services/kotisiivous
Sweden: https://nuuduu.com/sv-SE/services/hemstadning
Estonia: https://nuuduu.com/et-EE/services/kodukoristus
Latvia: https://nuuduu.com/lv-LV/services/majas-uzkopsana
Lithuania: https://nuuduu.com/lt-LT/services/namu-tvarkymas

These are the exact service names real users type into search engines.

3. Canonical: Always Self-Canonical — Never a Global Master Page

The golden rule:

Each localized page must be canonical to itself.

Alternates must always include the canonical URL.

Finland example (3 languages, one territory)

<link rel="canonical" href="https://nuuduu.com/fi-FI/services/kotisiivous">
<link rel="alternate" href="https://nuuduu.com/fi-FI/services/kotisiivous" hreflang="fi-FI">
<link rel="alternate" href="https://nuuduu.com/sv-FI/services/hemstadning" hreflang="sv-FI">
<link rel="alternate" href="https://nuuduu.com/en-FI/services/home-cleaner" hreflang="en-FI">

Sweden example

<link rel="canonical" href="https://nuuduu.com/sv-SE/services/hemstadning">
<link rel="alternate" href="https://nuuduu.com/sv-SE/services/hemstadning" hreflang="sv-SE">
<link rel="alternate" href="https://nuuduu.com/ar-SE/services/tnzyf-almnazl" hreflang="ar-SE">
<link rel="alternate" href="https://nuuduu.com/en-SE/services/home-cleaner" hreflang="en-SE">

The Baltics (shared RU + EN, different main language)

Estonia

<link rel="canonical" href="https://nuuduu.com/et-EE/services/kodukoristus">
<link rel="alternate" href="https://nuuduu.com/et-EE/services/kodukoristus" hreflang="et-EE">
<link rel="alternate" href="https://nuuduu.com/ru-EE/services/uborka-doma" hreflang="ru-EE">
<link rel="alternate" href="https://nuuduu.com/en-EE/services/home-cleaner" hreflang="en-EE">

Latvia

<link rel="canonical" href="https://nuuduu.com/lv-LV/services/majas-uzkopsana">
<link rel="alternate" href="https://nuuduu.com/lv-LV/services/majas-uzkopsana" hreflang="lv-LV">
<link rel="alternate" href="https://nuuduu.com/ru-LV/services/uborka-doma" hreflang="ru-LV">
<link rel="alternate" href="https://nuuduu.com/en-LV/services/home-cleaner" hreflang="en-LV">

Lithuania

<link rel="canonical" href="https://nuuduu.com/lt-LT/services/namu-tvarkymas">
<link rel="alternate" href="https://nuuduu.com/lt-LT/services/namu-tvarkymas" hreflang="lt-LT">
<link rel="alternate" href="https://nuuduu.com/ru-LT/services/uborka-doma" hreflang="ru-LT">
<link rel="alternate" href="https://nuuduu.com/en-LT/services/home-cleaner" hreflang="en-LT">

4. One Domain, Many Countries: Authority Compounding

Subdomains split authority:

fi.example.com  
se.example.com  
ee.example.com

Path-based locales consolidate it:

example.com/fi-FI/...  
example.com/sv-SE/...  
example.com/et-EE/...

This lets strong markets lift weaker ones.

Every backlink benefits the whole ecosystem.

5. Hreflang Must Be a Closed Loop

Each page must:

list all alternates in that territory
include itself as an alternate
use correct country codes
avoid linking to redirects or 404s

This prevents Google from collapsing similar pages or showing the wrong market version.

6. Anchor Text in a Multi-Territory Setup

Below are anchor text examples using “Order home cleaning” in each local language. The anchor link text is a strong signal to Google on what the page is about.

Provided as HTML code and exampled.

🇫🇮 Finland (FI–FI)

HTML

<a href="https://nuuduu.com/fi-FI/services/kotisiivous">Tilaa kotisiivous</a>

Rendered:

Tilaa kotisiivous

🇸🇪 Sweden (SV–SE)

HTML

<a href="https://nuuduu.com/sv-SE/services/hemstadning">Beställ hemstädning</a>

Rendered:

Beställ hemstädning

🇪🇪 Estonia (ET–EE)

HTML

<a href="https://nuuduu.com/et-EE/services/kodukoristus">Telli kodukoristus</a>

Rendered:

Telli kodukoristus

🇱🇻 Latvia (LV–LV)

HTML

<a href="https://nuuduu.com/lv-LV/services/majas-uzkopsana">Pasūtīt mājas uzkopšanu</a>

Rendered:

Pasūtīt mājas uzkopšanu

🇱🇹 Lithuania (LT–LT)

Markdown

< a href="https://nuuduu.com/lt-LT/services/namu-tvarkymas">Užsakyti namų tvarkymą</a>

Rendered:

Užsakyti namų tvarkymą

7. The Payoff When Everything Works Together

With correct slugs, canonicals, alternates, architecture and localized anchor text:

Google shows the correct language in each country
Pages stop competing against each other
Each region ranks independently
Authority compounds globally
Adding new markets becomes predictable and safe

Closing Thoughts

Scaling internationally requires a real SEO-aware routing system:

Localized slugs
Self-canonical pages
Closed-loop alternate tags
Shared authority under one domain
Localized anchor text

Once this foundation exists, expanding to new markets is safe, predictable and technically clean.

Winning the SEO Game

Antti Kaipila — Fri, 07 Nov 2025 15:57:19 +0000

The challenge

My company's business model — earning just €1 + €1 per order — delivers incredibly affordable prices for customers and fair earnings for partners. But it comes with one big challenge: marketing costs.

Take Google Ads, for example. One click costs €1–2, and it easily takes 20–30 clicks to get a single order.

The math doesn’t work.

The Solution?

Simply win the SEO game and rise to the top of Google search results organically. When your page appears on the first search page and your prices are unbeatable, the outcome is clear: success.

But how do you succeed when everyone else is trying to do the same?

The good news is that the competition largely consists of small, non-technical companies using the same off-the-shelf templates and marketing agency-built websites.

Nuuduu is different — we are a technology company with no employees but a network of independent professionals.

From these differences comes our SEO advantage.

SEO Principles

Search engine optimization is divided into two main parts:

1. Technical SEO

Keyword placement, meta descriptions, site structure, Schema.org tags, OG tags, canonical links, multilingual versions, loading speed, consistent rendering — the whole classic list.

You have to hit the nail on the head.

Our advantage is that we control the entire platform ourselves. We are tech nerds.

A competitor like www.joe-the-plumber.com is ultimately a prisoner of whatever CMS platform their agency or cousin chose.

Does it support all the required features? Does the creator know how to optimize them? Sometimes yes, often no.

For us, this is a religion.

We dynamically generate content from the database into JSON-LD tags, provide search bots with proper OG tags, create unique content using LLMs, automatically process partner images, use server-side rendering, test, polish, and repeat.

Two months of intensive optimization — hundreds of documents read and dozens of test rounds.

Few competitors invest anywhere near that much.

Knowledge about SEO is everywhere, but much of it is, frankly, nonsense.

SEO is an industry where everyone wants to buy something — and so someone is always ready to sell a “magic bullet.”

In reality, only Google knows what truly works inside its algorithm — and they’re not telling.

Luckily, there’s one real method: look at who already ranks at the top, analyze, and learn.

Hard work, but necessary.

2. Backlinks

This is the hardest part. Technical SEO is in your own hands, but backlinks depend on others.

There’s no shortage of advice:

“Create valuable content that others want to link to.”

“Pay an SEO guru who will handle it for you.”

The first is difficult and time-consuming; the second is expensive and risky. These SEO gurus don't have any magic potion but instead often rely on link farms and Google punishes those — and the guru will have vanished by the time the penalty hits.

Five-Step Backlink Strategy

I decided to do this myself to retain full control and avoid landing in a link farm.

The following five strategies support each other and form a complete system:

1. Microsite Network for Key Keywords

Thanks to LLM tools, this is no longer difficult. Ideas are generated with ChatGPT, and each site gets its own domain and small but growing PageRank, which it shares back to Nuuduu.

Examples:

All of these sites link back to the main site. As they age and collect links their page rank increases and feeds it back to the main site and these sites cross link between them selves.

Cost: about €20 and two days of work.

2. Open Source Visibility

By publishing small open-source tools on GitHub and Bitbucket, you gain visibility on high-PageRank platforms:

Create something that you can publish out in the wild and share it on well known platforms. These create natural backlinks among technical audiences.

3. Blogs and Guest Articles

I founded several free blogs for different service types, plus this one about building a startup, I guess they call this a PBN or a Private Blog Network:

In addition, I publish articles on Medium, Vocal.media, and Uusi Suomi Puheenvuoro. Finding sites that allow dofollow links is getting harder and harder. One trick on medium is to publish your article on a blog you control and import it to medium providing you with a canonical link from medium back to the blog you control.

This way, I get natural backlinks in the “About the author” sections — a fair trade: content for them, backlinks for me.

4. Automatic Partner Microsite Network

Since Nuuduu’s partners are independent entrepreneurs, they want visibility.

That’s why I created the nuuduu.partners domain and a wildcard system that automatically generates a personal microsite for every partner with a a unique slog of their first name as the host part of the domain name.

Descriptions, links, and content are AI-generated during registration. I also enrich the pages by cross delivering relevant content from the blog on my main site (described below)

The result: hundreds of unique, profession-specific pages, on separate sub domains, all linking back to nuuduu.com.

BOOM.

5. Nuuduu BOOST Campaign

Instead of spending money on Google Ads, we use it to reward partners.

Each partner gets their own page under the nuuduu.com domain, showing only their services. They can share the link on social media or add an Order Button to their own website.

Every month, the best-performing partners are rewarded.

This creates natural social media shares and dozens — later thousands — of genuine backlinks to nuuduu.com.

Exactly what Google values.

The international structure further strengthens the whole system:

A Finnish plumber and a Lithuanian dog-sitter benefit each other, as all traffic flows to the same domain.

Other relevance signals Google uses

Another signal Google uses is of course what people search for, what they click and how log the linger on the site. So how can we manufacture this kind of behaviour?

We are about to start running campaign that encourages people to search for gift cards hidden on our main site. We publish a set of two clues on our Facebook page 2-3 times per week and users have to come up with a search term and find out page using this search term on Google, then find the hidden gift card on the site.

This leads to all sort of niceties.

Facebook page gets followers because that's there we publish the treasure hunt clues.
Google sees search traffic that lingers.
Gifts cards lead to orders.

The gift cards stay active an discoverable for as long as someone finds and uses them.

Rich blog content on the main site

I created an internal tool for me to relatively easily product rich blog content on to my own site. The system uses LLM to write the articles but the quality and relevance of the articles is the magic sauce here.

I control the whole platform so what I do is I created a list of about 10 different seed topics like "5 being tips", "best time to buy" and combine these with the 30 or so different service categories we offer on the platform.

I give the LLM background information by feeding it service descriptions, our customer manual and terms and conditions.

The output comes out not as general blaa blaa but as content relevant to the services offered on the platform and the way we offer the services.

After building this system I am able to generate topic relevant content with a few clicks. I could just mass generate tons of these articles but I still like to actually read and tweak them a little, set them up with pictures and so on.

Summary

The SEO game is not a single trick but a complete system:

technical perfection + content consistency + backlink ecosystem.

We’ve combined automation, AI, and a partner network into a natural SEO engine that grows with every new partner and article.

This is a long-term game — and we’ve come to win it.

AI Agent Count — A New Metric for Measuring AI Adoption

Antti Kaipila — Thu, 06 Nov 2025 07:06:50 +0000

Metric for Measuring AI Adoption

Companies love to say they “use AI” — but what does that actually mean? Owning a ChatGPT subscription doesn’t make you an AI-driven business. It’s time for a clearer measure: the AI Agent Count — a concrete metric showing how deeply AI has been integrated into a company’s daily operations.

What is an AI Agent?

An AI agent can be thought of as a digital employee, built around three core components:

Large Language Model (LLM) — such as GPT-5, capable of understanding and producing natural language.
System Prompt — a precise description of the agent’s role and behavior: “You are a customer support representative,” or “You are an accounting assistant.”
Tools — access to APIs, databases, and other systems that allow the agent to actually do something.

Each agent is designed to handle one specific business process — and when combined, they form a kind of AI workforce inside the company.

One Agent, One Process

You can roughly think of one agent corresponding to one business process:
Customer support? One agent.
Lead qualification? Another agent.
Invoicing and reconciliation? A third one.
As companies deploy more of these specialized agents, they start to build an autonomous operational layer — a digital organization that scales without salaries, shifts, or onboarding time.

AI Agent Count: A Practical Measure of AI Maturity

The AI Agent Count tells us how deeply AI has been operationalized. It’s a simple but powerful proxy for organizational AI maturity.

Typical AI Agent Count

0 Not understood AI is still seen as futuristic or irrelevant.

1-5 Light adoption A few external tools with AI features are used, like Copilot or ChatGPT.

5-20 Moderate adoptio*n Some automations and internal integrations are in place.

>20 AI-native company Dozens of agents handle entire workflows — hiring, support, marketing, analytics, etc.

The agent count could soon be more meaningful than headcount — a leading indicator of how automated and adaptive an organization really is.

The scaling vector of AI agents compared to headcount is infinite. One agent for one purpose once perfected can scale as the business grows.

The Real Challenge: Scattered Data

The biggest barrier to AI adoption isn’t ambition — it’s data fragmentation.

Company data is often scattered across multiple disconnected systems: CRM, accounting, eCommerce, marketing automation, project management

AI agents can’t perform effectively without unified, structured access to information. This makes data integration and connectivity the true bottleneck of enterprise AI adoption.

With disconnected business tools a new kind of void starts forming in the center of companies AI org chart.

A New Software Category: The Agent Management Platform

This challenge has already given rise to a new kind of software: the Agent Management Platform — a control layer that connects, manages, and governs AI agents inside an organization.

Such a platform allows companies to:

Connect agents to data sources and APIs
Manage and version their system prompts
Monitor and label agent responses — tracking performance, reliability, and reasoning quality
Control access rights and audit logs
Coordinate workloads between multiple agents

If an ERP manages human employees, an Agent Management Platform manages digital ones.

It becomes, in essence, the AI brain of the company.
Conclusion

The real AI revolution won’t come from isolated pilot projects or shiny demos — it will come from companies building networks of specialised agents that automate full business processes.

In the years ahead, a company’s AI Agent Count may become one of the most telling metrics of its technological maturity — and for those who grasp it early, perhaps the most powerful competitive edge of the decade.

About the author
Antti Kaipila is the founder and CEO of Nuuduu, a next-generation direct service platform designed from the ground up for AI agents. Nuuduu leverages more than 40 different agents to automate its operations across Europe.

Detached agent chains that work like people do

Antti Kaipila — Wed, 05 Nov 2025 22:44:24 +0000

Most AI deployments wait for a prompt. Real operations do not. A manager defines a goal, delegates work, and people go collect facts, make decisions, and return results. What we need is AI agents that have an intent and initiate conversations with a goal.

The system below mirrors that familiar pattern with task templates, action chains, a watchdog that moves work forward, and scout agents that initiate conversations. It is simple to describe, strong in practice, and it scales. Everything described here is doable with existing technologies.

Let's look at the main components that make up such a system:

System Components

1) Task template

A task template is the starting point. It contains an action chain plus basic metadata like name and description. Task Templates are executed by an event, on a schedule, when criteria are met, or on demand by a human.
When a template is executed, it becomes a task instance with its own state and audit trail.

2) Action chain

An action chain is a list of actions that run in order inside a task. Each action has an embedded AI agent and the watchdog oversees execution. Actions come in a few types:

Scout Agents. Initiates a conversation with one or more counterparties and collects structured data. Scouts are detached by design. People can ignore a message or take time to answer, so the system must tolerate latency and silence. Scout agents are implemented by temporarily replacing the normal support agent in the support chat with it self and initiating a conversation with a pre-defined templatable greeting.
Create PDF. Uses a Word template and a simple JSON [key name]/[description] schema to generate a finished document as an artifact.
Compose and send email. Writes an email from instructions and can attach artifacts created earlier in the chain.
Perform other actions. Talk to an API, make a change in the database, post to a queue, or anything else needed. The sky is the limit here.

Every action in the task template and subsequently in the executed task instance includes:

Instructions. Plain language guidance that becomes part of the system prompt. For example, write an email about subject X or collect data about subject Y.
Input. The output of the previous action or fields fetched from the task data store.
Output. A structured payload for the next action and for the task data store.
Timeout policy. What to do if the action does not complete in time. Abort the task, alert a human, or ignore and continue.

The task has a general data store. Any action can write to it. Subsequent actions can read from it. This keeps the chain loosely coupled and easy to extend. LLMs are fantastically adopted to reading loosely structured data and understanding it’s context.

3) Watchdog

The watchdog advances the chain. It starts actions, monitors completion, enforces timeouts, and dispatches the next step. If an action stalls, the watchdog retries, switches channel, escalates to a human, or bails out per policy. It also keeps the log, timestamps each event, and stores artifacts. Think of it as the manager who checks progress and pushes the work forward.

4) Document templates

A document template is a Word file with {{tags}} and a schema for expected data. Templates can attach to a PDF action to produce a signed or branded document. They can also attach to a scout action so the scout knows exactly which fields to collect. The schema is the contract. It gives the system a way to validate outputs before side effects happen.

Three simple use cases

1) Collect pizza orders from the whole office

Let's start off with something simple to illustrate the functionality of each component in the chain.

The task dispatches scouts to everyone on the team.
Each scout opens a chat, gathers order, size, and special notes, then writes the result into the task data store.
When all scouts complete or hit a timeout, the next action composes a single email to the pizzeria and sends it. Instructions for the final action that composes an email, has access to the task instance data store and an agent doing the job in the task template can be as simple as:

Write a pizza order to some,local@pizza.joint using the available requests in the data store, deliver the order to our office at Some Street 123 and invoice Acme Inc account number #123 or bring a card terminal.
The role of the timeout and bail out functions here would be that if a scout fails to collect the pizza wishes from a team member, ignore it and send the order out at 11:30 anyway.

2) Help a customer file an insurance claim

Next let's examine a different flow and another way to trigger an execution of a task instance.

In support chat, the customer tells the default agent that they need help filing an insurance claim.
The default agent selects the “insurance claim” task template and it is executed.
The template includes a Word document and a schema that lists incident fields.
A scout opens a conversation with the customer, collects the fields, and writes the JSON to the task data store and outputs it as the input to the next action.
A PDF action fills the Word template and outputs a final document.
An email action writes a short cover note and sends the PDF to the insurer mailbox.

Task templates automatically becoming tools that your customer default support AI agent can dispatch is the revolutionary part here.

How many different jobs performed by some human in the office are essentially: Go talk to someone about something, write a document about it and send it to someone? Think about it, quite a lot.
Now just by defining the necessary task templates in an admin interface for each of these and selecting them to become available as tools lets you automate a lot of processes.

3) Update all customer phone numbers

The task dispatches scouts to every customer.
Each scout asks for the current phone number and returns a structured payload.
A database action validates the format and updates the record. Errors are logged for human review.

This example is powerful because the scout agents can all run in parallel. Updating the customer database can be done in 30 minutes for all customers compared to humans doing the work in a series.
Timeout and bail out parameters here can be set to like a day and alerts generated for scouts that were not able to complete their tasks.

Why full platform control matters

Scout agents only work when you control the full system where the chain runs. You need to open and manage chat threads with a separate chat agent handling them, attach forms and files, tag each message with a correlation id, and capture consent. You need to validate outputs against a schema before taking any action. You need clean audit logs. Stitching this across many unrelated tools is brittle.

This is why platforms like Salesforce, Odoo or others are natural homes for detached chains. Both combine workflow, data, and communications. You can create records, run server actions, post to chat, and store artifacts in one place. You can also add human review steps that see the full context. If your organization cannot use these, you can build the ecosystem in house. The key is ownership of channels and data paths.

Chat control is critical. A scout that cannot open a thread, send a reminder, switch channel, or record consent will fail in real life. The watchdog needs reliable signals from the chat layer to decide whether to retry, escalate, or move on.

How this mirrors human work

This model behaves like a team. A manager defines a mission. Scouts go talk to people, collect facts, and return structured notes. Specialists draft a document or update a system. The watchdog is like a manager. It keeps time, nudges when needed, and escalates when something blocks. The result is predictable progress from intent to artifact without long back and forth.

The difference is that the system never forgets a field, always writes in the right format, and enforces time limits. It also produces an audit trail that shows who said what and when, what data was used, and which decision happened next. And all this can run in parallel and scale on demand.

Implementation checklist

Pick a platform where you can control chat, data, and actions in one place.
Define a task template with a short action chain that solves one narrow problem.
Write the scout instructions and the output schema. Keep it concrete.
Add a Word template with {{tags}} for any document you need to produce.
Implement timeout policies. Soft timeout for a nudge. Hard timeout for failover.
Validate every output against the schema before any side effect.
Log transcripts, artifacts, and decisions for audit and training.
Track a few simple metrics. Completion time, first try rate, timeout rate, and cost per completed task.

The power factor is the scout

The scout is what turns a static workflow into a living one. It is the agent that can initiate conversations and create new information by talking to people. Everything else is plumbing. Once you can send respectful, purpose built scouts that collect exactly the fields you need, the rest of the chain becomes routine. The watchdog moves the work along. The PDF has the right data. The database stays clean. Customers get what they need without long calls.

That is how detached agent chains become practical. They behave like people, run inside systems you control, and leave artifacts other systems can trust.

Author bio: Antti Kaipila is the founder of Nuuduu, a direct service platform that uses AI to automate service delivery.