LX Series Deep Dive: Brand as a Platform - How the 6 Sovereignties Actually Get Delivered

Nearly every AI module solution on the market that claims to be "customizable" stops at swapping the logo and changing the UI colors.

What the LX Series sets out to do is hand over six complete areas of technical sovereignty, one at a time, to the brand owner.

1. Brand Sovereignty Anxiety: The Most Expensive Anxiety of 2026

It is 2026 and a toy brand has just sold 100,000 AI toys. What happens next?

Ending one (the common one): the app carries the solution provider's logo, user data runs in the solution provider's cloud, and subscription revenue is split 3:7 or 4:6 with that provider. The more you sell, the further you drift from your own users. Who owns the online activity and retention data across those 100,000 units? Nobody knows. Who owns the users' conversation preferences and content consumption data? Nobody knows. Who owns the channel that reaches them when you push a new story for the holidays? Nobody knows.

Ending two (the rare one): the app carries your logo, user data accumulates in your own back end, and 100% of the subscription revenue is yours. The more you sell, the thicker your user asset gets: 100,000 online users are a gold mine you can push content to with precision, use to iterate the product, and even use to incubate new SKUs.

The difference is not the hardware, because hardware converges sooner or later. The difference is control.

These six sovereigns of the LX Series are every technical capability the second ending requires, packaged up and delivered into the brand owner's hands.

2. Start with the Architecture: Device, Cloud and Application - Three Layers, Each One a Brand Asset

The LX Series four-layer architecture: application layer → cloud layer → device layer → AI layer

In a conventional module cloud platform solution, the architecture puts "platform first, brand behind": the brand owner is just one customer of the platform, and the data, the rules and the iteration cadence of every layer are defined by the platform.

The LX Series turns that around: brand first, platform behind. Nablai supplies the foundation - the module, AMS, the model framework and the app SDK - but the specific configuration of each layer, the operating rules and data ownership all belong to the brand owner.

Understand that architectural difference and the six sovereigns below will make sense.

3. Sovereignty 1: An Independent Account System - You Decide Who the User Belongs To

Deliverables

The brand owner runs its own user registration and login system (phone number, WeChat, Apple ID and other login channels)

Core user information, device binding relationships, usage records and AI conversation logs are all retained in full

The account system is deployed independently inside the brand owner's AMS console, physically isolated from Nablai's own technology platform

What does this give the brand owner?

(1) User assets that compound

A child who talks to a toy for 20 minutes a day generates 10 hours of real interaction data in a month. That data covers:

which topics, stories and facts the child likes

the child's language habits and emotional patterns

the household's spending bracket and the topics parents are sensitive about

peak conversation hours and how often the device gets used

Sitting in the brand owner's own database, this data is a mine worth working:

precise segmentation when you push a new story or a new feature

real usage data to back the next generation of the product

a user-profile foundation when you move into value-added services

(2) Cross-device user recognition

An independent account system means a user's multiple devices - the plush toy in the living room, the story player by the child's bed, the small speaker they take out with them - can all be bound to one user ID and form a single unified profile. With module cloud platform solutions, the account systems behind different devices do not interoperate, and that is the single biggest obstacle to building user LTV (Life Time Value).

(3) Full control of the membership system

Subscription membership, points, growth levels and paid content tiers are all defined by the brand owner. The AMS console on the LX Series supports flexible membership configuration: monthly / quarterly / annual / trial plans, auto-renewal, tiered pricing and member-only features.

Technical detail: how is the account system built?

The LX Series account system is delivered white-box:

Independent account system delivery checklist: the four layers of database, login service, security and user operations

Real scenarios

The first thing one designer toy brand did after adopting LX was to use the app registration rate as a proxy for product reputation. For every new SKU launch they look at the ratio of registered accounts to units sold: a high registration rate means people really use it and retention is high; a low rate means the product failed to hold their attention. With a module cloud platform solution that data is simply unobtainable, because every user is a platform account.

4. Sovereignty 2: Your Own Branded App - Your UI, Your Accounts, Your Store Listing

Deliverables

App UI, UX, icons, splash screen and motion, all built to the brand owner's design files

The app is published under the brand owner's own developer account (Apple Developer Account / Google Play Console / domestic app stores)

App feature modules (device binding, provisioning, conversation history, subscriptions, settings) are built on the LX SDK

What does this give the brand owner?

(1) The app store listing is yours

This is the sovereignty that most often gets overlooked. Plenty of brand owners publish their app under the solution provider's developer account, which means:

the risk of the app being pulled or the account being banned sits with the solution provider's account

store reviews and ratings are outside your control

the app name and icon are managed by the solution provider

With the LX Series brand-owned app, ownership of the app store listing is 100% the brand owner's. Even if the brand owner ends its relationship with Nablai one day, the app and the users stay in your hands.

(2) Brand consistency

The LX SDK ships with a complete UI component library and interaction framework, but the visual language is defined by the brand owner:

brand colors, typefaces and icon style

emotional design for onboarding, empty states and error pages

micro-interaction details (button feedback, transition motion, loading animations)

We delivered an app for a "panda IP" AI toy where the entire UI followed a gentle, soothing aesthetic, and even the error page was an animation of the panda rubbing its eyes. No OEM white-label app can deliver that level of detail.

(3) Freedom to extend features

Brand owners can build brand-specific features of their own on top of the LX SDK:

IP collaboration campaign pages (a limited-time Kakalong x animation-IP campaign)

a membership growth-points system (users earn 10 points for every hour of conversation)

a content store (virtual goods sold alongside subscriptions)

community features (parent community, user-generated content)

Technical detail: how is the app delivered?

The LX Series offers two app delivery models:

Model	Use cases	Delivery lead time
White-box SDK	Brand owners with their own app development team	1-2 weeks to integrate
Custom development	Brand owners without an app team	45 days on average

Under both models, ownership of the finished app belongs to the brand owner: source code, developer account and store listing are all handed over.

5. Sovereignty 3: A Custom LLM - You Define the AI Persona, and Accountability Is Clear

Deliverables

LLM integration: DeepSeek, Qwen, ERNIE Bot, ChatGLM, Llama and other mainstream models are all supported

Model fine-tuning framework: build a dedicated persona from conversation samples supplied by the brand owner

Safety filtering: an on-device sensitive-word library plus cloud-side content moderation, a double safeguard

MCP tool-calling protocol: the model can plug into external capabilities such as weather, music and calculation

What does this give the brand owner?

(1) An exclusive AI persona

Custom LLM fine-tuning workflow: samples → annotation → LoRA fine-tuning → A/B testing → independent deployment

One case we ran involved a dinosaur-IP toy. After fine-tuning, the model spoke like "Wow, you're amazing, kid. Want to come exploring with me?", where the general-purpose model answered "Sure, okay". That difference in tone is exactly what makes an IP recognizable.

(2) Freedom to choose the model

Different models are strong in different areas:

DeepSeek: exceptionally good Chinese conversation and solid logical reasoning

Qwen: a rich Chinese training corpus and smooth multi-turn conversation

ERNIE Bot: a strong domestic compliance posture

Open-source Llama / Qwen: friendly to private deployment

Hybrid routing architecture: use a lightweight model for everyday chat to hold costs down, and route deeper conversations to a stronger model

The LX Series does not lock you into a model. Brand owners pick the base model that best fits their product positioning, budget and compliance requirements.

(3) Clear lines of accountability for safety

AI content safety is the most sensitive compliance topic of 2026. The LX Series uses a two-layer safeguard:

Layer	Mechanism	Role
On-device	Local sensitive-word library + keyword-triggered blocking	No network dependency, zero-latency response
Cloud	The LLM's built-in safety filters + the brand owner's own moderation rules	Compliance checks in complex contexts

The key point: responsibility for content safety sits with the brand owner,not with the solution provider. The LX Series provides the tools and the mechanisms, but responsibility for designing conversation content and for compliance review rests clearly with the brand owner. That gets written into the contract, which avoids arguments later.

Technical detail: wiring up the MCP protocol

MCP tool-calling architecture: custom LLM ↔ MCP protocol ↔ tool server (weather / music / calculator / translation / stories)

Brand owners can connect any MCP-compatible tool service their product needs, so the capabilities of an AI toy keep evolving.

6. Sovereignty 4: Your Own AMS Console - Devices, Users and Content, All Under Your Control

Deliverables

AMS (AI Management System) is deployed independently into the brand owner's own cloud environment (public cloud / private cloud / hybrid cloud)

Device management: activation, binding, unbinding, status monitoring and OTA updates

User operations: account management, behavior analytics, user segmentation and push channels

Content operations: story library management, knowledge base management, conversation template configuration and holiday campaign configuration

What does this give the brand owner?

(1) Device lifecycle management

With 100,000 AI toys online, no two units are in the same state:

which devices are online and which are offline

which devices are low on battery and which have a faulty microphone

which devices upgraded successfully and which failed and need a rollback

which devices produced conversation content that tripped a sensitive word

The AMS console provides a complete device monitoring dashboard, so hardware after-sales issues are settled with data instead of manual troubleshooting.

(2) Your own data dashboards

User activity data, content consumption data, conversation topic data and retention curves: the AMS console provides a complete set of dashboards:

Dashboard	Metrics
User dashboard	DAU / WAU / MAU, new registrations, retention curves, churn alerts
Device dashboard	online rate, activation rate, error rate, OTA success rate
Content dashboard	top stories played, conversation topic distribution, sensitive-word trigger statistics
Operations dashboard	push delivery rate, subscription conversion rate, membership renewal rate

With a module cloud platform solution you either cannot get this data at all, or you only get an anonymized aggregate, or the platform deletes it after a few years under its own retention rules. With the LX Series the data is 100% the brand owner's: exportable and retainable indefinitely.

(3) Your own content operations

The brand owner's operations team can do all of the following on its own:

publish new stories (with no solution provider involvement)

configure holiday campaigns (Spring Festival, Christmas and Children's Day specials)

update the knowledge base (product changes, brand campaign sync)

adjust push strategy (user segmentation, send windows, A/B tests)

Technical detail: how is AMS deployed?

Deployment model	Best for	Security level
Public cloud SaaS	Small and mid-sized brands, fast validation	High
Brand private cloud	Large brands, data-sensitive industries	Very high
Hybrid cloud	Balancing security and cost	High

The Dongxiaobao project runs on a brand private cloud: insurance data never leaves the client's private environment, which satisfies the compliance requirements of the financial industry. The capability envelope the LX platform supports holds up just as well in cross-industry scenarios.

7. Sovereignty 5: On-Device Capability - Works Without a Network, and Sensitive Words Always Get Blocked

Deliverables

Local TF card playback (LX-F / S editions): offline stories, offline music, offline knowledge content

On-device VAD (voice activity detection), AEC (acoustic echo cancellation) and KWS (keyword spotting)

On-device sensitive-word blocking: a local word library plus keyword triggers

On-device caching: a temporary conversation buffer that syncs once the network comes back

What does this give the brand owner?

(1) Usable on weak networks or with no network at all

Children use toys in all kinds of situations:

in the car (poor signal)

on a plane (no signal)

in remote mountain areas (weak signal)

traveling abroad (roaming is expensive)

parents who would rather their child was not online at all (privacy concerns)

The on-device capabilities of the LX Series keep the core functions working with no network at all: wake word, conversation templates, TF card story playback and sensitive-word blocking all run locally.

(2) On-device sensitive-word blocking is the compliance floor

Why does sensitive-word blocking have to run on the device?

On-device	Cloud
Zero-latency response, blocked in milliseconds	200-1000 ms of network latency
No network dependency: it still blocks when offline	Stops working the moment the connection drops
Blocked content is never uploaded, so privacy is protected	Content has to be sent to the cloud to be judged
Blocking logic is controlled locally and customized by the brand owner	Depends on the model vendor's rules

The Law on the Protection of Minors and the Provisions on Cyber Protection of Children's Personal Information set explicit requirements for sensitive-word blocking in children's conversations. On-device blocking is the compliance floor, not an option.

(3) A device-cloud architecture

Device-cloud architecture: child speaks → VAD → KWS → sensitive-word blocking → cloud / TTS → cache

This architecture keeps the product usable on weak networks without giving up the intelligent cloud experience. It is the mainstream technical path for AI toys in 2026.

8. Sovereignty 6: Open Extension (MCP) - AI Capabilities That Keep Evolving, Without Vendor Lock-In

Deliverables

Native support for MCP (Model Context Protocol)

A standardized tool-calling interface

Existing tool library: weather, music, calculator, translation, stories and knowledge lookup

Documentation for connecting the brand owner's own in-house tools

What does this give the brand owner?

(1) Capabilities that keep expanding

The AI toy capability everyone wanted in 2024 was voice conversation. In 2025 it was recognizing the world around you. In 2026 it is emotion recognition plus multimodal interaction. What will 2027 bring? Nobody knows.

MCP on the LX Series lets an AI toy's capabilities keep expanding: no device swap, no module swap, just connect a new tool in the cloud:

Timing	New capability	MCP tool
2026 Q3	Weather, news	QWeather, news APIs
2026 Q4	Music playback	QQ Music, NetEase Cloud Music
2027 Q1	Photo recognition	In-house vision tool
2027 Q2	Multilingual conversation	Translation API
2027 Q3	Personalized recommendations	In-house recommendation engine

(2) No lock-in to a single model

Multi-model intelligent routing: everyday → DeepSeek, deep reasoning → GPT-4o, emotional → in-house model, creative → Claude

The routing logic is defined by the brand owner and can be tuned against cost, performance and brand tone.

(3) The brand owner's own tools

In the Dongxiaobao case we connected an insurance-industry "compliance engine" as an MCP tool. It was built in-house by the brand, and once connected it became the AI assistant's compliance checking capability.

The MCP extension capability of the LX Series lets a brand owner's own know-how feed back into the AI toy. That is something a module cloud platform solution simply cannot do.

9. Cross-Industry Validation: What Does the Dongxiaobao Case Prove?

Dongxiaobao is an AI assistant Nablai built for the insurance brokerage industry on the LX platform. The point of this case is not insurance itself; it is the proof that the six sovereigns of the LX platform are a capability foundation that can be reused across industries.

The compliance challenges Dongxiaobao faced

Challenge	How the LX platform solves it
Insurance data must not leave the client's private environment	Sovereignty 4: AMS deployed independently into the brand owner's private cloud
A knowledge base of tens of thousands of insurance products	Sovereignty 3: a custom LLM plus deep training on a domain knowledge base
Mis-selling and compliance risk	Sovereignty 3 + Sovereignty 5: a cloud compliance engine plus on-device sensitive-word blocking
Brokers face a high barrier to entry	Sovereignty 3: a custom LLM shortens the training ramp for new hires

Actual business results

Product lookup efficiency: up 80%

Proposal generation time: 2 hours → 5 minutes

New-hire training period: cut by 50%

Compliance review coverage: 100% (real-time and full-volume vs. manual sampling)

What this case means for toy brands

One set of technical sovereigns reused across industries: insurance industry ↔ LX platform ↔ toy industry

If the LX platform can carry an industry as compliance-heavy and specialized as insurance, the toy industry is well within reach.

10. So How Long Does It Actually Take to Deliver These 6 Sovereignties?

A question many brand owners ask: the six sovereigns sound great, but how long does delivery really take, and what do we have to put in?

The answer is 45 days on average to the first batch of functionality, and 3-4 months for the full delivery. Many stages run in parallel: account deployment and app development at the same time, model fine-tuning and module integration testing at the same time.

Standard delivery timeline (parallel mode)

Stage	Duration	Parallel notes	What the brand owner provides
Requirements gathering + solution design	2-3 weeks	Stage 1	Product + operations leads
Account system + AMS console	2-3 weeks	Stage 2
Three tracks in parallel	Cloud environment + test accounts
App development (LX SDK)	45 days on average	UI/UX designs + app store account
LLM fine-tuning	3-4 weeks	Conversation samples + persona definition
Module integration testing + gray release	3-4 weeks	Stage 3	Toy ID + beta users
Mass production	-	Wrap-up	Production handoff

Total timeline: roughly 3-4 months (from kickoff to mass production, with three stages running in parallel)

A brand owner with its own app team can compress that to 2-3 months. Even with a fully custom build, the parallel model keeps delivery inside 4-5 months.

The key insight: the six sovereigns are not a serial pipeline. Account deployment, app development and model fine-tuning all move forward at once, and the bottleneck is app development (45 days on average), not anything else. That is why "45 days on average" is the core delivery commitment of the LX Series.

Recommended resource commitment

Brand owner resource	Commitment level	Description
Product lead	50% x full duration	Owns product definition and launch cadence
Operations lead	30% x full duration	User operations, content operations, data analysis
App development (optional)	100% x 8 weeks	If you go with the white-box SDK model
Legal / compliance	Part-time	Privacy policy, user agreement, compliance review

Which brand owners is this right for?

A good fit	Not a good fit
Brands that want to own the user relationship themselves	Brands that only want a one-off hardware sale
Brands with a distinct IP tone that need a differentiated AI persona	White-label players with no clear IP positioning
Growth-stage brands selling 10,000+ units a month	Brands still testing the water with unstable volumes
Running global expansion and a domestic business at the same time	Single channel, single market
Long-term players with a 3-year-plus plan	Short-term opportunists who will not invest in operations

11. In Closing: These 6 Sovereignties Are Not a Technology Checklist, They Are a Choice of Business Model

Back to the question we opened with: once you have sold 100,000 AI toys, have you accumulated a user asset, or have you built someone else's business?

The answer does not depend on hardware specs. Anyone can buy an ESP32, microphone array pricing is transparent, and motor driver designs are broadly similar. The answer depends on control:

Whose database does the user data accumulate in?

Who defines the AI persona?

Whose account does the app store listing belong to?

What ratio is the subscription revenue split on?

Who drives content operations?

What the six sovereigns of the LX Series set out to do is write the answers to those five questions into the brand owner's balance sheet.

We are not selling you a module component. We are helping you build an AI capability castle of your own: the walls, the gate and every brick inside it are yours.

The most direct way to understand how the six sovereigns of the LX Series actually get delivered is to look at a real client's full implementation. Dongxiaobao is one of the most complete cases we have done: work backwards from insurance to the toy industry and you will find the underlying logic is exactly the same.

*Further reading:*

AI Toy Module Selection Guide: LX Series vs TY Series

LX Series: Brand as a Platform

Case study: How Dongxiaobao built an insurance AI assistant on the LX platform