Problem & Opportunity
Vietnamese office workers are a large, underserved market for AI productivity tools.
88%
Nominal AI awareness
Urban office workers report using AI - Microsoft WTI 2024
19%
Actual daily use
Urban Vietnam - WIN World AI Index 2026
15pp
Personal vs. work gap
53% personal, 38% at work - WIN Index 2026
70%
Self-provision AI tools
Rather than wait for company rollout - Microsoft 2024
Two Primary Barriers
Barrier 1 - Pricing mismatch
Monthly subscriptions don't fit Vietnamese usage patterns
$20/month for ChatGPT does not fit the irregular, bursty usage patterns of most Vietnamese office workers. Grey market evidence: ChatGPT Plus resold at 89K VND/month (Threads) and 40K VND/month (TikTok). Entire companies share one account.
Barrier 2 - Skill barrier (Prompt Barrier)
Most users cannot extract good results from open-ended chat AI
Users don't know how to prompt effectively. "Can Claude really do this without knowing how to code?" - 404 reactions in Claude Vietnam Facebook group. Users publicly share prompts in communities - highest-engagement content in Vietnamese AI groups. A clear signal of unmet need.
Additional Validated Signals
- Top unmet needs - explicitly named: Document processing (PDF to Excel/Word), format-preserving translation, and meeting minutes generation are the top-cited unmet needs across all research platforms.
- Claude is the preferred model but inaccessible: Users acknowledge Claude as the best-quality model for writing tasks but direct access is blocked by price and account instability (bans, token limits, degrading quality).
- Work AI lags personal AI by 15pp: Vietnamese workers use AI more for personal tasks than work - driven by perceived complexity and lack of company-approved tools.
Assumptions to Validate (Research Phase 2)
Open hypotheses
- H1 - Whether the prompt barrier, not lack of motivation, is the primary reason office workers do not use AI
- H2 - Whether users will pay per task in VND if output quality feels "worth it" compared to a monthly subscription
- H6 - Whether document-processing Skills (PDF, Excel, meeting minutes) will generate the highest conversion rates
High Level Approach
Klerio is positioned as a Solution-first, Skill-based AI tool - directly contrasting with the chat-first model of all major international competitors. Users do not write prompts; they describe a problem and Klerio returns a ready-to-use output through a guided form-based workflow.
Core Workflow
User describes a task or problem
↓
System identifies and suggests the right Skill
↓
System collects inputs via guided form
↓
Returns ready-to-use output
↓
User lightly edits / copies / uses immediately
|
Chat-first (ChatGPT, Claude.ai, etc.) |
Klerio - Solution-first / Skill-based |
| User input | Free-form prompts / requests | Problem description + guided form |
| Skill required | Prompt engineering | None |
| Output | Suggestions / instructions / drafts needing heavy editing | Ready-to-use, immediately usable |
| Consistency | Depends on prompt quality | Stable, defined standard per Skill |
| Pricing | Monthly subscription ($20/month) | Pay-per-task in VND |
User Narratives
Huong
TA Specialist - Hanoi
Each week, Huong needs to draft 2-3 JDs and several interview emails. She tried ChatGPT but felt "you have to know how to ask to get good results." With Klerio, she selects the Write JD Skill, fills in basic fields, and receives a complete, properly formatted JD. No monthly subscription.
Estimated: ~80,000 VND/month
Minh
Job Seeker
Minh manually searches LinkedIn, TopCV, VietnamWorks every morning - scrolling, filtering, copy-pasting, terrified of missing a good opening. The repetitive grind has made him exhausted and pessimistic. With Klerio's Job Search Skill, he uploads his CV once. The system surfaces a shortlist with a match score and recommended actions per job. Usage is bursty - intense over a few weeks, then zero once he lands a role.
Estimated: ~100,000-200,000 VND/active month
Quan
Marketing Manager
Needs a full PR campaign dashboard. Klerio's Skill connects to his Notion workspace via OAuth, collects campaign parameters via form, and auto-generates the complete database, relations, and views in 5 minutes.
Estimated: ~150,000-250,000 VND/month during active campaigns
Goals
- 1
Differentiated positioning: Become the first AI tool in Vietnam clearly recognized for "Solution-first / ready-to-use output," not chat-first.
- 2
Emotional positioning - "Tam tri thanh thoi" (Peace of Mind): Position Klerio as reducing cognitive stress, not just increasing speed.
- 3
Viral messaging - "1 nguoi = ca team": Primary social message targeting resource-constrained SMEs and individual contributors.
- 4
Individual-first GTM (B2C first, B2B later): Capture individual self-provisioning behavior before approaching enterprise sales.
- 5
User retention: A Skill library broad enough to cover the majority of common tasks for the target segment, reducing the need to switch to other tools.
- 6
Sustainable revenue: Pay-per-task credit structure achieving sufficient ARPU. Prioritize Skills with immediate revenue conversion over free user growth.
- 7
Guardrail metric: Time from sign-up to first useful output must not exceed 2 minutes.
Non-Goals
- Not competing on free-form chat - chat fallback is a safety net, not a feature
- Not targeting developers or power users - this segment already has suitable tools
- Not building Skills for every industry in Phase 1 - Job Seeker and Marketer only
- Not training proprietary AI models - continue using existing APIs (Anthropic, OpenAI, Google)
- Not building enterprise / on-premise version in Phase 1 - focus on SaaS cloud for individuals and small teams
Solution
Key features, Skill roadmap, and core product logic.
Key Features
| # |
Feature |
Description |
Status |
Priority |
| 1 |
Skill Library |
Core differentiator. Each Skill: name + task description, input form, backend-engineered prompt, standardized output format. |
Not built |
P0 |
| 2 |
Smart Skill Matching |
User describes a problem in natural language; system suggests 1-3 relevant Skills as selectable options. |
Not built |
P0 |
| 3 |
Pay-per-task Credit System |
Pricing in VND per task (e.g., 10,000 VND per meeting minutes Skill). Pay-as-you-go structure. Per-Skill pricing TBD. |
Existing |
P0 |
| 4 |
Quick Prompt Buttons |
4 shortcut action buttons on home screen. Post-description, system surfaces Skill suggestions as selectable options. |
Existing |
P1 |
| 5 |
Privacy Guard |
Client-side PII detection blocks sensitive data from leaving the device before reaching any AI model. Adds transparency layer: users can see exactly what data is being used and have full control over it. |
Existing (masking) + Not built (transparency layer) |
P1 |
| 6 |
Prompt Rewrite Suggestions |
Opt-in tool at the main search/input bar. User can choose to rewrite their input before submitting; system suggests multiple rewritten versions. Not automatic, not tied to Skill-matching failure - purely user-triggered. |
Not built |
P2 |
| 7 |
Multi-model Backend |
Backend selects the optimal model per Skill. Primary model: Claude. Model selection UI under consideration - pending UX evaluation. |
Existing |
P1 |
| 8 |
Chat Fallback |
Available for tasks without a matching Skill. Demand-collection channel to identify new Skills to build. Not marketed as primary feature. |
Existing |
P2 |
| 9 |
Notes Module |
Retained as-is. Low priority in current phase. |
Existing |
P3 |
| 10 |
Mobile Support |
Strong signal that Vietnamese users (especially outside major cities) require mobile access. To be evaluated - consider Remote Control via mobile. |
Not built |
TBD |
| 11 |
Workspace Context Layer |
Workspace evolves from file storage into a context layer personalizing Skill outputs per team/department. Full design deferred pending Recommendation v2. |
Basic exists |
P2 |
| 12 |
Third-party Integrations |
Skills connect to and operate directly inside Notion, Google Drive, etc. Two output types: Discrete deliverable and Workspace artifact. Deferred pending Recommendation v2. |
Not built |
P2 |
| 13 |
Mini Apps (Future) |
Users create custom mini apps tailored to their needs, managed from the left sidebar. Not yet scoped. |
Not built |
Future |
Phase 1 Skill Roadmap
Priority order based on segment urgency and revenue conversion potential. Two packs, 12 Skills total.
- Job Search Happy Case
- CV-JD Matching & Optimizer
- CV / Cover Letter Writing (ATS)
- Mock Interview
- Marketing / Creative Brief Priority #1
- Content Calendar
- Product / Service Description
- Video Script (TikTok / Reels / Shorts)
- Recruitment Post & Poster
- Report for Boss
- Customer Feedback Summary
- Market Research Summary
Deferred: Accountant Skills. Accounting involves complex workflows and strict VAS regulatory requirements. Requires 3-5 in-depth interviews with accountants before any build commitment.
Key Logic
- 1
Solution-first, not chat-first: Free-form chat is a fallback only, never marketed as a primary entry point.
- 2
Fresh-form-per-task architecture: Each Skill execution is a clean, independent session - prevents context degradation ("AI getting lazy") and ensures output consistency.
- 3
Build by data, not intuition: Skill priority determined by research signals. Unmatched chat requests are logged to identify next Skills to build.
- 4
Each Skill is an independent billing unit: Users see the VND credit cost of each Skill before use.
- 5
Output must be immediately usable: Never return "suggested approaches" or "structure recommendations" - results must be finished deliverables.
- 6
Output types: Discrete deliverable (JD, email, CV, report) produces standalone text/file output. Workspace artifact (dashboard, board) is built directly inside the user's existing tool - deferred.
- 7
Missing or invalid input: System prompts the user to complete input before processing. Never produce a low-quality output.
- 8
Output does not meet expectations: Allow regeneration; charged at a lower rate than first generation, or offered free once.
- 9
Skill versioning: When prompts are improved, older versions remain accessible; quality is tracked per version.
Launch Plan
Two parallel tracks: Technical (build) and Research (validate). Current status: Recommendation v1 approved 12 Jun 2026.
Track 1 - Technical Team
| Timeline | Milestone | Scope | Status |
|---|
| 19 Jun 2026 |
Happy Case |
Build Job Search Skill (CV in, aggregated job list + fit score + recommended actions out) and Skill engine foundation. Full team review to assess Solution-first / Skill-based model feasibility. |
In progress |
| Jul 2026 |
Full Development Kickoff |
Based on Recommendation v2. 12 Skills across 2 packs + Smart Skill Matching, Workspace Context Layer, Third-party Integrations (Notion, Google Drive), Mobile evaluation, Onboarding flow, Unmatched task tracking. |
Not started |
| Aug 2026 |
Development & QA |
Continued build and QA. Per-Skill quality bar defined and tested against real user tasks before release. |
Not started |
| Sep 2026 |
Closed Beta Launch |
Beta with Job Seeker and Marketer segments. Community feedback collected via social channels. |
Not started |
| Q4 2026 |
Continuous Improvement |
Ongoing feedback loop from beta and early users. Skill library expansion based on unmatched task signals. Market research continues. |
Not started |
Track 2 - Research Team
| Timeline | Milestone | Scope | Status |
|---|
| 30 Jun 2026 |
Research Phase 2 |
User research (in-depth interviews: 5 segments - Job Seeker, Marketer, HR/TA, Admin, Accountant) + Market research update + Hands-on competitor experience (Claude Cowork, Codex Remote, ChatGPT Pro, Gemini Pro, AI Hay, Slider.ai). Output: Recommendation v2. |
In progress |
| Jul - Aug 2026 |
Community Hire |
Community Marketing Executive onboarded. Engages social platforms (Threads, Facebook groups, TikTok, LinkedIn) to collect real user feedback. Findings fed back to Product on a rolling basis. |
Not started |
| Sep 2026 |
Beta Feedback Synthesis |
Collect and synthesize feedback from closed beta users via community channels and direct interviews. |
Not started |
| Q4 2026 |
Ongoing Market Research |
Continuous signal collection and hypothesis tracking. New findings fed into PRD updates and next product cycle. |
Not started |
Operational Checklist
| Item | Status | Owner |
|---|
| Happy Case review session scheduled (19 Jun) | In progress | Product |
| Skill QA process (quality bar definition per Skill) | Not defined | Product |
| Per-Skill pricing model (VND per task) | Not built | Product / Finance |
| Onboarding flow for Solution-first / Skill-based UX | Not designed | Design |
| Mechanism for tracking unmatched tasks (demand signal) | Not built | Engineering |
| Model selection UI decision (keep or remove) | Under consideration | Product |
| Fresh-form-per-task session architecture | To design | Engineering |
| Mobile UX evaluation and feasibility assessment | Not started | Product / Engineering |
| API agreements in place (Anthropic, OpenAI, Google) | TBD | Legal / Finance |
| Community Marketing Executive hiring | Not started | People |
| VND payment gateway integrated | Existing | Engineering |
| Multi-model backend cost optimization | Existing | Engineering |
| Client-side PII masking (Privacy Guard - masking layer) | Existing | Engineering |
| Privacy Guard - transparency and control layer | Not built | Engineering |
| Prompt Rewrite Suggestions (search/input bar, opt-in) | Not built | Product / Engineering |
Competitive & Positioning
Three competitive groups, positioning messages, pricing rationale, and SWOT.
Competitive Landscape
ChatGPT (OpenAI)
81% platform share in Vietnam. Recently launched $100/month Pro tier to segment users by willingness to pay. Free Go plan at 132K VND/month across 16 Asian markets.
Claude (Anthropic)
Highest-rated for Vietnamese writing quality. Fable 5 strong on spreadsheet processing and legal documents. Organic fan base in Vietnam but too expensive for daily use.
Google Workspace AI
Docs Live and Gmail Live directly overlap with Klerio email and document use cases. Native Gmail/Docs integration is the most dangerous moat. Rolling out summer 2026 at $19.99/month.
Microsoft Copilot Cowork
GA June 16, 2026. Runs on Claude Opus 4.8 + Sonnet 4.6. Distributed inside M365 that Vietnamese enterprise workers already use daily. Pay-per-use credits ($0.01/credit) now active.
FPT AI Assistant
Expected launch H2 2026. Most direct localization competitor. Combines local language, local trust, and potential enterprise distribution. Priority to monitor.
AI Hay
Notable satisfaction rate (47%) due to local context understanding. Usage is 24% but satisfaction nearly matches ChatGPT.
Careernote + MistWriter
Niche tools targeting CV optimization and thesis writing. Overlap with the Job Seeker segment. Watch for feature expansion.
WPS 365
Low-cost Microsoft Office alternative with integrated AI penetrating Southeast Asia. Vietnamese SMBs are a natural target.
Manual processes
In traditional industries (especially Accounting), the default competitor is doing it by hand or "Excel gymnastics." Building trust is as important as building features.
Positioning Messages
| Message | Rationale |
|---|
| Solution-first / Skill-based | Directly addresses the prompt barrier - the number one reason non-technical users abandon AI tools |
| "1 nguoi = ca team" | Highest-viral framing in Vietnamese social media; resonates with SMEs and under-resourced teams |
| "Tam tri thanh thoi" (Peace of Mind) | Mental wellness is a top priority in Vietnam 2026; positioning AI as cognitive stress relief creates emotional differentiation |
| Official Provider Layer | Klerio uses Claude/GPT APIs directly from official providers, ensuring quality, security, and account stability |
| Privacy Guard | Sensitive data is blocked before it ever leaves the user's device; users see exactly what data is used and have full control - extends trust beyond PII masking |
| Stable Claude Layer | Solves specific fears of users who tried Claude directly: account bans, token limits, response quality degradation over long sessions |
Pricing Strategy
Currency
VND, not USD. Makes pricing psychologically accessible to Vietnamese users. Grey market anchor: 40K-89K VND/month. Klerio pricing must compare favorably to reseller prices, not official ChatGPT.
Model
Pay-per-task. e.g. 10,000 VND per meeting minutes Skill. Specific per-Skill pricing to be calculated based on operational costs. Each Skill is an independent billing unit visible before use.
Revenue Priority
Conversion over growth. Vietnam startup funding down 87% YoY. Prioritize Skills with immediate revenue conversion (Job Search, CV-JD Matching) over free user growth. Track conversion rate per Skill from day one.
SWOT
Strengths
- Solution-first / Skill-based model eliminates prompt barrier entirely
- Official API layer: stable, secure, and quality-guaranteed
- Pay-per-task matches Vietnamese payment behavior
- Fresh-form-per-task architecture prevents output degradation
- Privacy Guard is an uncontested differentiator - no competitor offers client-side PII masking
Weaknesses
- Skill library needs sufficient depth for retention before launch
- Trust-building required vs. established international tools
- Dependency on third-party model API quality and availability
- Small team - Research Phase 2 hard deadline is a real constraint
Opportunities
- Accountant segment: 105+ posts asking about AI - completely underserved
- Individual-first entry into enterprises (15pp personal vs. work gap)
- Frontier models becoming free/cheap - value shifts fully to guided workflows
- Copilot "for entertainment only" disclaimer viral - users distrust enterprise AI
Threats
- Microsoft Copilot enterprise top-down deployment
- FPT AI Assistant launching H2 2026 - most direct local competitor
- Google Workspace AI deepening Docs/Gmail integration
- Brain-rot anxiety (AI dependency fear) reducing adoption intent
Risks & Open Questions
Known risks with mitigations, and open questions requiring further research or product decisions.
Risks
Risk
Users abandon Klerio for Claude or ChatGPT free chat because their specific task has no matching Skill.
Mitigation
Focus Phase 1 on 2 segments only. Log all unmatched tasks. Set clear retention metrics per Skill.
Risk
Klerio's value proposition collapses if the output still requires substantial work to be usable.
Mitigation
Define explicit quality bar per Skill before build. QA each Skill against real user tasks before release. Never ship a Skill that doesn't meet the bar.
Risk
A well-funded local competitor or a native-integration player captures the market before Klerio achieves product-market fit.
Mitigation
Build defensibility through deep localization (VN formats, dialects, legal docs) and superior Skill quality, not model access.
Risk
Full development kickoff is blocked until Recommendation v2 is delivered. A slip here cascades to beta timeline.
Mitigation
Hard deadline: 30 Jun for Research Phase 2 output. No extensions.
Risk
Klerio runs out of runway before achieving sustainable revenue if early growth metrics are non-paying users.
Mitigation
Prioritize Skills with immediate revenue conversion. Track conversion rate per Skill from day one.
Risk
External API changes or deprecations break Skills that produce Workspace artifacts.
Mitigation
Build with fallback behavior (export to file if API fails). Workspace artifact Skills are deferred to later phase.
Open Questions
To be answered by Research Track
H1Is the prompt barrier the primary reason office workers abandon AI tools? Or is it lack of motivation, trust, or perceived relevance?
H2Will users pay per task in VND if the output feels worth it? Compared to a monthly subscription - what is the "worth it" threshold?
H6Will document-processing Skills have the highest conversion rates? Or will Job Search or CV-JD Matching outperform?
Which of the 5 segments has the highest willingness to pay and lowest Skill complexity?
To be answered by Product / Technical Team
Per-Skill pricing structure: flat fee per task, by output length, or by underlying model cost?
Smart Skill matching: LLM-based or rule-based?
Model selection UI: keep or remove? Under consideration pending UX evaluation.
Mobile: full responsive web first, or native app required? What is the MVP scope for Remote Control via mobile?
Skill quality measurement: user rating? Edit distance between raw output and final version used?
Does the OAuth flow in third-party integrations break the 2-minute guardrail?
Private Skills from Phase 1 or defer? Should users be able to create private Skills early?
Terminology Changes
| Old Term | New Term | Reason |
|---|
| Template | Skill | More accurately reflects what the product does: not just a document template, but an executable guided workflow that produces a ready-to-use output |
| Template library | Skill library | Consistent with Skill terminology |
| Template-first | Solution-first / Skill-based | More accurately reflects both the positioning (solution, not chat) and the product mechanism (Skill-based guided workflows) |
Changelog
-
v0.1-v0.6 02/06/2026
Initial drafts: LacViet AI baseline to Klerio positioning, solution-first / template-based model, pay-as-you-go rationale, Workspace context layer, third-party integrations, translated to English.
-
v1.0 12/06/2026
Post-Recommendation v1 (Founder approved 12 Jun 2026): Terminology changed to Skill throughout. Model selection UI moved to under consideration. Target segments confirmed: Job Seeker (#1), Marketer (#2). Happy Case defined: Job Search Skill. Job Seeker Pack (4 Skills) + Marketer Pack (8 Skills) = 12 Skills total. Fresh-form-per-task architecture added. Competitive landscape updated: FPT AI Assistant, Google Workspace AI (Docs Live, Gmail Live), WPS 365. Positioning messages formalized. Revenue-first priority added. Roadmap restructured into 2 parallel tracks.
-
v1.1 19/06/2026
Founder-requested additions: Privacy Guard added (expands PII masking with transparency/control layer). Prompt Rewrite Suggestions added (opt-in at main search/input bar, user-triggered). Key Features renumbered (#5-13). Operational Checklist updated.