# Architectural Decisions Log

Detailed reasoning for the structural choices made in the Tech Sequence Holdings architecture. Captured from the design conversation that produced versions v1 through v3.1.

---

## D1. Holding company architecture: Singapore parent + Vietnam OpCos

**Decision:** Tech Sequence Holdings Pte Ltd (Singapore) as parent, with each operating entity structured as a Singapore HoldCo + Vietnam OpCo pair.

**Why this over alternatives:**

- Vietnamese parent only (rejected): Foreign VCs avoid investing directly into Vietnamese entities due to FDI rules, currency controls, SBV foreign exchange regulations, and exit friction. Caps capital at local/regional sources.
- Cayman/BVI top layer (deferred): Adds complexity only justified if specifically targeting US listing or US-based late-stage investors. Premature at current stage. Architecture leaves room to add Cayman layer later if needed.
- Singapore parent (selected): Standard playbook for venture-backed Vietnamese tech (VNG, Tiki, Sky Mavis, Momo). English common law, predictable courts, no capital controls, USD-denominated, tax treaties with 90+ countries, familiar to global investors, clean exit paths.

**Implication:** Each venture has its own clean cap table. Investors fund individual entities, not the parent.

---

## D2. Founder equity placement

**Decision:** Founder shares sit at Tech Sequence Holdings Pte Ltd (Singapore), not scattered across Vietnamese entities.

**Why:** Vietnamese individuals owning Singapore entities is tax-feasible but requires proper structuring. Concentrating founder equity at one Singapore vehicle simplifies reporting, tax treatment, and future transactions.

**Implication:** Need a Vietnamese tax advisor specializing in cross-border holding structures (PwC, KPMG, or specialist firm like Acclime) before incorporation.

---

## D3. ESOP at entity level, not group level

**Decision:** Each venture has its own ESOP pool (10-15%), not a group-wide ESOP.

**Why:** Group-level ESOPs are hard to make compelling for talent betting on a specific venture's upside. Entity-level ESOPs let CoreHR engineers benefit from CoreHR's success without dilution from unrelated ventures.

**Implication:** Standard ESOP infrastructure replicated at each entity. Slight admin overhead, large talent-attraction benefit.

---

## D4. Groups as governance constructs, not legal entities

**Decision:** The five operating groups (Ventures, Infrastructure, Services, Culture and Creative, Capital) are management constructs, not legal sub-holdings, at least initially.

**Why:** Creating legal sub-holdings prematurely adds tax, reporting, and restructuring cost without operational benefit. Groups serve internal management, investor narrative, and brand architecture purposes.

**Future trigger:** Promote a group to a legal sub-holding when it raises capital as a unit (e.g. TS Services carve-out IPO, TS Block fintech group fundraise from a strategic bank partner).

---

## D5. IP ownership and licensing framework

**Decision:** Build an explicit IP licensing framework between TS Intel and the ventures from day one.

**Why:** AI built by TS Intel is consumed by Hogcare, Argus, Gather AI, Beacon, Intelomics, Signet, TS Health. Without a licensing framework, future Series A in any AI-touching venture gets messy in diligence. Investors will ask who owns the underlying models.

**Implication:** Standardized internal licensing terms; transfer pricing documentation required for related-party transactions.

---

## D6. Intercompany agreements and transfer pricing

**Decision:** Clean intercompany service agreements between all entities from incorporation.

**Why:** When TS Services consults for Argus, when TS Pay processes payments for Mira, when TS Intel licenses models to Hogcare, these are related-party transactions. Vietnamese tax authorities and future auditors will scrutinize. Sloppy intercompany flow costs months in diligence.

---

## D7. Banking and treasury separation

**Decision:** Each entity has its own bank accounts and books. No co-mingling at TS Shared Services level despite the temptation.

**Why:** Co-mingled cash is the number one cause of failed diligence in conglomerate-style structures. Separation is operationally inconvenient early but saves months of pain at Series A.

---

## D8. Why TS Intel sits in Infrastructure, not Ventures

**Decision:** TS Intel (AI development arm) is grouped with TS Data Services as shared infrastructure, not as a peer venture.

**Why:** TS Intel's primary role is internal leverage. If treated as a venture, every other AI-touching company in the portfolio either duplicates AI capability internally or competes with TS Intel for the same talent. As infrastructure, it builds shared capability that every venture consumes, and partnership candidates (AWS, NVIDIA, Microsoft, Google) match this framing.

**Alternative considered:** TS Intel as standalone services company selling AI development externally. Rejected as primary framing because it undersells the internal leverage role. External AI services can be a TS Intel product line over time without breaking the structure.

---

## D9. TS Block as fintech holding, TS Pay as first product

**Decision:** TS Block is scoped as a fintech holding, not a single product. TS Pay (B2B payments) is its first product line. Future products (lending, wealth, insurance) become additional nested children.

**Why:** Fintech alone is too vague to be a strategy. Payments specifically is the most defensible fintech wedge in Vietnam, and every venture in the portfolio consumes payments: B2B settlement (Mira), payroll (ViecHomNay), event ticketing (Pavilo), rent collection (Unit Desk), supplier payments (WarehouseOS). High leverage multiplier across the group.

**Implication:** TS Block's group placement (TS Capital and Financial Services) is correct for licensing/governance clarity. Long-term, when TS Block has 3+ product lines, it likely spins out as its own TS Financial Services group.

---

## D10. Why Culture and Creative gets its own group

**Decision:** TS Studio, TS Media, Saigon Volume, and TS Live form a distinct group, separate from Ventures and Services.

**Why:** Creative entities run on different unit economics than SaaS:
- Creative talent pool (not engineers)
- Project- or production-cycle revenue (not ARR/NRR/CAC payback)
- Facility/asset-based for some (Saigon Volume LED stage, TS Live venue partnerships)
- Brand- and partnership-driven GTM (not sales funnels)

Grouping them protects the "Ventures = software" mental model for investors. Stamping a virtual production studio into TS Ventures alongside Mira and CoreHR is technically possible but structurally misleading.

**Strategic upside:** Most Vietnamese tech holdings (FPT, VNG, Vingroup tech arm) don't have a serious creative arm. Building this well creates a moat. Tech Sequence becomes the holding that owns the brand, content, and cultural surface of Vietnamese tech.

---

## D11. Why Saigon Volume is not TS-prefixed

**Decision:** Virtual production studio is named "Saigon Volume", does not carry TS prefix.

**Why:** Serious creative studios brand independently of their parents. Industrial Light and Magic is not "Disney Studios A". Pixomondo, Trilith, Weta, none use parent-company nomenclature. When a Netflix Vietnam production is shot at the LED volume, the credit needs to read as a standalone brand. The TS prefix would commodify the studio in industry credits and ad agency conversations.

**Naming rationale:**
- "Saigon": culturally weighted, internationally recognizable, signals place
- "Volume": industry term for LED virtual production stage; immediately signals premium category to industry buyers; carries craft connotation that "Production" doesn't

**Rejected: "Saigon Production"**: "Production" is generic, doesn't differentiate from commodity video houses, drags pricing down, and is hard to trademark.

**Implication:** Trademark Saigon Volume in Vietnam (Cuc So huu tri tue). Register domains: .com, .vn, .studio, .co.

---

## D12. Shared Services attached to parent, not in any operating group

**Decision:** TS Shared Services is a centralized back-office layer attached to Tech Sequence Holdings, not classified under any of the five operating groups.

**Why:** Shared Services covers finance, legal, HR ops, IT, procurement, treasury, internal audit, functions that serve every entity in the group. Placing it under any one operating group would distort its mission. At 28+ entities, distributed back-offices are where conglomerates bleed money.

**Implication:** Can be an internal function or a legal entity (with internal service agreements). Tax structuring drives the choice.

---

## D13. Capital strategy varies by venture type

**Decision:** Each venture's capital strategy is set at entity level, not group level.

| Venture profile | Typical capital |
|---|---|
| Hogcare, Vitrine, Unit Desk | Strategic capital from sector incumbents (agriculture, museums/gov, real estate) |
| Signet, CoreHR, Gather AI, Mira | VC-shaped (high growth, software margins, regional expansion) |
| TS Block | Strategic capital from bank or fintech partner due to licensing |
| TS Data Services, TS Intel | Strategic partnership with AWS/Google/Microsoft/NVIDIA |
| Saigon Volume, TS Live | Project finance + strategic (media partners, venues) |
| Intelomics | Specialist biotech investors, longer R&D cycle |

**Why this matters:** VC capital usually comes cleaner with growth pressure. Strategic capital often comes with strings (anchor customer rights, ROFR on exits, board veto rights). Choosing per venture protects cap table cleanliness at entity level.

---

## D14. Auditor selection from year one

**Decision:** Engage a Big 4 auditor (likely KPMG or Deloitte for Vietnam) from year one.

**Why:** Switching auditors before a fundraise is painful and signals to investors that something is off. Big 4 from incorporation = credible audited financials from day one, smoother diligence.

---

## D15. Consolidation candidates flagged for future review

The following pairs/groups are worth evaluating for consolidation rather than remaining as separate entities:

- Gather AI + Beacon: both AI/data products, overlapping customers, shared talent. Could be one company with two product lines.
- Vitrine inside Signet: Vitrine is essentially a vertical application of Signet (museum-specific digitalization). Hold as Signet product line until Vitrine has standalone revenue justifying separation.
- CoreHR + ViecHomNay + TS HR: talent stack fragmented across three entities. Could anchor a future "TS People" sub-group with shared GTM, or potentially merge depending on revenue trajectory.

---

## D16. TS Land — placement, structure, and Unit Desk relationship

**Decision:** TS Land is added as a TS Ventures entity under the Marketplaces and Networks sub-category, structured as a single entity with internal product lines (listings marketplace, agent CRM, 3D visualization, mapping/zoning, AI valuation, project evaluation network) rather than a sub-holding with nested ventures.

**Why a single entity, not a TS Block-style sub-holding:** TS Land's product lines all share the same customer base (agents, buyers, renters, developers), the same data layer, and the same regulatory regime. They are not separated by distinct licensing requirements the way payments, lending, wealth, and insurance are under SBV. The TS Block pattern is justified by regulatory separation; TS Land has no equivalent driver. Internal product lines stay inside one entity, with optional carve-out later if a product line (e.g. the AI valuation engine sold to banks for mortgage underwriting) develops independent fundraising potential — same playbook as the open Vitrine-inside-Signet question.

**Relationship to Unit Desk (no overlap, designed complementarity):**

- Unit Desk = landlord and operator back-office (B2B SaaS for managing tenants, leases, maintenance across a real estate portfolio).
- TS Land = transaction marketplace and agent ecosystem (front-of-house, consumer and agent surfaces).

They share a sector but serve different actors. Designed data interchange from day one:
- TS Land listings and transaction data enrich Unit Desk market intelligence
- Unit Desk portfolio inventory becomes potential listings supply for TS Land

This interchange should be specified in the IP licensing framework (D5) at the time TS Land is incorporated, not retrofitted.

**Capital strategy:** Strategic capital (per D13), not venture capital. Vietnamese PropTech incumbents (PropertyGuru-owned batdongsan.com.vn, Carousell-owned Nhatot, Meey Group, Homedy, RealEstate.com.vn) are well-capitalized; a marketplace cannot bootstrap against them at venture pre-seed scale. Realistic launch: $3–5M seed; $10–15M Series A.

**Sequencing trigger:** Strategic anchor LOI from a Vietnamese real estate developer (Vinhomes, Novaland, Sun Group, Capitaland Vietnam, Khang Dien). The anchor brings listings supply and distribution; TS Land brings the technology platform and the agent ecosystem. Without such an anchor, the venture should not be incorporated — bootstrapping a marketplace against incumbents is the failure mode.

**Brand consideration (open):** TS Land follows the TS-prefix convention for now. The standalone-brand argument (D11 for Saigon Volume) applies partially: consumer real estate marketplaces in Vietnam typically use standalone brands. If consumer trust, SEO, and agent recognition prove materially weaker with the TS prefix during go-to-market, rebranding to a standalone consumer name remains an option. Trademark both TS Land and at least one standalone alternative defensively at incorporation.

---

## D17. Government-customer blast-radius management

**Decision:** A separate Vietnamese sub-holding (working name **TS Public Sector Holdings Vietnam JSC**) sits between TS Holdings Pte Ltd and any entity with material government-customer exposure. This is an explicit exception to D4 ("groups are governance constructs, not legal entities"), justified by blast-radius containment rather than capital-raising convenience.

**Why this breaks D4:** D4 holds that legal sub-holdings add cost without operational benefit at the current stage. Government-customer exposure is the case where the operational benefit (containing investigation blast radius) exceeds the cost. Vietnamese contagion patterns are real and have hit unrelated parts of conglomerate-style structures in recent investigations (AIC Group, FLC, Việt Á, SCB):

- Banking freezes across related parties via asset-preservation orders (Lệnh kê biên)
- Legal representative (Người đại diện theo pháp luật) personal criminal liability cascading to all entities where the same person holds the role
- Tổng cục Thuế (General Department of Taxation) investigating all related parties under transfer-pricing scrutiny once one entity is flagged
- License revocation by beneficial-owner suitability — sector regulators (MIC, SBV, MoH, MoCST) deny or revoke licenses based on "responsible person" assessment
- Exit bans (Lệnh cấm xuất cảnh) as preliminary measures, often pre-charge
- Reputational cascade among Vietnamese strategic partners (banks, telcos, SOEs, ministries) closing doors to all related entities

Driving laws: Luật Đấu thầu 2023 (Public Procurement), Luật Phòng chống tham nhũng 2018 (Anti-Corruption), Luật Bảo vệ bí mật nhà nước 2018 (State Secrets), Luật An ninh mạng 2018 + Nghị định 53/2022 (Cybersecurity + Data Localization), Nghị định 13/2023 (PDPD).

**The structural firewall:**

```
TS Holdings Pte Ltd (Singapore, founder owns)
  └── TS Public Sector Holdings Pte Ltd (Singapore intermediate sub-holding)
        └── TS Public Sector Holdings Vietnam JSC (Vietnamese sub-holding,
              Vietnamese majority where required, incorporated just-in-time
              before first Tier-1 launch)
              └── Tier-1 operating entities (each its own Vietnam OpCo)
```

- Each operating entity has its own Vietnamese-citizen legal representative — never the founder
- Founder owns the Singapore parent only and is not Người đại diện anywhere in the Vietnamese stack
- Banking, office space, employees, and intercompany flows are segregated between gov-facing entities and the rest of TS

**Two-tier classification:**

| Tier | Definition | Treatment |
|---|---|---|
| **Tier 1** | >40% revenue from gov / SOE / public sector, OR sector requires Vietnamese majority | Sit under TS Public Sector Holdings, separate banking, separate legal rep, Vietnamese strategic co-investor minority where possible |
| **Tier 2** | <40% revenue from gov, but real gov touch | Stay in normal groups, adopt Tier-2 compliance discipline (separate legal rep, no intercompany loans against gov contracts, gov-contract documentation isolated) |
| **Tier 3** | No government touch | Standard discipline (D6, D7) sufficient |

Initial tier assignment is tracked in [../README.md](../README.md). Reassessed at each entity's incorporation and annually thereafter.

**Entity reassessments forced by this lens:**

1. **Argus** — almost certainly Ministry of Public Security / A05 cyber unit as anchor customer. Foreign ownership caps probable. Move under TS Public; consider Vietnamese majority structure. Material change to D1's Singapore-HoldCo-+-Vietnam-OpCo default for this entity. Requires sector-specific legal scoping before incorporation.

2. **TS Data Services** — split into a Tier-1 gov-cloud arm (under TS Public, Vietnamese majority for the gov cloud) and a Tier-3 commercial-cloud arm (standard structure). One entity cannot easily straddle both.

3. **Vitrine inside Signet** — D17 inverts the prior open question. If Vitrine targets MoCST and provincial museum systems (Tier 1) while Signet serves commercial digitalization (Tier 3), they cannot be the same entity without contaminating Signet with Tier-1 blast-radius exposure. **Recommendation: separate Vitrine into a standalone Tier-1 entity under TS Public, opposite of the prior consolidation lean.**

**Vietnamese SOE / quasi-government co-investor strategy:**

For Tier 1 entities specifically, a 10–15% minority Vietnamese strategic co-investor materially reduces investigation risk by creating internal political resistance. Candidate co-investors:

- SCIC (State Capital Investment Corporation) — sovereign-style minority
- FPT Corporation — tier-1 political cover, common pattern in Vietnamese tech
- Viettel Group (military-owned telecom) — for cybersecurity, telecom, gov-cloud entities
- VNPT — similar profile for telecom-adjacent
- Vietcombank, BIDV — for fintech-adjacent

**Operating disciplines:** The 12 mechanical compliance practices that apply to every Tier-1 and Tier-2 entity are specified in [../docs/government-customer-compliance.md](../docs/government-customer-compliance.md). These are part of the incorporation checklist for any new gov-facing entity, not retrofitted later.

**Implication for D5 (IP licensing framework):** Government data acquired through any one entity cannot flow to other entities without explicit licensing terms. TS Intel's shared AI infrastructure is the highest-risk channel — if government data trains shared models, the model itself becomes contaminated for unrelated use cases. D5 framework must address this explicitly before TS Intel ingests any gov-source data.

**Implication for D7 (banking separation):** D7's no-co-mingling applies generally. For gov-facing entities, the rule is stricter — no intercompany loans, period. Cost-plus services billing only, with full legal review at arm's length. Easy intercompany loans become investigation evidence of related-party benefit.

**Implication for D1 (Singapore-HoldCo + Vietnam-OpCo default):** For Tier-1 entities in sectors with foreign ownership caps (Argus / cybersecurity, TS Data Services gov-cloud, possibly TS Land for surveying/mapping data), the default structure does not work. Vietnamese-majority Vietnamese JSC under TS Public is the alternative pattern. D1 holds for Tier-3 and most Tier-2 entities.

---

## D18. CEO-first incorporation rule

**Decision:** No TS entity is incorporated (Singapore HoldCo or Vietnam OpCo) until an operator-CEO has signed an offer letter committing to a 3-year mandate. This is the gating discipline that paces the 8-year aggressive growth plan and prevents the single most common failure mode in portfolio-style structures: entity drift while waiting for CEO recruitment to complete post-launch.

**The four gates required before incorporation:**

| Gate | Requirement |
|---|---|
| **CEO** | Operator-CEO offer letter signed; equity terms agreed; performance milestones documented; D&O insurance from Vietnamese carrier committed; succession plan drafted |
| **Customer** | Validated customer wedge — signed paying customers (B2B/SaaS) or demonstrated demand at scale (B2C / marketplace) |
| **Capital** | LOI from strategic anchor OR VC term sheet OR founder-shareholder cap table closed with operator-CEO |
| **Capacity** | Shared Services has confirmed bandwidth to onboard another entity (gov affairs, compliance, finance, IT) |

All four gates must pass before legal incorporation begins. The CEO gate is the hardest — see [../docs/operator-ceo-recruiting.md](../docs/operator-ceo-recruiting.md) for the recruiting playbook this rule depends on.

**Why this is non-negotiable:** founder belief that a CEO can be recruited *after* incorporation is the most fatal pattern in aggressive growth portfolios. Post-launch recruiting takes 6–12 months while burn accumulates and the entity drifts without leadership. Better to hold the launch date for a quarter than launch with the wrong CEO.

**Enforcement:** Senior CEO Recruiter (TS Shared Services, hired Year 2) maintains this gate as a hard requirement, not a suggestion. Before that role exists, the founder enforces personally. Exception requires founder + Shared Services Legal sign-off + post-incorporation CEO commitment deadline of 30 days with documented escalation if missed.

**Implication for the 8-year plan:** entity launch dates in [../docs/aggressive-8-year-growth-plan.md](../docs/aggressive-8-year-growth-plan.md) are targets, not commitments. If CEO recruiting capacity is constrained in any year, launches slip rather than CEO quality degrades. The annual portfolio review (D18 ongoing) checks "did we hold the quality line" as an explicit metric.

---

## D19. M&A discipline framework

**Decision:** M&A becomes the primary entity-growth mechanism from Year 3 onward. Without acquisitions sustaining ~30–50% of annual entity additions in Years 4–8, the 8-year timeline does not work. This decision establishes the discipline framework so M&A is a capability, not opportunism.

**Build the M&A function in Shared Services by Year 3:**

- **M&A Lead** — full-time hire by mid-Year 2, ramping by Year 3. Profile: ex-investment-banking VP or corporate-development director with Vietnamese M&A experience. Reports to founder.
- **Standing advisory relationships** — top 3 Vietnamese M&A advisors (Avalon, ASART, KPMG M&A, PwC Deals Vietnam, EY) on retainer or first-look
- **Target pipeline CRM** — 30–50 named candidates by end of Year 3, refreshed quarterly
- **Treasury reserve** at TS Holdings parent (per D21) for opportunistic deals

**Target screening criteria:**

| Criterion | Standard |
|---|---|
| Sector fit | Maps to existing TS group structure; doesn't create new governance complexity |
| Revenue stage | $1M–$10M ARR typical (under $1M: too early; over $10M: too expensive) |
| Team retention | Incumbent CEO or key leadership willing to stay 18–24 months minimum |
| Cap table cleanliness | No nominee shareholders, undisclosed liabilities, related-party transactions requiring unwind |
| Customer concentration | No single customer >40% of revenue |
| Tier classification | Targets classified per D17 before close — Tier 1 acquisitions move under TS Public Sector Holdings |

**Valuation discipline (Vietnamese market benchmarks, 2026):**

- **SaaS / recurring revenue**: 4–6× ARR healthy, up to 8× for fast-growing market leaders, never above 10×
- **Services / project revenue**: 1–3× revenue, 5–8× EBITDA
- **Asset-based businesses**: asset-based plus operational premium
- **Distressed**: 0.5–1.5× revenue, structured with earnouts and clawbacks

**Vietnamese M&A diligence specifics requiring extra scrutiny:**

- Related-party transactions in target's history (common; require cleanup before close)
- Transfer pricing positions for any cross-border element
- Undisclosed labor liabilities (severance, social insurance back-payments, 13th-month payments)
- Real estate / lease commitments (often material, sometimes off-balance-sheet)
- Government licenses and renewal cadence
- Beneficial ownership reality vs. registered ownership
- For Tier-1 targets: state-secret handling history, gov-customer contract terms

**Integration playbook (100 days):**

- Days 1–14: leadership alignment, immediate quick wins, no major changes signaled to customers or employees
- Days 15–45: finance + reporting integrated into Shared Services cost-plus billing (D6), banking transition, brand alignment decisions
- Days 46–90: tech stack rationalization, customer communication, team rationalization if needed
- Day 100: formal integration review with founder + Group MD; explicit go/no-go on continuing integration vs. unwind

**M&A pace target:** 11 acquisitions over Years 3–8 per the aggressive growth plan. 30–50% expected to underperform expectations within 24 months — this is normal at this pace. The framework reduces but does not eliminate this rate.

---

## D20. Product-line spin-out playbook

**Decision:** Mature product lines within existing TS entities are spun out as standalone entities when they achieve independent fundability. This becomes a primary new-entity source from Year 4 onward, accounting for ~15–25% of annual entity additions in Years 4–8. Pre-defining the playbook avoids ad-hoc negotiation each time and removes friction for the spin-out decision.

**Independence test — all three must apply:**

1. **Independent customer base** — customers buying the product line wouldn't naturally buy other products of the parent entity, or vice versa
2. **Independent fundraising path** — a dedicated investor segment exists (e.g., insurance-focused investors for an insurance product line spun out of TS Pay)
3. **Independent CEO available** — operator-CEO who can run the spin-out as a standalone business has been identified (per D18 CEO-first rule)

If only 1–2 apply, the product line stays inside the parent entity and continues maturing.

**Pre-spin validation period:** product line operates as a separate P&L within the parent entity for at least 18 months before formal spin-out. This validates the independence test before incurring incorporation cost.

**Standard spin-out cap table:**

| Stakeholder | Allocation |
|---|---|
| Parent entity | 40–60% retained |
| Operator-CEO + founding team | 20–30% (founder-shareholder model) |
| ESOP | 15% |
| External investors (concurrent seed/Series A) | 0–25% |

Parent retention is highest (50–60%) when the product line is core strategic infrastructure (e.g., a vertical AI from TS Intel that other portfolio entities consume); lowest (40%) when the spin-out is purely opportunistic.

**IP licensing terms (back to parent):**

- Shared foundational IP licensed back at favorable rates (1–3% royalty on revenue)
- Spin-out-specific IP (post-spin innovations) stays with the spin-out
- Pre-spin shared customer relationships: pro-rated revenue share for 18–24 months
- Trademark / brand: spin-out adopts new brand identity (TS-prefixed or standalone per D11 considerations)

**Transition services agreement (TSA):**

- Shared Services provides finance, HR, IT, legal at cost-plus 8% for first 12–18 months (D6)
- Office space access at fair-market sublease rates
- Joint customer success for shared customers (sunset over 6 months)
- TSA expires automatically at month 18 unless renewed by both sides

**Likely spin-out sources by Year 8:**

| Parent entity | Likely spin-outs (target Year) |
|---|---|
| TS Block | TS Lending (Y4), TS Wealth (Y6), TS Insurance (Y8) |
| TS Intel | Vertical AI products — medical AI (Y5), agri AI (Y7) |
| Signet | Vertical digitalization products as customer demand justifies |
| TS People | Executive search, specialized payroll, learning platforms (Y6–Y8) |
| TS Tech Consulting | Vertical consulting practices when revenue justifies P&L separation |

---

## D21. Capital recycling discipline

**Decision:** Liquidity events from mature TS entities (Series B/C sales, strategic sales, IPOs) are explicitly allocated to fund next-wave entity launches and M&A, not distributed as dividends. This sustains the 8-year aggressive growth plan's ~$800M–$1.1B cumulative capital deployment without requiring continuous founder capital injection.

**Allocation framework for liquidity events:**

| Allocation | Use |
|---|---|
| **50%** | Treasury reserve at TS Holdings parent — opportunistic M&A (D19) and next-wave entity seed capital |
| **30%** | Next-wave entity launches (Year N+1 to N+3 per the growth plan) |
| **20%** | Strategic M&A reserved for larger or transformational acquisitions outside the standard pipeline |

**No dividends to founder personal level** until the 8-year plan is funded (or scope is reduced). Founder treats TS Holdings parent as a permanent capital vehicle, not a distribution vehicle. Personal liquidity comes from a small annual draw (calibrated to lifestyle, not wealth accumulation) and from eventual carry-equivalent participation in major IPOs.

**Expected liquidity events Years 4–8:**

| Year | Event | Estimated proceeds to TS Holdings |
|---|---|---|
| Y4 | First Series B sale of an early-wave SaaS venture | $5–15M |
| Y5 | TS Services prepares for HOSE/HNX carve-out IPO (cash event Y6) | — |
| Y6 | TS Services Vietnamese IPO completes | $50–100M |
| Y6 | TS Block strategic raise from bank partner | $50–100M (mostly to TS Block, partial liquidity to TS Holdings) |
| Y7 | Series B/C sales across 2–3 portfolio entities | $30–60M cumulative |
| Y8 | Additional carve-out or strategic sale | $50–150M depending on which |

**Cumulative expected liquidity over Years 4–8: $200–500M.** Funds approximately 40–50% of the total $800M–$1.1B deployment required.

**Banking discipline:**

- TS Holdings parent maintains a dedicated treasury account at DBS Singapore (or comparable) separate from operating banking
- Treasury reserve target: minimum 12 months of planned next-wave entity launches + 6 months of M&A pipeline
- Below target: pause new entity launches until reserve replenished
- Above 24 months reserve: opportunistically deploy into accelerated launches or M&A

**Operating cash from cash-generative entities:**

TS Services (consulting + managed services), TS People, and eventually TS Pay should be cash-generative within 12–24 months of launch. Operating-level cash flows up to TS Holdings via:

- Cost-plus 8% billed from TS Shared Services (D6)
- Annual dividend from cash-generative entities (decided per entity board, typically 30–50% of free cash flow)
- License royalty from TS Brand & IP Holdings (D5)

Estimated annual upstreamed cash by Year 8: $30–80M/year from operating entities alone.

**Implication for founder personal capital:** founder personal capital is deployed primarily in Years 1–3 ($5–15M cumulative). From Year 4 onward, founder personal capital is rarely needed for new launches — recycled exit capital and operating cash carry the plan.

---

## D22. TS Field — drone + IoT + applied robotics in one entity, no robotics R&D

**Decision:** A single TS Ventures entity (working name **TS Field**) operates the physical-world data and automation layer for the portfolio and external customers, with three integrated product lines: **Aerial** (drone services + drone-data platform), **Ground** (IoT sensors + edge + IoT platform), and **Applied Robotics** (integration and deployment, explicitly *not* R&D). Tier 1 under TS Public Sector Holdings Vietnam JSC. Y4 launch via M&A wedge.

**Why one entity instead of three separate ones:**

The original framing considered drone, IoT, and robotics as three separate plays. Combining them into one entity is justified by:

- **Customer purchase pattern**: Vietnamese buyers of physical-world automation think "I have a physical-world problem, solve it." They don't separate the answer into drone vs. IoT vs. robotics. Single sales motion captures the integrated value; three separate motions fragment it.
- **Integrated data layer**: drone imagery + ground IoT sensors + robotics telemetry are far more valuable combined than separated. Building three independent platforms duplicates effort and forgoes cross-modality analytics.
- **Operating overhead**: three entities require three CEOs, three boards, three Compliance Officers, three Tier-1 strategic co-investor relationships. One entity is dramatically lower overhead at this stage.
- **Operator-CEO scarcity**: finding three Tier-1-quality CEOs simultaneously is harder than finding one CEO + one CTO pair.

**Why "applied robotics" not robotics R&D:**

Robotics hardware R&D has structurally hostile economics:
- Multi-year development cycles for new robot designs
- Capex-intensive (prototype tooling, certification, manufacturing setup)
- Dominated by international incumbents (FANUC, ABB, Yaskawa, KUKA, Universal Robots) with decades of accumulated engineering depth
- Hardware margins fundamentally lower than software margins
- Vingroup's VinSmart failure is a cautionary tale for Vietnamese hardware R&D ambitions at scale

Robotics **integration and deployment** has different economics:
- SaaS-adjacent (recurring services revenue with capex amortization)
- Value capture in operational software layer, not hardware manufacturing
- No supply-chain or inventory risk on robot hardware (vendor-managed)
- Customer-relationship-led (the value is matching robot to problem, configuring, deploying, operating)
- Direct adjacency to existing TS portfolio (WarehouseOS warehouse AMRs, TS Health lab automation, agricultural ground robots paired with Hogcare)

TS Field acts as **operator and integrator**, not manufacturer. Partners with FANUC, Locus Robotics, Geek+, OTSAW, KUKA, Universal Robots, Geek+ as channel suppliers; wraps best-in-class robotics with the TS Field data and operations layer.

**Why Tier 1 (under TS Public Sector Holdings):**

The whole entity inherits Tier 1 classification because of drone operations:
- CAAV operator certification + MoD vetting required
- ≥51% Vietnamese ownership mandatory for commercial drone operations above small-recreational scale
- Sensitive imagery (border zones, military facilities, government sites) triggers state-secret handling per Luật Bảo vệ bí mật nhà nước 2018
- Dual-use applications (surveillance, mapping, inspection) carry national-security considerations

Sub-arms (IoT, applied robotics) are typically Tier 2 in isolation, but housing them in a single entity simplifies governance vs. fragmenting across two entities at different tiers.

**Strategic co-investor strategy:**

**Viettel Group** is the primary co-investor candidate (15–20% minority). Reasons:
- Military-owned conglomerate with deep MoD relationships
- Existing drone operations (cyber + dual-use, not commercial-scale-focused, so partnership preferable to competition)
- Viettel IoT division — complementary not competitive when scoped correctly
- Manufacturing + agricultural customer relationships through Viettel's commercial arm
- Political cover for Tier-1 sensitive work
- International expansion alignment (Viettel operates in Cambodia, Laos, Myanmar, Mozambique, Tanzania, Peru — natural channel for TS Field regional expansion)

Alternative candidates: FPT (less defense-aligned, stronger commercial); VNPT (telecom-IoT-aligned); SCIC (sovereign, political cover only).

**Mechanism: M&A wedge.** Y4 launch starts with acquiring an established Vietnamese drone operator (primary target: MiSmart or comparable). The acquired entity provides:
- CAAV operator certification in good standing (transferable post-acquisition with regulatory consent; greenfield certification takes 18–24 months and may not be granted)
- Trained pilots + fleet (the team is the asset; cannot be hired from scratch in Vietnamese market quickly)
- Existing customer book (MARD programs, agricultural cooperatives, surveying contracts)
- Operational rhythm

IoT platform and applied-robotics integration capability are built greenfield on top of the acquired drone-operations foundation over Y4 Q3 → Y5 Q2.

**Cross-portfolio role (strongest in the portfolio):**

8+ internal TS entities are natural customers from day one: Hogcare (agri drones + livestock IoT), Blue Dot (environmental drones + sensors), TS Land (real-estate aerial + smart buildings), Unit Desk (building IoT), WarehouseOS (warehouse IoT + AMRs), TS Health (hospital IoT + lab automation), Mira (F&B outlet IoT), TS Data Services Gov-Cloud (co-sell to gov customers). Internal billing at cost-plus 5% (favorable vs. 8% Shared Services rate) encourages adoption.

**Capital and timing:**

- Y4 Q2 launch via M&A
- Total Y4 deployment: $5–11M ($3–6M M&A + $2–4M IoT platform + $500K–1M founder direct)
- Y5–Y6: additional $5–10M for platform scale + applied robotics buildout
- Series A from strategic co-investor target Y6 Q1 ($10–20M)
- Cumulative through Y6: $10–20M

**Implications for the 8-year plan:**
- Y4 cumulative entity count becomes 25 (was 24); TS Field replaces one previously-planned generic Y4 acquisition
- Y4 capital envelope unchanged (TS Field absorbed into M&A budget)
- TS Public Sector Holdings cluster grows to 5 Tier-1 entities at maturity (Blue Dot, Vitrine, Argus, TS Data Services Gov-Cloud, **TS Field**)
- Strongest internal cross-flow of any Ventures entity; central to TS Intel and TS Data Services value flywheel

**Implications for D17:** TS Field added to Tier-1 entity list under TS Public Sector Holdings VN JSC. State-secret handling protocols apply to drone imagery of sensitive sites. Vietnamese-majority structure required.

**Implications for D19 (M&A):** Y4 acquisition pipeline includes a Vietnamese drone-operator target with operator-certification transferability as critical diligence item.

---

## D23. Split TS Intel into TS Intel (commercial) and TS National AI (gov-facing)

**Decision:** TS Intel's planned scope is bifurcated into two legal entities mirroring the [TS Data Services Commercial / Gov-Cloud split](#) under [D17](#):

- **TS Intel** (existing, Y2 launch) — Foundation-model R&D + commercial AI capability. Sits in TS Infrastructure, Tier 3. Serves commercial TS portfolio entities and external commercial customers. Can raise from international AI investors.
- **TS National AI** (new, Y3 Q4 launch) — Applied AI vendor exclusively for government customers and Tier-1 TS entities. Sits under TS Public Sector Holdings VN JSC, Tier 1, Vietnamese majority. Licenses foundation models from TS Intel with strict one-way data flow.

**Why split AI specifically (more than other entities):**

AI contamination has properties that justify legal-entity separation rather than just operational discipline:

1. **One-way and irreversible.** Once a model is fine-tuned or trained on government data, it cannot be "un-trained." The model itself carries the contamination forever. A commercial customer would inherit gov-data-contaminated weights even if no gov data flows through inference. **Unlike data leaks or IP issues, this cannot be remediated after the fact.**
2. **Source-code review provisions** in Vietnamese gov contracts require Vietnamese-citizen-only model + source-code access. If TS Intel shared codebase with gov AI products, the commercial side inherits these constraints — slow iteration, restricted talent pool, export-control friction.
3. **Talent pools fundamentally differ.** Foundation-model R&D requires international AI talent (returnees from OpenAI, Anthropic, Google DeepMind). Government AI applications require Vietnamese citizens with clearance amenability. Forcing one entity to recruit both is impossible HR.
4. **Capital strategies fundamentally differ.** TS Intel commercial can credibly raise Series A/B from tier-1 international AI investors (Lightspeed, Sequoia, A16Z India). TS National AI gov raises from Vietnamese strategic capital (Viettel primary). Mixing poisons both fundraising tracks.
5. **Investor + customer optics.** International commercial customers and their compliance teams are increasingly cautious about AI workloads on infrastructure that serves government surveillance / dual-use applications.

**One-way data flow (the structural enforcement of D23):**

- TS Intel → TS National AI: foundation models licensed at favorable internal rate (2–4% royalty on TS National AI revenue from licensed-model applications, per D5 framework)
- **Gov-customer training data, fine-tuned weights, and derived improvements stay in TS National AI's tenancy. TS Intel never receives gov-derived training data, weights, or telemetry.**
- Technical enforcement: separate infrastructure (TS National AI on TS Data Services Gov-Cloud Y4+; TS Intel on commercial cloud), separate access controls, separate model registries, separate auditing
- Operational enforcement: Vietnamese-citizen-only access for gov-data handling; logged + audited every access
- Contractual enforcement: explicit IP licensing terms with one-way-flow language; annual third-party audit confirming no contamination

**Strategic co-investor for TS National AI: Viettel** (primary candidate, mirroring TS Field):
- Military-owned, MoD relationships
- Existing AI capability (Viettel AI — partnership preferable to competition)
- Same co-investor across multiple Tier-1 entities (TS Field, TS National AI) creates leverage and consistency
- Sovereign-AI positioning aligns with Viettel strategic narrative
- 18–20% minority stake target

Alternative: FPT (stronger commercial, weaker defense). Decline path: delay to Y4 and reassess if no acceptable Vietnamese strategic co-investor.

**When the split applies vs when operational discipline is sufficient (the broader question):**

The split pattern (TS Data Services Commercial/Gov-Cloud per D17; TS Intel/TS National AI per D23) applies when:
- Contamination/exposure is irreversible at the platform layer (AI, infrastructure)
- The two arms have fundamentally different talent/capital/regulatory profiles
- The shared resource is structurally a platform consumed by other entities (not a vertical sold to a specific customer base)

The split pattern does **NOT** apply to:
- **Tier-2 operating entities** (TS Tech Consulting, TS Managed Services, TS Health, TS Education, TS Land, Hogcare, Mira, TS People). For these, [government-customer-compliance.md](../docs/government-customer-compliance.md) Tier-2 disciplines are sufficient. Splitting would over-engineer without comparable benefit.
- **Tier-1 vertical entities** (Blue Dot, Vitrine, Argus). These are sold to specific government customer categories and don't have a "commercial" arm to separate.
- **TS Field** despite its drone + IoT + applied robotics scope (per D22) — the whole entity is already Tier 1 and unified under TS Public Sector Holdings.

**Rule:** apply split when (a) contamination is irreversible at platform level, AND (b) the two arms have fundamentally different talent/capital/regulatory profiles. Otherwise, Tier-2 operating disciplines are the right tool.

**Implications:**

- **TS Public Sector Holdings cluster grows to 6 Tier-1 entities** at maturity: Blue Dot (Y2), Vitrine (Y3 Q1), TS National AI (Y3 Q4), Argus (Y4), TS Data Services Gov-Cloud (Y4), TS Field (Y4)
- **Y3 cumulative entity count: 19** (was 18); Y8 cumulative: 43 (was 42). Capital impact: ~$2–5M Y3 (applied AI team + sovereign infrastructure setup), absorbed within Y3 envelope (~$50M total)
- **TS Intel positioning narrowed** — commercial only; gov customers route through TS National AI; TS Intel can pursue international VC fundraising without gov-customer concentration risk
- **All Tier-1 entity plans must reference TS National AI** for their AI capability (Blue Dot environmental AI, Vitrine cataloguing AI, Argus cyber AI, TS Field drone-imagery AI, TS Data Services Gov-Cloud workload hosting — all served by TS National AI not TS Intel directly)
- **D5 IP licensing framework** central to the structural integrity of this decision — must specify one-way-flow terms in writing before TS National AI incorporates

**Implications for D8 (TS Intel sits in Infrastructure, not Ventures):** D8 still holds. TS Intel remains shared infrastructure for commercial portfolio. TS National AI is its gov-facing applied sibling, not its replacement.

**Implications for D17 (firewall framework):** D17 extended with explicit row for TS Intel/TS National AI split, mirroring the existing TS Data Services Commercial/Gov-Cloud entry.

---

## Open questions / future decisions

1. TS Live name: placeholder. Resolves once events business direction is set (corporate B2B vs consumer concerts).
2. TS Studio external positioning: if external sales dominate over internal portfolio work, may benefit from non-TS brand.
3. Sequencing: which 3-5 entities to launch first. Drives near-term capital, hiring, and structural priorities.
4. TS Foundation: eventual philanthropy/impact vehicle. Add when group-level revenue justifies it.
5. Real estate / asset-backed entities: small TS Realty or TS Logistics ops companies as reference customers for Unit Desk and WarehouseOS. Adds tangible assets to balance sheet for debt financing optionality.