Sustain claimed 70%-plus topline growth
DSG's Series B announcement cites over 70% topline growth in 2024 and 5M+ products sold over two years, company-published figures quoted as their claim.
company claim, DSG Series B post (~Feb 2025); unaudited
What was done
- 01Classify all open Blood roles
- 02Break it down task by task
- 03Classify augment / automate / keep human
- 04Generate playbook for repeatable systems
Audit rows
107
across 10 open roles
Augment
54
AI assists, human owns
~53% of role hours
Automate + review
34
AI drafts, human approves
~27% of role hours
Keep human
19
the seat itself
~20% of role hours
Absorbed by OS workflows
41
rows the six systems below pick up
What the company is trying to do
five goals · read from public signal
Five goals read from public signal, each traced to the workflows on this page that serve it.
Indonesia expansion beyond e-commerce-only
Indonesia is live but e-commerce-only and recently licensed. BPOM registration and the mandatory halal deadline of October 2026 are a plausible gate on deeper entry; a Livestream Host (ID) posting in Jakarta is staffing the push.
verified market status (cascade research, 30 Jun 2026) + public Jakarta posting created 21 Apr 2026, captured 10 Jul 2026
India market entry
A Country Lead posting in Bangalore under Founders Office, created 30 March 2026, signals a live India entry not yet in the company's public market list.
public posting, created 30 Mar 2026, captured 10 Jul 2026
No Days Off cold-start launch
NDO is a confirmed second brand with a distinct tired-parent voice, DTC and marketplace presences to build from zero, and traction not yet public. A dedicated Performance Marketer posting in Kuala Lumpur is staffing the launch.
verified second brand (Blood LinkedIn, live nodaysoff.co) + public posting created 2 Mar 2026, captured 10 Jul 2026
Capital discipline through Series B
The company ran a long bootstrapped phase before a DSG-led Series A and a follow-on DSG-led Series B whose amount is undisclosed, so margin and headcount leverage carry the growth plan.
first-party DSG announcements; Series B amount undisclosed
System map
a drop means the system serves that goal · absorbs counts the audit rows it takes over
The role-task capability audit
10 roles · 107 tasks · scored 10 Jul 2026
All ten pslove roles public on 10 Jul 2026, decomposed into tasks, each task scored: augment, automate with review, or keep human.
- Task framing follows the Anthropic Economic Index: roles are scored as bundles of tasks, not job titles.
- Verdicts anchor to graded deliverables, per GDPval: the unit of scoring is what a reviewer can check.
- Verifiability sets the ceiling, per the SML rubric: output that cannot be checked stays augment or human.
- Independent spend authority scores keep human: where the JD grants it, the whole task stays with the owner.
- Friction “none” means no data or integration blocker once the absorbing workflow ships; one of the six workflows runs today, five are builds, sequenced in Layer 3.
- Time-weighted, automate with review covers ~27% of role hours, augment ~53%, keep human ~20%; the row counts above weight every task equally.
Scored from pslove's public job descriptions, captured 10 Jul 2026. Time shares marked est. are scoring assumptions; unmarked shares come from the JD. The org-wide audit is week-2 work with internal inputs; this page shows the method on the ten seats pslove has published.
ROLE-01 · BLR · FOUNDERSCountry Lead, India6 augment · 2 auto+review · 2 human
ROLE-02 · SG · MARKETINGDigital Marketing Executive6 augment · 3 auto+review · 2 human
ROLE-03 · SG · FOUNDERSHead of AI5 augment · 6 auto+review · 1 human
ROLE-04 · JKT · E-COMMLivestream Host (ID)3 augment · 4 auto+review · 1 human
ROLE-05 · KL · MARKETINGPerformance Marketer, NDO7 augment · 3 auto+review · 1 human
ROLE-06 · SG · MARKETINGPerformance Marketing Manager4 augment · 3 auto+review · 2 human
ROLE-07 · KL · SUPPLYRegional Demand Planning Manager6 augment · 4 auto+review · 2 human
ROLE-08 · SG · MARKETINGRegional Head of Brand & Innovation6 augment · 3 auto+review · 3 human
ROLE-09 · BKK · MARKETINGSenior Brand Manager, Thailand5 augment · 2 auto+review · 3 human
ROLE-10 · KL · MARKETINGSocial Media & Content Executive6 augment · 4 auto+review · 2 human
ten roles at a glance · the full row-level audit for every role continues below, unchanged
Country Lead Bangalore · Founders Office live posting 6 augment 2 automate 2 human
| Task | JD quote | Time | Verdict | Friction | Absorbed by | What unblocks it |
|---|---|---|---|---|---|---|
| Build and lead the founding India team | “Build the India business unit from zero to one” | ~15%est. | keep human | trust ceiling | none | stays human; hiring judgment and founder-proxy leadership can't be delegated |
| Incorporation, regulatory and competitive groundwork | “from incorporation to pricing, go-to-market” | ~10%est. | augment | none | Market-Entry Research Agent | |
| Quick-commerce partnership terms and co-activations | “agree on listing terms, maintain stock levels and plan marketing co-activations” | ~15%est. | augment | data readiness | KAM & Trade Comms | first months of India sell-through exports so platform comms argue from data |
| Promotion, discount and visibility optimization | “optimization of promotions, discounts and visibility” | ~15%est. | augment | data readiness | Settlement Reconciliation | live payout and campaign exports to decompose promo profit per SKU |
| Brand direction, influencer and PR relationships | “Direct brand efforts across digital, influencer, performance marketing and PR” | ~10%est. | keep human | trust ceiling | none | stays human; intimate-category voice and press judgment in a new market |
| Performance ad copy and creative variants | “on-platform growth tools, performance marketing and influencer” | ~10%est. | automate + review | none | Governed Creative Gate | |
| Fill rate monitoring and logistics partner management | “Ensure fill rate on sales channels” | ~10%est. | augment | integration gap | tooling augment | candidate cluster: fill-rate alerting off platform and 3PL stock feeds |
| Marketplace storefront queries and buyer chat | “Partner with quick-commerce platforms and marketplaces” | ~5%est. | automate + review | integration gap | Marketplace Chat First Responder | port chat connectors from Shopee/Lazada to India marketplace messaging |
| Team onboarding on SOPs and brand guidelines | “hiring and managing digital channels, marketing activities and operations teams” | ~5%est. | augment | data readiness | SOP & Onboarding Q&A | index SG/MY SOPs and brand guidelines into a Q&A corpus |
| Country P&L management and reporting | “Comfortable with P&L management” | ~5%est. | augment | integration gap | Settlement Reconciliation | platform payout reports and 3PL invoices wired into the netting pipeline |
| Month 1 With the OS behind them, month 1 ends with a source-cited India regulatory and competitor brief plus claims-gated launch creative already in human review; without it, that month goes to assembling the same groundwork by hand. | ||||||
Digital Marketing Executive Singapore · Marketing live posting 6 augment 3 automate 2 human
| Task | JD quote | Time | Verdict | Friction | Absorbed by | What unblocks it |
|---|---|---|---|---|---|---|
| Cross-platform paid campaign planning and buying | “Own end-to-end performance marketing” | ~25%est. | augment | integration gap | tooling augment | ad platform API access so drafted campaign structures push to live |
| Budget pacing and ROAS target management | “Manage budgets, pacing, and ROAS targets” | ~15%est. | augment | trust ceiling | tooling augment | review-gated reallocation proposals with an audit trail; spend stays human |
| Retargeting segment strategy | “Understand retargeting segments and implement strategies” | ~10%est. | augment | data readiness | tooling augment | first-party purchase data joined and exportable for segment hypotheses |
| Structured test design and iteration | “Design and run structured tests” | ~10%est. | augment | workflow fit | tooling augment | a shared test log the assistant reads and writes |
| Daily ad metric monitoring | “monitor ad metrics daily” | ~10%est. | automate + review | integration gap | tooling augment | scheduled metric pulls from Meta, TikTok, Google, Shopee, Lazada into one daily digest |
| Scale or kill spend decisions | “Scale up investment into winning ads” | ~5%est. | keep human | trust ceiling | none | nothing near-term; committing budget needs a human owner |
| Creative briefs and content angle planning | “Plan performance-driven content angles, formats, and messaging” | ~7%est. | automate + review | none | Governed Creative Gate | |
| Self-editing ad creative iterations | “Self-edit ad creatives for faster minor iterations” | ~4%est. | augment | none | tooling augment | |
| Starring in own ads | “comfortable starring in your own ads” | ~2%est. | keep human | trust ceiling | none | nothing to build; real faces carry UGC trust |
| Performance data to creative learnings | “Translate performance data into clear creative learnings” | ~2%est. | augment | data readiness | Governed Creative Gate | tagged creative library joined to per-ad metrics so learnings cite numbers |
| Weekly and monthly performance reporting | “Track and report weekly/monthly performance results” | 10% | automate + review | integration gap | tooling augment | ad platform and marketplace exports wired into one weekly pull |
| Month 1 With the creative gate running, month 1 goes to launching campaigns and tests off claims-checked briefs and a wired weekly report; without it, the same month goes to hand-building brief templates and stitching channel numbers. | ||||||
Head of AI Singapore · Founders Office live posting 5 augment 6 automate 1 human
This capability audit is itself the first output of the "Maintain role-task AI-capability audit" row: AI-drafted, human-reviewed before shipping, per that row's own verdict. The review pass rescored two draft verdicts and two friction calls before this version shipped.
| Task | JD quote | Time | Verdict | Friction | Absorbed by | What unblocks it |
|---|---|---|---|---|---|---|
| Rank org-wide AI opportunity portfolio | “build and maintain the prioritised portfolio with clear ROI logic” | ~10%est. | augment | trust ceiling | tooling augment | AI drafts the scoring; leadership trust in the ranking accrues from live wins |
| Pressure-test and kill weak use cases | “kill the low-value ones early” | ~10%est. | augment | trust ceiling | tooling augment | AI drafts the adversarial case; the kill call stays human |
| Ship marketplace creative and content throughput | “creative/content throughput for Blood & NDO across TikTok Shop/Shopee” | ~10%est. | automate + review | data readiness | Governed Creative Gate | ingest Blood/NDO voice and claims corpus; gate itself runs today |
| Build KAM and trade comms workflows | “KAM and trade comms” | ~7%est. | automate + review | data readiness | KAM & Trade Comms | clean sell-through feeds from key accounts before comms argue from data |
| Run market-entry and competitive research | “market-entry and competitive research” | ~7%est. | automate + review | none | Market-Entry Research Agent | |
| Support demand and supply planning | “demand and supply planning support” | ~5%est. | augment | data readiness | tooling augment | clean sales and inventory history first; candidate workflow if planning cluster recurs |
| Stand up finance and reporting workflows | “finance and reporting tasks” | ~5%est. | automate + review | integration gap | Settlement Reconciliation | seller-center payout exports wired into the reconciliation pipeline |
| Deploy SOP and onboarding Q&A | “SOP and onboarding content” | ~5%est. | automate + review | data readiness | SOP & Onboarding Q&A | SOPs consolidated and current before indexing |
| Run stakeholder discovery and iteration | “Use stakeholder time efficiently across user discovery” | ~10%est. | augment | workflow fit | tooling augment | AI preps guides and synthesizes notes; the sessions stay live |
| Coach function leaders to adoption | “training, hands-on pairing, and removing friction” | ~15%est. | keep human | trust ceiling | none | stays human; leaders adopt through pairing with a person they trust |
| Own platform foundations and guardrails | “speed of response by AI, easy human-in-the-loop input and correction” | ~10%est. | augment | integration gap | tooling augment | API access to marketplace seller centers, ERP and comms stack |
| Maintain role-task AI-capability audit | “AI-capability audit that maps opportunities for AI augmentation or replacement” | ~6%est. | automate + review | none | tooling augment | |
| Month 1 With the OS behind them, month 1 ends with the ranked portfolio drafted, the creative gate ingesting the Blood and NDO corpus in week 1, and the market-entry research agent built by week 2 and in daily use; without it, month 1 ends at the portfolio and one prototype. | ||||||
Livestream Host (ID) Jakarta, Indonesia · E-Commerce live posting 3 augment 4 automate 1 human
| Task | JD quote | Time | Verdict | Friction | Absorbed by | What unblocks it |
|---|---|---|---|---|---|---|
| Host live sales sessions on camera | “hosting engaging LIVE sales sessions on TikTok & Shopee” | ~50%est. | keep human | trust ceiling | none | stays human; viewers buy from a live face |
| Build LIVE promos from master pricelist | “Create LIVE promo from the master pricelist” | ~8%est. | automate + review | data readiness | Governed Creative Gate | ID claims register and Bahasa voice corpus seeded into the gate |
| Draft selling scripts and demo tactics | “Plan and prepare new sales demo tactics / scripts” | ~8%est. | automate + review | data readiness | Governed Creative Gate | ID claims register and Bahasa voice corpus seeded into the gate |
| Per-session reporting and insight write-ups | “Meticulously report and analyze each LIVE sales session” | ~10%est. | automate + review | integration gap | tooling augment | per-session metrics export from Shopee/TikTok dashboards |
| Pre-sales viewer chat and Q&A follow-up | “pre-sales engagements with our viewers” | ~5%est. | automate + review | integration gap | Marketplace Chat First Responder | hook Shopee/TikTok Shop chat into the first-responder queue |
| Read live audience, adjust selling in real time | “Monitoring real-time audience interactions” | ~10%est. | augment | integration gap | tooling augment | live comment feed tap for a second-screen top-questions digest |
| Cross-team brand consistency on promo materials | “Collaborating with our marketing, creative and ecommerce teams” | ~5%est. | augment | none | Governed Creative Gate | |
| Monthly schedule planning for peak-hour coverage | “120 Livestream hours/month monthly KPI” | ~4%est. | augment | none | tooling augment | |
| Month 1 Month 1 with the OS, promo sheets and session scripts arrive pre-drafted and price-checked for the host to review, so the 20% planning block goes to refining tactics instead of writing from scratch; the 120 live hours themselves are unchanged. | ||||||
Performance Marketer - NDO Kuala Lumpur, Malaysia · Marketing live posting 7 augment 3 automate 1 human
| Task | JD quote | Time | Verdict | Friction | Absorbed by | What unblocks it |
|---|---|---|---|---|---|---|
| Plan content angles and messaging | “Plan performance-driven content angles, formats, and messaging” | ~15%est. | automate + review | data readiness | Governed Creative Gate | seed an NDO claims and voice corpus; the gate then screens every variant |
| Brief and iterate creatives with teams | “brief, test, and iterate on creatives” | ~10%est. | augment | none | Governed Creative Gate | |
| Translate performance data into creative learnings | “Translate performance data into clear creative learnings” | ~5%est. | augment | data readiness | Governed Creative Gate | tagged creative library joined to per-ad metrics so learnings cite numbers |
| Buy and optimise paid media | “Plan, launch, and optimise paid media campaigns” | ~20%est. | augment | integration gap | tooling augment | AI drafts campaign structures and copy; launch and spend execution stay in the consoles |
| Marketplace ad spend allocation | “marketplace ads e.g. Shopee CPC ads” | ~5%est. | augment | data readiness | Settlement Reconciliation | payout netting exposes per-SKU margin, so bid caps price to real profit |
| Manage budgets, pacing, ROAS targets | “Manage budgets, pacing, and ROAS targets” | ~10%est. | augment | trust ceiling | tooling augment | review-gated pacing and reallocation proposals with an audit trail |
| Design and run structured tests | “Design and run structured tests” | ~10%est. | augment | workflow fit | tooling augment | a shared test log the assistant reads and writes |
| Continuous performance metric analysis | “Continuously analyse performance metrics” | ~5%est. | automate + review | integration gap | tooling augment | same read-only metrics feed as pacing; candidate cluster: weekly ad-metrics digest |
| Weekly and monthly performance reporting | “Track and report weekly/monthly performance results” | ~10%est. | automate + review | integration gap | tooling augment | once exports land, draft reports assemble themselves; human signs the insights |
| Recommendations on positioning and promotions | “Use data to inform product positioning, promotions, and messaging” | ~5%est. | augment | none | tooling augment | |
| Star in own ads | “comfortable starring in your own ads” | ~5%est. | keep human | trust ceiling | none | stays human; real faces carry UGC trust |
| Month 1 With the creative gate seeded for No Days Off, month one ships a tested batch of claim-checked ad variants across Meta and TikTok; without it, those same weeks go to writing briefs and vetting claims by hand. | ||||||
Performance Marketing Manager Singapore · Marketing live posting 4 augment 3 automate 2 human
| Task | JD quote | Time | Verdict | Friction | Absorbed by | What unblocks it |
|---|---|---|---|---|---|---|
| Build, launch, optimise paid campaigns | “plan, launch, and optimise paid media campaigns” | ~25%est. | augment | integration gap | tooling augment | assistant drafts campaign structures and copy; execution stays in the consoles |
| Budget, pacing, ROAS and channel-mix calls | “Make independent calls on channel mix and budget reallocation” | ~15%est. | keep human | trust ceiling | none | AI preps the reallocation brief; spend authority stays with the owner |
| Retargeting segments and audience building | “implement strategies to create impactful audiences” | ~10%est. | augment | data readiness | tooling augment | clean first-party event and purchase data feeding platform audience tools |
| Testing roadmap and hypothesis design | “Own the testing roadmap; prioritise hypotheses” | ~10%est. | augment | workflow fit | tooling augment | a maintained test log the assistant can mine for next hypotheses |
| Daily metric monitoring, scale/kill flags | “monitor ad metrics daily and iterate on ad copy” | ~15%est. | automate + review | integration gap | tooling augment | read-only ad-account pulls into one daily digest; candidate workflow if it clusters |
| Creative briefs and weekly ad variants | “Define the creative strategy and brief for performance content” | ~10%est. | automate + review | none | Governed Creative Gate | |
| Self-editing and appearing in ad creatives | “self-editing ad creatives to speed up iterations” | ~5%est. | augment | none | tooling augment | assistant-class editing tools cut turnaround; on-camera work stays human |
| Weekly and monthly performance reporting | “Track and report weekly/monthly performance results with clear insights” | ~7%est. | automate + review | integration gap | tooling augment | scheduled channel pulls into a fixed template the hire signs off |
| Leadership POVs and positioning input | “bring sharp POVs to leadership on what's working” | ~3%est. | keep human | trust ceiling | none | AI supplies the evidence pack; the POV has to come from the person in the seat |
| Month 1 With the creative gate already running, this hire ships claim-checked briefs and weekly ad variants from week one and spends month 1 on budget calls and the test roadmap; without it, month 1 goes to drafting and vetting copy by hand. | ||||||
Regional Demand Planning Manager Kuala Lumpur, Malaysia · Supply Chain live posting 6 augment 4 automate 2 human
| Task | JD quote | Time | Verdict | Friction | Absorbed by | What unblocks it |
|---|---|---|---|---|---|---|
| Rolling demand forecast build and consolidation | “Develop, maintain, and finalize rolling demand forecasts” | ~20%est. | augment | data readiness | tooling augment | one clean multi-market sales history with promo and launch flags |
| Forecast risk early-alert flagging | “raising early alerts to commercial and supply stakeholders” | ~5%est. | automate + review | data readiness | tooling augment | live actuals-vs-forecast feed so drift auto-drafts alerts for planner review |
| Monthly S&OP cycle ownership | “Drive the monthly Sales & Operations Planning cycle by country” | ~10%est. | augment | workflow fit | tooling augment | assistant drafts pre-reads and decision logs; facilitation stays with the manager |
| KPI dashboards and standardized reporting | “Track forecast accuracy and publish performance dashboards” | ~10%est. | automate + review | integration gap | tooling augment | scheduled pipeline from ERP and channel exports into standing dashboards |
| Daily inventory exception monitoring | “Monitor inventory health daily” | ~10%est. | automate + review | integration gap | tooling augment | daily SKU-by-market stock feed across channels driving one exception report |
| Stock policy and Out-of-Stock root cause | “Set and govern safety stock, reorder points, and Weeks of Cover” | ~10%est. | augment | trust ceiling | tooling augment | AI proposes parameters and root-cause packs; manager signs off on the trade-offs |
| Regional interface with China factory teams | “primary regional interface with the Procurement and Quality teams in China” | ~10%est. | keep human | trust ceiling | none | stays human; Mandarin supplier negotiation and relationship trust carry the outcome |
| Supply scenario simulation | “Run simulations to evaluate the supply impact of demand shifts” | ~5%est. | augment | data readiness | tooling augment | lead times, capacity, and stock positions in one model AI can rerun on request |
| Planning SOPs and coordinator self-serve Q&A | “Establish standard operating procedures for finished goods planning” | ~5%est. | augment | data readiness | SOP & Onboarding Q&A | SOPs written and kept current, then loaded into Q&A for coordinator self-serve |
| Mentoring market planning coordinators | “Guide and mentor local market planning coordinators” | ~5%est. | keep human | trust ceiling | none | stays human; coaching judgment and team trust do not delegate |
| China outbound shipment coordination | “Ensure timely shipping of finished goods from factories in China” | ~5%est. | augment | integration gap | tooling augment | forwarder and factory milestones in one tracker; expediting calls stay human |
| Factory document matching for finance reconciliation | “Documentation & Financial Reconciliation: Check and match shipping and commercial” | ~5%est. | automate + review | data readiness | Settlement Reconciliation | factory invoices and shipping docs extracted into the matching engine |
| Month 1 With the OS behind them this hire ends month 1 with a standing forecast-accuracy dashboard, a daily inventory exception report, and reviewed factory-document matching; without it the same weeks go to assembling those spreadsheets by hand. | ||||||
Regional Head of Brand & Innovation Singapore · Marketing live posting 6 augment 3 automate 3 human
| Task | JD quote | Time | Verdict | Friction | Absorbed by | What unblocks it |
|---|---|---|---|---|---|---|
| Innovation roadmap and white-space scanning | “Building and maintaining the regional innovation and renovation roadmap” | ~10%est. | augment | data readiness | Market-Entry Research Agent | centralize review, CX and category sources the research agent can cite |
| Product and proposition brief writing | “Writing sharp product, packaging and proposition briefs” | ~6%est. | augment | data readiness | Governed Creative Gate | AI drafts structure and evidence blocks; the proposition stays with the hire |
| NPD ownership from concept to shelf | “Owning the NPD journey from concept to shelf” | ~8%est. | keep human | trust ceiling | none | stays human: launch accountability and cross-functional trade-offs need an owner |
| Launch operations tracking and readiness checks | “Managing the operational reality of physical-product launches” | ~6%est. | augment | integration gap | tooling augment | artwork, barcode and SKU status sit in ERP and factory systems with no shared feed |
| Competitive tracking and consumer insight agenda | “competitive tracking and the consumer insight agenda” | ~8%est. | automate + review | none | Market-Entry Research Agent | |
| Brand, portfolio and pack-price strategy | “brand guardrails, portfolio architecture, SKU roles, claims hierarchy” | ~12%est. | keep human | trust ceiling | none | stays human: portfolio bets set margin and channel economics, founder-level judgment |
| Authoring the marketing operating system | “NPD stage gates, product briefs, packaging briefs, claims checklists” | ~10%est. | augment | none | tooling augment | |
| Launch toolkit and sell-in asset production | “product one-pagers/explainers, sell-in narratives and claims decks” | ~12%est. | automate + review | data readiness | Governed Creative Gate | substantiated claims register per market feeds the deterministic gate |
| Country enablement and post-launch learning | “sharing best practices across SG, MY, ID, TH, IN” | ~8%est. | augment | data readiness | SOP & Onboarding Q&A | the SOPs and playbooks this hire writes become the RAG corpus |
| Design governance and asset approval | “creative asset governance and design approval standards” | 10% | augment | workflow fit | Governed Creative Gate | gate screens claims and copy compliance; shelf impact and taste calls stay with the hire |
| CX triage and voice-of-customer reporting | “turn customer complaints, reviews, DMs, marketplace feedback and service issues” | ~7%est. | automate + review | integration gap | Marketplace Chat First Responder | Shopee, TikTok Shop and brand.com inboxes wired into one triage queue |
| Cross-functional leadership and standards enforcement | “strong enough to enforce standards when needed” | ~3%est. | keep human | trust ceiling | none | stays human: enforcement runs on personal authority with GMs and founders |
| Month 1 In month 1 the hire ships the first claims-gated brief and sell-in pack plus a source-cited TH and IN category scan with the OS behind them; without it those same weeks go to writing templates and chasing sources by hand. | ||||||
Senior Brand Manager Bangkok, Thailand · Marketing live posting 5 augment 2 automate 3 human
| Task | JD quote | Time | Verdict | Friction | Absorbed by | What unblocks it |
|---|---|---|---|---|---|---|
| Thailand brand roadmap and narrative | “Translate Blood’s regional brand strategy into a Thailand-specific marketing roadmap” | ~5%est. | keep human | trust ceiling | none | positioning bets stay with a person; OS supplies cited inputs |
| Whitespace and local trend research | “Identify whitespace opportunities, category tensions, and local trends” | ~10%est. | automate + review | none | Market-Entry Research Agent | |
| 360 campaign creative and briefs | “Lead nationwide 360° brand campaigns across digital, social, influencer” | ~20%est. | augment | none | Governed Creative Gate | |
| Trial activations and sampling execution | “drive consumer trial and product demos such as campus activations” | ~10%est. | keep human | workflow fit | none | physical event and vendor work; AI drafts run-sheets only |
| KA sell-in pitches and JBPs | “develop sell-in pitches with marketing plans and consumer insights” | ~15%est. | augment | data readiness | KAM & Trade Comms | retailer sell-through feeds cleaned into one queryable table |
| Promo plans, decks, POSM toolkits | “Develop retailer JBPs, promo plans, campaign decks and POSM toolkits” | ~8%est. | augment | workflow fit | Governed Creative Gate | add trade-deck and POSM templates to the gate corpus |
| Shelf and PDP visibility checks | “planograms, promotions, extra displays, in-store activations, PDP, brand pages” | ~7%est. | augment | data readiness | tooling augment | store-audit photos and OSA reports centralized and queryable |
| Category, competitor, NPD insight gathering | “Regularly gather insights (both quals and quants) on category trends” | 15% | augment | none | Market-Entry Research Agent | agent covers the desk research; quals fieldwork and synthesis stay with the hire |
| Marketing performance reporting | “Track, analyze and report marketing performance metrics” | ~5%est. | automate + review | integration gap | tooling augment | ad, social, and retail metrics landed in one view |
| Agency and HQ management | “Manage external marketing and media agencies” | ~5%est. | keep human | trust ceiling | none | agency negotiation and budget accountability stay with a person |
| Month 1 With the OS behind them, month 1 ends with a source-cited Thailand category and competitor map plus gate-checked Thai campaign and trade copy ready for the first KAM sell-in; without it, those same weeks go to hand-assembling research and first-draft decks. | ||||||
Social Media & Content Executive (FMCG) Kuala Lumpur, Malaysia · Marketing live posting 6 augment 4 automate 2 human
| Task | JD quote | Time | Verdict | Friction | Absorbed by | What unblocks it |
|---|---|---|---|---|---|---|
| Creative direction and full-funnel messaging | “owning headlines, scripts, and messaging across Instagram, TikTok, emails” | ~10%est. | augment | none | Governed Creative Gate | |
| On-camera short-form video creation | “Create and appear in short-form video content” | ~15%est. | keep human | trust ceiling | none | stays human; a real face carries a period-care brand's credibility |
| Filming, production, and editing | “Film and produce content using mobile-first formats” | ~15%est. | augment | none | tooling augment | |
| Organic content calendar planning and scheduling | “Own and drive Blood’s organic social content calendar” | ~10%est. | augment | none | tooling augment | |
| Captions, emails, scripts, visual copy | “Write compelling captions, emails, scripts, and visual copy” | ~10%est. | automate + review | none | Governed Creative Gate | |
| Content performance tracking and reporting | “Track and report on creative and content performance” | ~5%est. | automate + review | integration gap | tooling augment | one consolidated weekly export from TikTok, Instagram, and ads managers |
| Trend and platform monitoring | “Stay ahead of platform best practices and cultural trends” | ~5%est. | augment | workflow fit | tooling augment | assistant-compiled digests from saved links; feed instinct stays human |
| Creator and affiliate identification | “Identify and collaborate with relevant creators, affiliates, and micro-influencers” | ~5%est. | augment | workflow fit | Market-Entry Research Agent | a creator-scouting variant of the cited-research agent |
| Creator outreach and campaign coordination | “Coordinate outreach, onboarding, briefing, product seeding, campaign timelines” | ~10%est. | augment | workflow fit | tooling augment | creator threads logged in one tracker an assistant can draft from |
| Affiliate campaign kits and guidelines | “Prepare content guidelines, key messages, offer mechanics, and campaign kits” | ~5%est. | automate + review | none | Governed Creative Gate | |
| Affiliate delivery and metrics tracking | “Track affiliate content delivery and basic performance metrics” | ~5%est. | automate + review | integration gap | tooling augment | standing export of affiliate posts and metrics into one sheet |
| Creator relationship building | “Build strong working relationships with external creators and affiliates” | ~5%est. | keep human | trust ceiling | none | stays human; creators show up for people they know |
| Month 1 With the creative gate running, this hire ships a month of voice-checked captions, scripts, and affiliate kits in the first week and spends the reclaimed hours on camera; without it, most of month 1 goes to blank-page drafting and manual formatting. | ||||||
The systems that absorb the work
six production lines · one runs today
Six workflows, each serving named goals and absorbing rows from the audit above. One runs today; five are the production designs behind it.
Governed Creative Gate WORKING DEMO
Corpus-grounded briefs and variants behind a deterministic claims and voice gate. The drafting and gate engine runs today on Blood and NDO's public market corpus; internal voice and claims ingest is week-1 work.
This workflow now runs: see it work at /agents updated 20 Jul 2026 · all other content as built 11 Jul
Brief and corpus
campaign ask plus claims register and voice guide
DEMO
Variant drafting
briefs, copy, scripts grounded in the corpus
Claims and voice gate
versioned rules; deterministic pass or fail
Owner review
named reviewer approves the weekly batch
Ship-ready creative
gated variants to Meta, TikTok, marketplaces
Working demo, runs todayProduction designHuman gateDeterministic gate
Market-Entry Research Agent
Source-cited competitive and regulatory research, verified by a human before it circulates. Assemble: model plus search and filings APIs; the citation gate is the part worth building.
Research question
market, regulatory, or competitor brief request
Source fan-out
agent pulls filings, listings, category and press sources
Cited drafting
every claim written with its source attached
Citation gate
claims without a live source fail the draft
Analyst verification
human spot-checks sources before it circulates
Cited brief
regulatory and competitor map ready to act on
Production designHuman gateDeterministic gate
KAM & Trade Comms
Retailer and JBP comms argued from sell-through data. Build on the data spine; the drafting is generic, the number check is the part worth building.
Sell-through feeds
retailer EPOS and marketplace data off the spine
Account read
per-account performance, gaps, and promo effects
Comms drafting
JBP decks and trade emails argued from the numbers
Number check
every figure traces back to the feed
KAM sign-off
account owner edits, owns the relationship, sends
Retailer-ready comms
sell-in stories with data behind them
Production designHuman gateDeterministic gate
Settlement Reconciliation
Marketplace payout netting and per-SKU profitability decomposition. Buy the parsing where recon SaaS fits; build the per-SKU margin decomposition.
Payout and cost files
marketplace settlements, factory invoices, 3PL bills
Extraction
line items parsed into structured records
Matching engine
deterministic netting of orders to payouts to fees
Margin decomposition
per-SKU profit after fees, promos, logistics
Exception review
finance owner clears unmatched lines
SKU profitability
true margin per SKU per channel, monthly
Production designHuman gateDeterministic gate
Marketplace Chat First Responder
Chat triage across Shopee, Lazada, TikTok Shop: template intents auto-send after the gate, sensitive ones wait for a human. Buy the queue from helpdesk tooling; build the policy gate.
Chat inboxes
Shopee, Lazada, TikTok Shop in one queue
Intent triage
order status, product question, complaint, urgent
Reply drafting
answers grounded in product and policy corpus
Policy gate
health claims and refund promises checked; PII stripped
Tiered review
health, refund, complaint intents wait for a human
Reviewed replies
template intents auto-send; the rest go out human-approved
Production designHuman gateDeterministic gate
SOP & Onboarding Q&A
Internal Q&A over SOPs and brand guidelines with cited passages. Assemble from off-the-shelf RAG parts; build the source gate and owner routing.
SOP corpus
SOPs and brand guidelines, versioned and current
Indexing
documents chunked for retrieval with owners tagged
Grounded answers
each answer cites the exact passage
Source gate
no matching passage means no answer
Owner escalation
unanswered questions route to the function owner
Self-serve answers
new hires and coordinators unblock themselves
Production designHuman gateDeterministic gate
Order of attack
- 1creative gate, corpus ingest
- 2research agent
- 3SOP Q&A
- 4chat responder
- 5settlement recon
- 6KAM comms
The first 90 days ship the top two: the gate touches 19 rows across every marketing seat and already runs; the research agent unblocks the India and Indonesia briefs.
ranked by rows absorbed, hours touched, data readiness · build-week figures are sizing estimates, revised in the monthly re-rank
Foundations
build once, every workflow reads from them
The corpus mechanic behind the creative gate, generalized: each function keeps its own store of ground truth, with campaign data treated as the best record of what customers actually respond to.
- One store exists today: the creative gate's corpus, about 1,600 provenance-tagged public reviews curated from a larger public pull (full corpus published); it is the template for the rest.
- Week-1 form per function: a shared folder of plain files with one named owner and a weekly append ritual; a database only when several writers or query needs arrive.
- Every AI output cites the store entry behind it; evidence lands with a provenance tag (organic beats sample beats gifted).
- Campaign results append automatically once readback ships; results are customer feedback with money attached.
Sell-through unified across every sell-point (marketplaces, retail EPOS, brand.com) is the dependency most workflows on this page name; it gets built once and every workflow reads from it.
- The grain is SKU by channel by week; one table, and every workflow reads from it.
- Sources: seller-center exports (Shopee, Lazada, TikTok Shop), retailer sell-through files (Watsons, Guardian and peers), brand.com orders.
- The unglamorous heart is a SKU naming map across channels, built by hand first; every channel names the same product differently.
- Weekly discipline before infrastructure: scheduled pulls where APIs exist, manual uploads where they don't, reconciled against finance settlements so the spine agrees with money.
Nothing reaches a customer or a retailer without a versioned rule set and an eval it must pass first; the creative gate already runs this way.
- A rule is five fields: id, pattern, reason, the evidence behind it, and a named owner.
- Agreement flow: drafted from evidence, signed by the function owner; voice rules belong to the Head of Brand, claims rules to a named compliance owner.
- Every change ships with a version bump and a test case; a change that breaks an existing fixture cannot ship.
- Disagreement becomes a rule-change proposal with an owner and a version number; refusals cite the rule id, and false flags are a tracked metric reviewed monthly.
Health-adjacent customer data gets role-scoped access and PII stripping before any model sees it, built into the workflow rather than written into a policy document.
- Each workflow reads only its own function's store; access is scoped to the role.
- Names and contact details are stripped before any model call; the pattern survives without the identity.
- Period care and baby categories are health-adjacent, so they sit in the strictest tier by default.
- The guardrail lives in the code path rather than in a policy document; the system cannot leak what it never receives.
Run weekly
each mechanism starts when its first workflow ships
The cadence the Head of AI installs. Each mechanism starts when its first workflow ships; none of them run today.
Weekly
Weekly batch review
AI output ships in reviewed batches on a fixed weekly cadence, each batch signed off by a named reviewer.
Weekly
Adoption from telemetry
Per-function usage comes from telemetry; a workflow nobody runs gets fixed or retired.
Monthly
Monthly re-rank
Each workflow reports its numbers into the portfolio; once a month the ranks get re-argued against them.
Monthly
Function-leader scorecards
Each function leader holds one scorecard: the metric their workflow moves, the adoption behind it, and the next unblock.
Day 90
Runbook handover by day 90
Every workflow ships with a runbook; by day 90 the function owner runs it alone.
How this stays alive: every verdict on this page is re-scored quarterly against the same rubric; a model-capability change that moves verifiability or cost triggers a re-score out of cycle; once workflows are in production, telemetry feeds the verdicts, and what teams actually run outranks what the audit predicted.