TINA  /  T9 Intelligent Agents  /  Tangent 9
01 / 23
Seed round · S$2.0M · 2026

A workforce of
AI agents, managed
like a group chat.

TINA gives a person, a family or a company a team of AI staff. They plan and carry out whole jobs across your tools, act only with your permission, and every action is recorded.

Agents as contacts Whole jobs, done Mini app marketplace Permission scoped Human approvals
FA
Ops team
4 people · 2 agents
Priya@Finance Agent pull the overdue invoices over 30 days
Finance AgentFound 7, total S$18,420. I can send reminders from your Xero account.
Send them, but not to Acme, I am handling that one
Approval required
Send 6 payment reminders via Xero, acting as Marcus Lee.
ApproveReject
Approved
6 sent · logged · reversible
Tangent 9 Pte Ltd  /  UEN 202541051K  /  Singapore
What TINA is

An AI workforce, plus the
engine that runs their work.

TINA is an agentic AI platform delivered as a messaging app. It gives you a team of digital staff: the agents appear as contacts, you brief them the way you would message a colleague, and behind the chat a workflow engine plans the job, checks what each agent is allowed to do, pauses for your approval and records every step. The chat is not the product. It is simply the easiest way ever invented to manage a workforce.

01

A workforce, not one bot

Many specialist agents: finance, admin, research, family life. Each has skills, a schedule and limits that you set.

02

Whole jobs, not answers

One brief becomes a plan with steps, carried out across your calendar, accounts, files and customer records.

03

Your rules, always

An agent can only do what you are allowed to do. Important steps pause for a named human, and everything is logged.

04

Zero learning curve

It works like the messaging apps people already use every day. If you can message, you can manage a workforce.

One platform serves one person, a family, or a company of thousands. Only the settings change, never the software.
Vision

Everyone should be able to hand
real work to an AI, safely.

Handing work to someone has always needed three things: trust, clear limits, and a record of what was done. We are building those three things for AI, so giving a job to an agent feels as normal and as safe as giving it to a colleague.

TODAY

A governed workspace

People and AI agents doing real work side by side, under per-person permissions and human approvals.

NEXT

A creator layer

Anyone can build and deploy an AI-powered mini app into that workspace. Approved partners can sell theirs in the marketplace, with safety enforced by the platform.

THEN

A governed economy

A marketplace where those apps sell products and services, under the same permission and approval rules.

The problem

AI is built for one.
Work is done by many.

Today's assistants serve one person at a time. That is fine while the AI writes emails. It breaks the moment the AI does something real: pays a supplier, updates a customer record, approves a leave request. Someone has to answer for what it did.

01

Work is shared. AI is not.

Running a home or a small business is teamwork. But today's AI talks to one person at a time, in private. Everyone gets their own separate AI that knows nothing about what the others have planned, so the same things get explained again and again.

02

Shared AI chat came, and went.

ChatGPT piloted group chats in late 2025 and retired them within a year. A shared chat that can only talk gives people no reason to stay. The job is doing the work, safely.

03

Safe AI costs too much.

The enterprise options assume an IT team, per-seat licences and a stack of identity and policy setup before anything is safe. Most small businesses here have none of that.

The one question today's tools cannot answer: under whose authority did the agent act?
Existing solutions

What others already do,
and what nobody does

ALREADY SOLVED BY OTHERS
  • AI agents inside team chat. Microsoft lets you add agents to Teams chats, and each answer respects what the asker is allowed to see.
  • A store for AI agents. Microsoft already runs a store where partners publish and sell AI agents to businesses.
  • Scheduled AI tasks. Daily briefings, reminders and recurring jobs are now standard in the big assistants. Table stakes, not a differentiator.
STILL UNSOLVED
  • A shared AI consumers actually kept. ChatGPT piloted group chats in November 2025 and retired them from July 2026. Talk-only did not stick. The version that does real work has only ever shipped for enterprises with IT teams, never for the rest of us.
  • Governed AI without the IT project. Microsoft agents can reach outside tools now, but making that safe is assembly required: licences, identity, policy setup, and the risk handling is on the customer. Nobody ships it safe out of the box.
  • An AI a family can actually share. No one offers a family AI with a shared workspace, spending limits per member, and parental control over what the AI may do.
Sources: OpenAI Help Center, group chats retirement notice, July 2026; Microsoft Teams and Copilot documentation, 2026.
For families and individuals

A life organiser,
not another chatbot

Most people do not want to chat with an AI. They want the school run remembered, the bills paid on time and the weekend planned. TINA comes with ready-made mini apps, staffed by its agents, so from day one it feels like a family organiser.

Family Calendar

"Who is picking up the kids on Thursday?" It checks everyone's week and books it in.

Bills & Renewals

"Car insurance renews in 12 days. Last year you paid S$1,140. Renew, or compare?"

Meals & Groceries

Plans the week's dinners, builds the shopping list, and reorders the staples.

Family Wallet

Tracks household spending and pocket money, and who paid for what.

Rewards

Earn points at partner shops and cafes, and redeem them right inside the chat.

Memory Vault

"When does Mum's passport expire?" Passports, policies and warranties, remembered.

Trip Planner

"Plan the June holiday." Flights, stays and a day-by-day plan, agreed in the family chat.

Family Health

"Dad's dental check is due." Appointments, jabs and medicine reminders for the whole family.

Home & Chores

"The aircon service is due again." Repairs, chores and the helper's schedule, organised.

FREE TO TRY

We are in discussions with AI providers to fund free starting credit, so a new family can try TINA at no cost, to them and largely to us.

For small businesses

Four everyday apps,
free for the first year

A small business does not buy "AI". It buys fewer chores. TINA launches with four business apps every small firm needs, staffed by its agents and free for the first year, in exchange for one thing.

CRM Lite

A simple customer system: members, sales opportunities and reward balances in one place. The agent keeps it current and chases the follow-ups.

Loyalty & Rewards

Points and perks that bring repeat customers back, run entirely from the chat.

Basic HR

Leave requests, rosters and approvals, handled inside the team chat.

Expense Claims

Snap the receipt. The agent files the claim. The boss taps approve.

THE DEAL

All four apps are free for year one when the business switches on our rewards engine for its customers. That one condition powers our growth, as the next slide shows. Free is the front door, not the business: paid plans for more agents, staff and tools sit behind it.

How the two sides feed each other

The rewards engine
that sells for us

One condition on the free business pack sets a loop in motion.

STEP 1

A shop joins free

It gets the four business apps and the rewards engine at no cost.

STEP 2

Shoppers earn rewards

Its customers collect points every time they spend there.

STEP 3

Shoppers join TINA

To track and redeem their points, they download the app.

STEP 4

They stay for the apps

The family organiser gives them a reason to keep it.

STEP 5

More shops follow

A growing audience makes joining worth it for the next shop.

EACH SIDE RECRUITS THE OTHER

Shops bring in shoppers. Shoppers make the network worth joining for shops. So the cost of winning each new customer falls as the network grows, instead of rising.

The same loop exists in loyalty and payment apps across the region. Ours differs in what sits behind it: a full AI workforce, not just a points card.
Product demo

One brief, carried out
end to end

A live prototype of all three surfaces is available to walk through. Every instruction takes the same five steps. Nothing skips the check, and nothing important skips the human.

01

Brief

A user types a request in a thread, or mentions an agent in a group.

02

Plan

TINA works out what is needed, picks the right agent and plans the steps.

03

Check

TINA checks what that person is allowed to do, tool by tool, action by action.

04

Approve

Anything important asks a human first. Who decided, and when, is recorded.

05

Do

The agent does the work in the connected tool. Every step is logged.

The user app

Web, iPhone and Android. Agent contacts, group threads, approval cards, tool connections and a mini app store.

The partner portal

Where partners submit apps, set prices, track orders and see exactly what they earn and when they get paid.

Our control centre

Where we manage customers, billing, partner sign-ups, app reviews, pricing and support, with a full activity log.

Built, and available to demonstrate today.
The core idea

The agent is never
more powerful than you

Every agent works under the identity and access of the person asking. In a group chat, five people get five different results from the same agent, because it borrows each person's access in turn, never everyone's access combined.

  • An agent can only do what the asking person is allowed to do, checked on every single action.
  • Important actions pause for a named human, and the decision is recorded.
  • Information coming back from connected tools is treated as data, never as instructions.
  • Every action is written to a permanent log that cannot be edited.
Governance

Built to Singapore's new
AI safety framework

In January 2026 Singapore's IMDA launched the world's first safety framework for AI agents. TINA was designed to its four principles before it was published. In this region, that alignment opens doors rather than adding cost.

Know and limit the risks

Every agent, tool and app must be granted to a person before they, or their agent, can even see it.

Keep humans accountable

Payments, permanent changes and outside messages pause for a human. Who decided, and when, is recorded.

Build in safeguards

Rules and spend caps on every action, defence against hidden malicious instructions, and an off switch per app.

Keep users in control

Users see what each app may do before it runs, watch agents work in the thread, and can review the full log.

PDPA baseline ISO/IEC 27001 aligned ISO/IEC 42001 readiness PCI scope at the payment provider
The framework is voluntary guidance with no certificate to hold. The claim is design alignment, principle by principle, not formal compliance.
Why this is hard to copy

Six reasons competitors
cannot simply follow

We hold no patents and we do not claim a smarter AI model. Our protection comes from how the product is built, where it is built, and what accumulates inside it.

Hard to retrofit

Per-person permissions cannot be bolted onto a product built for one user. Consumer AI companies would need to rebuild their foundations.

Regulatory head start

Designed to Singapore's AI safety framework from day one, with international security standards informing the build.

Hard to leave

A customer's tool connections, agents, approval rules, history and app subscriptions all live inside TINA.

Not tied to one AI

TINA works with any AI model and switches as better or cheaper ones appear. Falling AI prices improve our margins.

A complete blueprint

Over 150 documented requirements, three working prototypes and a full permission design, before production code.

A two-sided pull

Every partner app makes TINA more useful to customers, and every customer makes TINA more attractive to partners.

Stated honestly: none of these is a patent. Together they are a two to three year rebuild for anyone starting from a single-user product.
Market opportunity

A market big enough to matter,
a slice small enough to win

The regionS$11B

Over 71 million small businesses across Southeast Asia, plus every household in those markets. Everyone who could one day run an AI workforce.

Our first marketsS$1.0B

Digitally active small businesses in Singapore and Malaysia. Singapore alone has over 300,000 businesses, 99 percent of them small.

Our 5-year aimS$10M

About 4,000 paying customers, still under 1 percent of our first markets, won through resellers across the region.

Sized from the bottom up: customers multiplied by what each pays per year. The S$2M raise funds the first S$1.4M of this by year three; the aim is the trajectory it puts us on.

The region: ~7M digitally ready businesses x S$1,200 a year plus ~7.5M households. First markets: ~510k active small businesses x S$1,800. Aim: ~4,000 customers x ~S$2,500.
Beachhead

Start narrow,
then grow wide

FIRST

Small businesses in Southeast Asia

5 to 200 staff, on Xero, Google and everyday tools. They want AI that does real work, cannot afford Microsoft's option, and have no IT team.

NEXT

Families and individuals

The same product with simpler settings. Shared usage, spending limits per member, parental controls.

THEN

Big companies and resellers

Staff directory sync, private databases and local data storage, sold through regional resellers.

No family plan exists

ChatGPT sells to individuals, or to teams at USD 30 per person per month. There is no household plan at all.

Sharing a plan is not sharing a workspace

Google's family option now shares the subscription and even pools some usage credits. What no one shares is a workspace: no family threads, no per-member limits, no say over what the AI may do.

Same product for us

A family is simply a small team with friendlier settings. Underneath, it is exactly the same platform as a company.

Business model

Five ways we earn

Steady subscription income pays for the platform. The marketplace and payment fees are what make the business compound over time.

01 · Subscriptions

Plans for every customer type, from one person to a big company, with AI usage included.

02 · AI usage top-ups

Heavy users buy extra AI usage in top-up packs, instead of being cut off.

03 · Marketplace share

Partners sell mini apps at their own prices. We keep 20 percent of what subscribers pay.

04 · Payment fees

When mini apps sell products or services, we earn a small fee on every transaction.

05 · Resellers

Regional resellers sell TINA at wholesale pricing in markets we do not enter directly.

Next: a sixth rail

The AI Delivery Pod adds a build fee plus metered usage when customers create apps of their own. Not counted in this plan.

WHY IT COMPOUNDS

Every partner app makes TINA more useful to customers, and every customer makes TINA more attractive to partners. Our safety system is what lets outside apps in without letting outside risk in.

Go to market

Three motions, sequenced

Two free front doors bring people in. The paid platform sits behind them.

01 · FAMILIES

They come for the apps

  • Ready-made lifestyle mini apps from day one
  • Free starting credit via AI-provider partners
  • Rewards at local shops pull families in
  • Households spread it by word of mouth
02 · BUSINESSES

They come for the pack

  • Four business apps, free for year one
  • In return, they switch on the rewards engine
  • Resellers hand it to clients they already serve
  • Target: 25 paying customers in 12 months
03 · THE REGION

Grants and resellers

  • Singapore's MRA grant supports Malaysia entry
  • Wholesale pricing for regional resellers
  • Marketplace partners bring their own users
  • Cost per customer ~S$800, falling with the network
We rarely pay cash to win a customer: AI partners fund the trials, merchants recruit shoppers, resellers bring their clients, and a grant part-funds the region.
Competition

How we compare, honestly

Consumer AI
ChatGPT, Gemini
Enterprise agents
Copilot, Agentforce
Work chat
Slack, Teams
TINA
Group chat with people and AI togetherretired
Agent limited to the asker's own access··
Outside tools safe out of the box, no setup project···
AI agents without an IT department··
Family plan with a shared workspace and parental controls···
AI app store where partners earn from families and small firms···
Human approvals and a log of every action··
Publicly documented features, August 2026. ChatGPT retired its group chat pilot from July 2026; its new app directory reaches consumers, but partners cannot yet sell subscriptions inside it. Google's family sharing pools a bill and some credits, not a workspace. Microsoft's protections are real; extending them beyond Microsoft is the customer's project.
Traction

Two earlier products
already proved the idea

PALS proved businesses will let AI agents do real work. YAPA proved individuals will run their lives through a chat assistant. Neither was built to serve many customers at scale, with billing and a partner store. TINA is that rebuild.

2

Shipped products

Behind TINA, with real users and paying customers.

3

Working prototypes

User app, partner portal and control centre, demonstrable today.

5

Revenue rails

Designed into the platform from day one, not bolted on later.

S$0

Outside money raised

Everything above was built founder-funded.

A complete build blueprint, covering permission design, data model and the integrations list, was finished before production code begins. The build partner is engaged.

Financials

The numbers, and
where they come from

Figures are yearly income levels reached by each year end. Revenue = customers x plan price + AI usage top-ups + 20% of marketplace sales + payment fees.

Key measureYear 1Year 2Year 3Year 4 aimYear 5 aimDriver
Paying customers25150550~1,500~4,000Direct, then resellers
Revenue per customer per yearS$1,800S$2,000S$2,200~S$2,350~S$2,500Staff count and plan
Yearly subscription incomeS$45kS$300kS$1.2M~S$3.5M~S$10MCustomers x price
Marketplace and payment revenueS$5kS$30kS$150k~S$350k~S$800kPartner apps live
Gross margin after AI cost65%72%78%~80%~80%Usage metering
Monthly spend, netS$70kS$80kS$95knext roundnext roundHeadcount
Assumes S$15 per person per month, average customer 10 staff, launch Q4 2026, resellers from 2028. Years 1 to 3 are the funded plan. Years 4 and 5 assume the next round closes and the reseller network scales: aims, not commitments. The market slide's S$10M aim counts subscriptions only; marketplace revenue is upside on top. Full spreadsheet model on request.
Roadmap

The road ahead

NOW · PLATFORM V1

Launch and prove

  • All three apps launched
  • Tool connections live
  • Billing and plans running
  • First partner apps in the store
NEXT · AI DELIVERY POD

Build by describing

  • Create a mini app by describing it
  • Partners ship apps faster
  • Companies build private apps
  • Safety enforced by the platform itself
THEN · COMMERCE

Apps that sell

  • Mini apps that sell goods and services
  • Money held until delivery is confirmed
  • A fee earned on each transaction
  • Subject to legal review on payments
LATER · REGION

Scale through others

  • Resellers across the region
  • Local data storage per country
  • Private setups for big companies
  • Marketplace feeds its own growth
Each stage is funded by the one before it. Nothing on this list requires a new platform.
Risks, stated plainly

What could go wrong,
and our answers

Every early investment carries risk. These are the four we watch most closely.

Building too much at once

Three apps plus billing and payments in twelve months is a lot.

Our answerPhased delivery tied to grant milestones, and a strict written list of what version one will not include.

An empty marketplace

A store with no named partners is a promise, not a product.

Our answerSigning launch partners now. Two named partners at close is a condition we set ourselves.

Attacks that trick the AI

Hidden malicious instructions inside data are the most serious risk for AI that does real work.

Our answerTool output treated as data, never instructions. Risky actions default to human approval. An off switch per app.

The giants move down-market

Microsoft or OpenAI could come after small businesses.

Our answerNeither knows how to sell to, or price for, a five-person shop on mixed tools. Our protection is who we can reach, and how cheaply.

Team

Who is building this

Two shipped products and three working prototypes, delivered before raising a dollar.

FOUNDER AND PRODUCT LEAD

Tan Han Kiong

Twenty years building enterprise and consumer platforms: product head for enterprise cybersecurity at ST Engineering, led a 60-strong product and engineering organisation at SPH Media, and shipped a SaaS loyalty platform adopted by retail groups across Southeast Asia. Conceived and shipped PALS and YAPA, the two products TINA rebuilds as one platform.

linkedin.com/in/hk13

ENGINEERING LEAD

Hmu

More than 15 years building scalable enterprise systems for high-traffic workloads. Leads the TINA version one build single-handedly: the governed AI orchestration service with its agent tooling and retrieval (RAG) engine, the permission engine, billing, tool connections and all three apps, delivered against a fully specified blueprint of over 150 requirements.

COMMERCIAL AND PARTNERSHIPS

Augustine Ler

A technologist turned business developer with more than 30 years of BD experience, who sold the first company he founded for USD 5M. Runs channel businesses in connectivity as chief executive of Solutions4u Global and eSIM4U, selling through distribution partners across the region. Owns TINA's reseller and partner rails.

linkedin.com/in/augustine-ler-5638b92

BUSINESS DEVELOPMENT PARTNERS

Chew Boon Hong · BD and presales. Enterprise IT veteran across Beyondsoft and NCS, with hands-on enterprise AI delivery in Singapore's banking sector.

Ng Boon Teck · BD, Singapore and Vietnam markets. Digital transformation and business development, based in Vietnam and networked through the Singapore business community in Hanoi.

The ask

What we are asking for,
and what it buys

RAISING
S$2.0M

Funds 24 months of build and launch: S$1.8M planned spending plus a S$200k reserve.

USE OF FUNDS
Building and launching all three apps55%
Winning customers in Singapore and Malaysia20%
Signing partners and opening the marketplace15%
Security and safety readiness10%
WHAT THIS RUNWAY PROVES
Q4 2026Version one launched across all three apps
Q1 2027AI mini app builder launched
Q4 202725 paying customers in Singapore, with named logos
Q2 2028First 10 partner apps live and earning in the marketplace
Q3 2028Reseller agreement signed in Malaysia
Q4 2028S$330k in yearly recurring income, the proof needed to raise the next, larger round
Closing

The model is not the moat.
The accountability around it is.

AI models get cheaper and better every quarter, and none of that answers who is responsible for what the agent did. TINA is building that answer, and everything else plugs into it.

TINA  /  T9 Intelligent Agents  /  Tangent 9 Pte Ltd   ·   hankiong@tangent9.com  /  +65 8868 1386  /  tangent9.com
Speaker notes · press N to hide
Arrow keys, or swipe