Decisions get lost.
What was agreed in Monday's meeting is forgotten by Thursday.
On-prem and cloud AI agents, built in India
AI agents that run on your servers or in the cloud, in your team's language, on the channels they already use.
They keep track of decisions, chase follow-ups, and answer from your documents.
- Your servers or cloud
- Hindi, Hinglish, English
- A person approves
Answered from your SOPsop-qa-014.pdf, page 6
Nothing sent yetWaiting for your approval
Hindi, Hinglish, EnglishThe way your team writes
TSGAgentOS
Morning brief, 8:30
Morning brief, 8:30
Good morning. Three things need you today.
PharmaceuticalsBatch record BR-4412 is missing the QA sign-off.From sop-qa-014.pdf, page 6
CorporateThe vendor contract renewal is still with legal.From msa-playbook.pdf, page 4
ManufacturingLine 2 has been down since 6:00.From ops-handbook.pdf, page 12
Draft follow-up to QAShall I send the reminder for the BR-4412 sign-off?
ApproveEdit
Illustration. Names and files are examples.
Where work gets stuck
Work stalls between the meeting and the next one.
Follow-ups depend on memory.
Someone has to remember to chase every task.
Knowledge lives in people.
When someone leaves, their know-how leaves too.
What we do
Three ways we can help.
AI Agents
One agent for the whole team, or AI employees built for specific roles. On your servers or in the cloud.
- Answers from your documents
- Drafts emails for approval
- Works with your calendar
- AI employees for each role
- WhatsApp, in pilot
AI Consulting
Two weeks with us to see where AI would help, and where it would not.
Learn moreIndustry solutions
SmartFactory360 for manufacturing plants. TSG Elevate for everyday workflows.
TSG AgentOS
An AI teammate that knows your company.
It reads your documents, drafts the follow-ups, and waits for a person before anything goes out.
Answers from your documents
Ask in plain words. Get the answer and the page it came from.
SOPs, contracts, handbooks and policies. Your team stops asking the same three people.
Who can close a minor deviation?
The shift QA officer can close a minor deviation. Anything above minor goes to the QA head within 24 hours.
sop-qa-014.pdf · page 6
sop-qa-014.pdf · page 6
Every answer shows the document and page it came from.
It asks before it acts
It drafts the email. You decide if it goes.
Emails, reminders and changes wait for approval. You can always see which documents and tools it used.
Draft email · to Rakesh, Mehta TradersHi Rakesh, following up on the contract renewal we discussed on Monday. Could you confirm the revised terms by Friday?
Nothing is sent until a person approves it.
AI employees
An agent for each role, not one agent for everything.
A QA assistant, an executive assistant, a maintenance planner. Each one knows that role's documents, tools and approval rules.
QA assistantPharmaceuticals2 drafts to review
Executive assistantCorporateCalendar ready
Maintenance plannerManufacturing1 work order
Proof
What we have already built, in numbers.
Essay screening at national scale
100%
of fully translated copies caught in testing, across six language pairs
The same approach works for SOPs, contracts or batch records at volume. Software finds the problem and shows the evidence. Your people make the call.
A national organisation
Field-operations tracker
10 states
and 1,000+ people on the field, on one live dashboard for head office
The same setup works for sales teams, service engineers and site audits. Reports come in from the phone and the totals add up on their own, so nobody spends the evening chasing calls.
A national organisation
Civic quiz for students
2 people
must approve every question and every published result
Nothing that matters changes without a second person signing off, and every payout traces back to a record. That is the same rule our agents follow in your company.
Helping Soul
Who we work with
Different industries. The same stuck work.
Every company has decisions waiting on someone, questions that land on the same few people, and documents nobody can find.
Pharmaceuticals
QA, production and regulatory teams
SOPs people keep asking about, batch records waiting on a signature, and deviations that need an owner before the next shift.
Corporate
Leadership, finance, HR and legal teams
What was agreed in the Monday meeting, contracts sitting with someone, and policy questions that land on the same three people.
Manufacturing
Plant heads, maintenance and production teams
Downtime nobody has explained yet, the handbook no one can find, and energy costs. SmartFactory360 adds live machine data.
How we work
From first call to a working agent.
12 weeks
Audit
We sit with your team, map the work, and look for places where AI would save real time or money.
24 to 6 weeks
Build
We put one working agent into use in 4 to 6 weeks.
3Ongoing
Run
We stay on to support it and improve it. Your team still approves anything it does.
How it is built
Your data stays where you decide.
- Private by default
The model runs on your own servers by default. If your data policy allows it, we use a cloud model instead. Your choice, and you can change it later. - It asks before it acts
The agent drafts. A person approves anything that would change the business. Every answer shows how it was put together. - Your language, your apps
Hindi, Hinglish and English today, on the channels the team already opens. WhatsApp comes first.
Your serversCloud
Local modelDocuments stay on your networkOn
Approval rulesSame in either setupOn
Moving to the cloud later is a setting, not a rebuild.
About us
Built in India, and running wherever our clients want it: their servers or the cloud.
Think Small Group is based in New Delhi. We build AI agents that run on a company's own servers or in the cloud, in the team's language, on the channels they already use.
Questions
What people usually ask us first.
Anything else, ask us on the call. We would rather answer it straight than have you guess.
Does our data leave our servers?
Not unless you want it to. By default, the agent and the model run on your own hardware. If you would rather use a cloud model, we set it up with a provider your data policy allows.
Will the agent do things on its own?
No. It drafts the message, the task or the change, and a person on your team approves anything that sends, changes or deletes. You can see which documents and tools it used for every answer.
What is an AI employee?
A specialised agent set up for one role, like a QA assistant or an executive assistant. It works from that role's documents and tools, and it follows the same approval rules as the person it helps.
Which languages does it work in?
Hindi, Hinglish and English today. It works on the channels your team already uses, starting with WhatsApp.
Do we have to be a manufacturer?
No. We work with pharmaceutical companies, corporate teams and manufacturers. SmartFactory360 is built for plants. AgentOS and our consulting work for any team that runs on meetings, WhatsApp and documents.
How long until something is working?
The audit takes two weeks. After that, one working agent is usually in daily use within 4 to 6 weeks.
What happens on the 20-minute call?
We ask how the work runs today, where decisions get stuck, and what you have already tried. If AI is not the right answer, we will tell you. There is no slide deck.
Can we see it working first?
Yes. We run AgentOS inside our own company, and we can show you how we use it.
Next step
Not sure where AI fits in your business?
Book 20 minutes with us. We'll ask how the work runs today. No slide deck, no pressure.