AI summary
- 20X is a no-code AI agent platform for business operations, built by Setoo.
- Agents work from approved knowledge and connected systems, and they run on phone, web chat, WhatsApp, SMS, and email.
- It is not a website FAQ widget and not a stack of one-off automations. It is one place to run the work.
- Put a number on one workflow with a free tool, then launch that agent or book a strategy call.
What is 20X?
20X is a no-code AI agent platform for business operations. You use it to run voice and chat agents that finish the job: answering customers, qualifying leads, booking appointments, following up, and moving routine back-office steps along. The same agent can work on phone, web chat, WhatsApp, SMS, and email. It is built by Setoo.
If you remember one line, remember this one: a chatbot answers questions, and a 20X agent completes work.
What that means in practice
"AI operating system" is the short way to describe the category. It is not what you type into Google. You probably search for an AI voice agent, AI customer support, AI lead follow-up, or an AI receptionist.
Underneath the label, 20X gives you one place to set an agent's goal, give it knowledge you have approved, connect it to the systems your team already uses, decide what it may do on its own, and watch what happened afterwards. Your operations lead can own that, without waiting for an engineer.
Most of the value in AI is expected to show up in customer operations, sales, and other everyday functions rather than in exotic projects, which matches what we see in the field (McKinsey's research on generative AI value). The story of how we got here is in why we built 20X.
Chatbot, patched-together tools, or an agent
| Approach | Chatbot or FAQ widget | Stack of one-off automations | 20X agent |
|---|---|---|---|
| Main job | Answer questions | Pass data between apps | Complete a business workflow |
| Channels | Usually web chat | Whatever was wired up | Phone, chat, WhatsApp, SMS, email with shared context |
| Knowledge | Uploaded FAQs | Scattered | Knowledge you approve, with a fallback |
| When it fails | Guesses, or says it does not know | Breaks quietly overnight | Transfers, raises a ticket, or asks for approval |
| Who changes it | A vendor or an engineer | Whoever built it | Your operations team, without code |
Use that table the next time someone asks whether you need another bot. If the work ends at an answer, a chatbot may be enough. If it includes a lookup, a booking, a reminder, or a handover, you need an agent.
How an agent goes live
You describe the workflow in plain language. You decide what the agent may do, what it must never do, and when a person takes over. You test it against real examples before anyone outside the company hears it. Then you switch it on for one channel and watch the conversations.
That last part matters more than any feature list. The difference between a demo and something you are happy to put in front of a customer is the clarity of those boundaries, not the size of the model behind it.
Channels and where they are used
Voice covers real phone lines in English, Hindi, and a mix of both. Web chat sits on your site. WhatsApp carries sales and support threads. SMS handles reminders. Email picks up the slower conversations, with the same knowledge behind it.
The named solution areas follow the work rather than the technology: customer support, e-commerce and retail, appointment booking, lead qualification, internal operations, and onboarding and HR. If you sell online, the public example is how Thrifti runs resale operations on 20X.
Who it suits, and who it does not
It suits you if you are somewhere between a one-person business and a mid-market operation, and your week is shaped by call volume, chat volume, enquiries, or appointments. It also suits agencies that want a repeatable way to run operations for clients.
It is a weak fit if you only need an FAQ box on a website, if you must have a completed SOC 2 Type II certificate today, since we are ready for that audit rather than certified, or if the decision genuinely needs a doctor or a lawyer. Very low volume work rarely pays back either.
What it costs, in plain terms
The platform is usage based. Your agents run on tokens, where 1 token is ₹3 and a standard voice minute inside a package costs about one token, and one balance covers voice, chat, WhatsApp, and email. Top-ups do not expire, so idle seats do not quietly drain the budget the way a per-user tool does. Committed pricing exists once you are running high volume.
If you want help rather than only software, the packages are on the pricing and packages page: Explore at ₹4,999, Start at ₹25,000 plus ₹9,999 a month, Transform at ₹75,000 plus ₹24,999 a month, and Scale from ₹1,50,000 plus ₹49,999 a month. Consulting fees pay for people. Tokens pay for the platform running your agents. The two are billed separately so you always know which is which.
Before you argue about budget, put a number on the workflow with the voice cost estimator or the ROI calculator. A calculator result is a planning estimate, not a quote.
Two ways to start
You can start on your own: describe the workflow, test it, switch it on. That path exists so you do not wait on a proposal to put an agent on one inbox.
You can also build it with our team. That is worth it when the workflow is messy, when phone lines and more than one CRM are involved, or when you want a runbook your team can operate after we leave. Neither path removes the need for knowledge you have approved and a clear handover to a person. If you are weighing them up, book a strategy call. If the workflow is already obvious, start free.
Guardrails worth insisting on
Any agent that talks to customers will eventually be asked something it should not invent. Insist on answers grounded in approved content, a log of the questions it could not handle, a way to transfer or raise a ticket, approval before anything sensitive, and a record of who changed what.
Those controls are why we do not describe 20X as fully autonomous. Completing work also means knowing when to stop.
How to evaluate this in an afternoon
Do not start with twelve workflows. Take the automation opportunity quiz, or simply name the pain you already know about: missed calls, slow leads, or no-shows. Put a number on it with the matching free tool, open the relevant solution page to check channels and handover, then either start free on that one workflow or book a discovery call.
One workflow live is a better decision than a rollout plan nobody has tested.
What the first month usually looks like
Week one is scoping, and it is shorter than people expect. You pick one workflow, agree what the agent may do, and gather the knowledge it needs. Most of the effort here is deciding which documents are actually current.
Week two is testing against real conversations from last month, including the awkward ones. This is where the boundaries get corrected, because a real caller always asks something the script did not anticipate.
Week three is a limited launch: one channel, one call or message type, often only the hours when nobody is at the desk. You read the transcripts yourself rather than a summary.
Week four is the decision. Either the work is completing cleanly and you widen the scope, or the transcripts show a pattern worth fixing first. Both outcomes are progress. What you want to avoid is launching six workflows and having no idea which one is working.
Questions worth asking any vendor, including us
Where does the answer come from? If it cannot point at approved content, it is guessing.
What happens when it does not know? There should be a defined path: transfer, ticket, callback, or approval. "It will try its best" is not an answer.
Who can change it next week? If every change needs the vendor, your operations team does not own the workflow.
What does a quiet month cost? Usage-based pricing should mean a quiet month is cheap. If it does not, you are buying seats with a different label.
What can you show me? Ask for a working example on your own call or message type before anything is signed. Our public one is Thrifti.
What 20X does not do
It does not decide anything that should need a professional. It does not read your entire website and treat it as fact. It does not remove the need for someone to own the workflow and read the transcripts occasionally. And it does not make a shapeless process work by moving it somewhere faster.
Being clear about that is not modesty. Every failed automation project we have seen started with a promise in one of those four categories.
Where to go next
If you already know the workflow that hurts, start free and put an agent on it. If you want a number before you decide, run the matching free tool and bring the figure to the conversation. If the workflow spans phone lines, more than one system, and a team that needs a runbook, book a strategy call and we will tell you honestly whether the guided route is worth it.
One workflow live, measured for a month, is the fastest way to find out whether this belongs in your business.
Frequently asked
Questions, answered.
- 20X is a no-code AI agent platform for business operations, so your team can build and run agents that complete work on phone, web chat, WhatsApp, SMS, and email.
Written by
20X team
Editorial · Setoo
20X is built by operators, engineers, and consultants at Setoo. We write about the operational work AI agents should complete, not the slogans around them.