Omnichat Raises $10M to Put AI Agents in Sales Chats

Direct answer: Omnichat, a Hong Kong-based platform built around marketing and customer experience in chat apps, has raised US$10 million in a Series A round led by Openspace Capital. The money will go towards scaling its Agentic AI platform, deepening work on its Omni AI Agents, and expanding beyond its Asia Pacific footprint. For a UK small business, the interesting part is not the headline figure; it is the shift the funding points to, from chat as a place where customers ask questions to chat as a place where a business markets, sells and supports in one continuous conversation.

What happened

Omnichat announced the round in a release dated 26 October 2026 in the United States and 27 October 2026 in Hong Kong. Openspace Capital led the investment, with participation from Darwin Ventures and Acorn Pacific Ventures. The stated uses of the funds are straightforward: scale the Agentic AI platform, deepen research and development on the Omni AI Agents, and accelerate global expansion.

That combination matters because it tells you where the company thinks the next stage of its growth will come from. This is not framed as a round to add another inbox or another dashboard. It is framed as a round to build agents that do work inside conversations, and to take that approach to more markets. Omnichat already describes itself as an AI-native marketing and customer experience platform, and the funding gives it more room to turn that description into a product that more brands can actually run day to day.

Openspace Capital, the lead investor, is a multi-strategy asset manager focused on Southeast Asia and headquartered in Singapore. Its past investments include GoJek (GoTo), Kredivo, Fano, Video Rebirth and TradeGecko. That background is worth noting for a simple reason: investors who have backed consumer and commerce businesses in Southeast Asia tend to look closely at whether a product fits how people in a region actually buy. In much of Asia, that buying journey already runs through chat apps in a way that is still less familiar in the UK, where email, search and the website checkout remain dominant. Omnichat’s pitch is, in effect, that the chat-first pattern is spreading, and that businesses elsewhere will need tools built for it rather than adapted to it.

The timing also fits a wider pattern in 2026. Plenty of software now claims to use AI agents, and plenty of it amounts to a chatbot with a new label. A funding round does not settle whether a product is genuinely agentic, but the details Omnichat has published about its suite give a clearer picture than the label alone. The company says its platform has helped 5,000 clients generate over US$100 million in revenue, and it names clients including Dyson, FILA, Watsons, L’OCCITANE, VF Corporation, SMCP and Mead Johnson. It is also described as a Meta Business Partner, a WhatsApp Business Solution Provider and a LINE Biz-Solutions Tech Partner. Those are claims from the company itself, and they should be read as such, but they do explain why the round has drawn attention: this is a business with an existing client base and existing platform relationships, raising money to push further into agents rather than starting from scratch.

The original announcements behind this story are available from PR Newswire and TechNode Global: the PR Newswire release on Omnichat’s Series A and the TechNode Global version of the same announcement. Both set out the round, the investors and the planned uses of the funds, and they are the sources for the figures in this article.

Key facts at a glance

Timeline

Founder and chief executive Alan Chan has described the round, in paraphrase, as marking a move from conversational commerce to AI-driven commerce. His argument, again in paraphrase, is that traditional messaging tools still need heavy human effort to turn a conversation into a sale, whereas Omni agents are built to sell, market and support around the clock. That is a strong claim, and it is worth testing against how the agents are described as working, rather than accepting it at face value. The next section does exactly that.

How the Omni AI Agents actually work

The useful distinction here is between a chat tool and a chat commerce platform. A chat tool gives a business a shared inbox: messages arrive, someone replies, the conversation ends. A chat commerce platform tries to keep the whole customer journey inside the conversation, from first message to purchase to follow-up support, and to link that journey to revenue. Omnichat’s claim is that its agents do the second job, not just the first.

That depends on two things. The first is a single customer profile across channels. If a customer asks a question on Instagram, buys through WhatsApp and later complains on Facebook Messenger, the platform is meant to see one customer, not three separate threads. The second is revenue attribution: the ability to say which conversation helped produce which sale, including sales completed in a physical store. Both are valuable if they work, and both are hard to get right. Attribution in particular has limits, which are covered in the limits section below. For now, the point is that the agents are only as useful as the profile underneath them. A business with messy customer data, duplicate profiles and no consistent way of identifying repeat buyers will get messier results, however good the agent’s wording sounds.

The AI Marketing Agent, step by step

The Marketing Agent is designed to build a campaign from a single prompt. In practice, the steps look like this: the business states a goal, such as promoting a new range or re-engaging lapsed customers; the agent builds a customer segment from the unified profile; it drafts the campaign message and any supporting content; and the campaign is sent to the chosen channels. The important detail is the step in between drafting and sending. Omnichat describes human review and approval as part of the process, and that step should be treated as mandatory, not optional. An agent that can generate a complete campaign from one prompt can also generate a plausible error at scale: the wrong offer, the wrong segment, a claim the business cannot support, or a tone that does not fit the brand. The control is simple: approve every campaign before it goes out, check a sample of the segment against real customers you recognise, and start narrow, with one audience and one offer, before widening the use.

The reported growth figures give some sense of uptake. Omnichat says the number of brands using the Marketing Agent tripled within two months of OmniClaw’s June launch, though it does not publish the starting number in the announcement, so tripling is a direction, not a scale. It is also worth remembering that using a tool is not the same as getting results from it. A campaign that is easy to create still needs a genuine reason for the customer to respond, and a segment built from incomplete data will still miss the people it should reach.

The AI Sales Agent, step by step

The Sales Agent is meant to recommend products and close orders inside the conversation itself. The steps, in simplified form, are: the customer asks about a product or a need; the agent draws on the customer’s history and the business’s catalogue to suggest something suitable; it answers questions about price, availability or delivery; and it completes the order without sending the customer to a separate checkout page. The value is the removal of friction. Every extra step between a customer’s question and a completed order is a chance for that customer to drop out, and chat removes several of them at once.

That only works if the product data underneath is clean. An agent that recommends from an out-of-date catalogue, quotes a price that has changed, or promises stock that is not there will create more problems than it solves, and it will do so politely and at speed. The sensible control is to keep the catalogue accurate and to let the agent recommend within clear limits: what is in stock, what the price is, and what the delivery or collection options actually are. There is also a tone question. A good sales agent does not have to push for a sale at every turn. The most useful version helps the customer reach the right decision, including the decision not to buy today, because that is what keeps the customer coming back to the chat channel in the first place.

The AI Customer Service Agent, step by step

The Customer Service Agent handles enquiries around the clock. For standard questions, such as opening hours, delivery status, return policy or how to use a product, it can answer directly from the business’s knowledge base. For anything more complex, sensitive or high-value, its job is to hand over cleanly to a human, with the full conversation attached, so the customer does not have to repeat themselves. That handover is the part that matters most. A support agent that tries to handle everything will eventually handle something badly. A support agent that knows its limits, and passes the conversation with context, saves staff time without damaging trust.

The out-of-hours benefit is real for a small business. Customers message when it suits them, not when the shop is open. An agent that can resolve the routine enquiries overnight, and queue the rest for a human in the morning, means fewer missed messages and fewer customers who give up and go elsewhere. It does not mean the business can stop checking its messages. It means the morning check becomes a review of flagged conversations rather than a backlog of unanswered ones.

OmniClaw: the analyst in the room

OmniClaw is the fourth component, and it is the one that turns conversations into numbers a business owner can act on. It tracks customer behaviour, chat trends and store-level revenue attribution in real time. In plain terms, it tries to answer questions such as: which conversations are leading to sales, which topics are coming up most often, and whether a campaign is actually moving revenue or just generating replies. For a business that runs both online and in-store sales, the store-level part is the most interesting, and also the most difficult to trust without checking.

The distinction to keep in mind is between a campaign effect and a background buying rate. If a shop normally sells a certain number of a product each week, and a chat campaign runs in the same week, some of those sales would have happened anyway. Attribution that simply credits the last conversation before a purchase will overstate the campaign’s effect. The honest way to use OmniClaw is as a direction indicator: compare periods with and without a campaign, look at the same weeks year on year where possible, and treat a single week’s attribution figure as a starting point for questions, not as proof. Joining an online chat to a store purchase also needs a reliable link, such as a loyalty identifier, a code, a booking reference or a consented link between the chat profile and the till. Without that link, the attribution is an estimate, and it should be described as one.

The adoption figures for OmniClaw are worth reading in the same spirit. Omnichat says around 300 brands took it up within two months of its June launch. That is a sign of interest and of a product that is easy to start using. It is not, on its own, evidence that those brands are getting better results. A business considering OmniClaw should ask for the same kind of evidence it would ask for from any analytics tool: a worked example, a clear explanation of how attribution is calculated, and a way to export the underlying data and check it against its own sales records.

What it means for UK small businesses

Start with an honest caveat. Omnichat’s footprint is Asia Pacific, and its named clients and platform partnerships reflect that. A UK business will not find the same LINE or WeChat audience, and it will not find the same chat-first buying culture. What it will find, and what makes this round relevant here, is that several of the channels Omnichat covers are already central in the UK: WhatsApp, Facebook Messenger, Instagram and TikTok as a discovery channel. The question is not whether UK customers will suddenly start buying everything through chat. It is whether the businesses that already receive enquiries through those channels are treating chat as a sales channel, or merely as a support inbox.

That distinction is the practical starting point. Most UK small businesses already get questions on WhatsApp or Messenger: opening hours, availability, price, delivery, a request to book. Many of those conversations end with a member of staff replying and hoping the customer follows through. A chat commerce approach tries to close the loop inside the same conversation: answer the question, recommend the right option, take the booking or the order, and follow up afterwards. The gap between those two approaches is often not technology. It is process: who replies, how quickly, with what information, and what happens next if the customer does not reply.

There is also a framing point worth making plain. The website funnel, where a customer finds you through search, reads a page and checks out, is still important. Chat does not replace it. It adds a second journey, one that starts with a message rather than a search. A customer who discovers a product on Instagram, asks a question on WhatsApp and then buys in store has travelled through both journeys. A business that only measures website conversions will miss that customer entirely, and will undervalue the chat channel as a result. That is why the attribution discussion in the OmniClaw section matters even for a business that never uses Omnichat itself.

A playbook by business type

For a retailer, the starting point is catalogue honesty. A chat agent is only useful if the products it recommends are actually available at the price it quotes, with delivery or collection options that match reality. A beauty retailer, for example, might use an agent to help a customer choose between shades or sizes, check stock, and complete the order in chat. The routine questions, such as shade availability or delivery times, can be handled automatically. The sensitive ones, such as a reaction to a product or a complaint about a faulty item, should go straight to a person. The boundary between routine and sensitive is the most important decision a retailer makes when setting up a chat agent, and it should be reviewed regularly as new types of enquiry appear.

For a hospitality business, the value is in availability and booking. A restaurant, hotel or attraction that receives booking enquiries on WhatsApp or Messenger can use an agent to check availability, confirm a booking and send a reminder, all inside the conversation. That removes the step where a customer has to leave the chat, find a booking page and start again. The judgement calls, such as a large group booking, a dietary requirement that affects the kitchen, or a complaint about a stay, should still be handled by a person. The agent’s job is to make the straightforward bookings effortless, not to replace the host.

For a services business, such as a trades firm, a salon or a consultancy, the value is in triage and booking. A customer who messages about a leak, a haircut or a consultation does not need a sales pitch. They need to know whether the business can help, when someone can visit or see them, and what it will cost to find out. An agent can collect the basic details, offer available slots and confirm the appointment. It should not quote a price for work it has not seen, and it should not promise a timescale it cannot keep. The honest boundary is the same in every case: let the agent handle the booking, let the person handle the judgement.

Staffing and out-of-hours economics, in general terms

It is tempting to read a chat agent as a way to cut staff. The honest version is more modest: an agent changes what staff spend their time on. Instead of answering the same ten questions all day, staff handle the conversations that need judgement, the complaints that need care, and the sales that need a human touch. That is a better use of skilled people, but it is not an automatic saving. Someone still needs to review the agent’s answers, check the flagged conversations, keep the catalogue and knowledge base up to date, and handle the handovers. A business that treats an agent as a set-and-forget tool will get set-and-forget results: answers that slowly drift out of date and customers who learn not to trust the channel.

The out-of-hours economics are clearer. A small business cannot staff a phone or a chat inbox around the clock, but customers message around the clock. An agent that captures the enquiry honestly overnight, gives a useful immediate answer where it can, and sets a clear expectation for when a human will follow up, is worth more than a silent inbox. The key word is honestly: an agent should not pretend to be a person, and it should not promise a human reply by a time the business cannot meet. A simple, accurate overnight message, followed by a prompt human confirmation in the morning, builds more trust than an impressive-sounding agent that overpromises.

Consent and the WhatsApp rules, in brief

Chat commerce runs on permission, and the rules are stricter than many businesses expect. On WhatsApp in particular, business-initiated marketing messages require opt-in, and messages must use approved templates and categories. That means a business cannot simply import a list of phone numbers and start messaging offers, and it cannot treat a number given for delivery updates as permission for marketing. The agent must not be allowed to decide opt-in on its own: the business needs a clear record of who opted in, for what, and when, and a simple way for a customer to opt out. These rules are not a footnote. A business that breaks them risks losing access to the channel entirely, which would undo the whole point of building a chat audience.

The measurement that matters is simple. Count three numbers each week: enquiries by channel, conversions to sales or bookings, and reply time. Compare them after a month of using a chat agent against the month before. If enquiries are up but conversions are flat, the agent is generating noise, not sales. If reply time is down but complaints are up, the agent is answering quickly and badly. The numbers will not tell you everything, but they will stop you from mistaking activity for progress.

For readers who want to see how this fits with the wider move to agents in messaging, our earlier coverage explains how AI agents now live inside text messages on WhatsApp and SMS, and our guide to AI live chat for small businesses in the UK covers the practicalities of choosing and running a chat tool.

Honest limits / what to watch

The first limit is the simplest: a funding round is not a product review. US$10 million is a meaningful Series A, but it says more about investor confidence and about the size of the opportunity investors see than about whether the product works for a particular business. The figures in the announcement, from 5,000 clients to over US$100 million in revenue attributed to the platform, are company claims. They are useful context, but they are not independently verified results, and they do not tell a UK business what its own results would be.

The second limit is geography. Omnichat’s experience, clients and channel mix are rooted in Asia Pacific. A platform that works well where customers already buy through LINE or WeChat may behave differently where WhatsApp and Messenger dominate and where buying habits are different. The core ideas transfer, but the details, from message templates to payment flows to customer expectations, may not. A UK business should test on its own channels and its own customers, not assume that results from another market will repeat.

The third limit is data. A unified customer profile is only as good as the data that feeds it. Businesses with duplicate records, inconsistent product information or no clear consent trail will get less reliable segments, less reliable recommendations and less reliable attribution. Cleaning that data is unglamorous work, and it is easy to skip when a new tool promises to do the thinking. Skipping it is the fastest way to turn a promising agent into a source of errors.

The fourth limit is the agent’s confident error. AI agents are good at producing fluent, plausible answers, including answers that are wrong. A wrong price, a wrong stock level, a wrong policy or a wrong recommendation delivered confidently in chat can do real damage, especially if it reaches many customers before anyone notices. Human approval for campaigns, clear limits on what the agent can promise, and regular checks of a sample of conversations are not bureaucracy. They are the controls that keep a useful tool from becoming a liability.

The fifth limit is comparison framing. When choosing a chat tool, it helps to compare like with like. A shared inbox, a general agent platform and a chat commerce platform are different products with different prices and different promises. A shared inbox helps a team reply faster. A general agent platform helps a business build agents for many jobs. A chat commerce platform tries to turn conversations into revenue across channels. Omnichat is making the third claim. That claim is worth testing, but it should be tested against other chat commerce tools, not against a simple inbox, and the test should use the business’s own numbers.

What to watch next is practical. Look for UK or European results, not just Asia Pacific case studies. Look for how store attribution works for businesses without a sophisticated point-of-sale system. Look for how human approval works in practice when a business is busy, because approval that is skipped under pressure is not really approval. And watch how WhatsApp and Meta messaging rules evolve, since those rules set the boundaries for everything a chat agent can do.

The practical conclusion is modest and useful. Audit the conversations you are already losing: the enquiries that go unanswered, the bookings that never complete, the customers who ask and then disappear. Then test one agent on one job for a month, with honest measurement, and keep a human in the loop throughout. If the numbers improve, widen the use. If they do not, you have learned something valuable without betting the business on a single funding announcement.

FAQs

What is Omnichat’s $10 million Series A for?

The funding, led by Openspace Capital with participation from Darwin Ventures and Acorn Pacific Ventures, is intended to scale Omnichat’s Agentic AI platform, deepen research and development on its Omni AI Agents, and accelerate global expansion from its Asia Pacific base. The announcement was dated 26 October 2026 in the United States and 27 October 2026 in Hong Kong.

What do the Omni AI Agents actually do?

The suite covers three jobs inside chat. The AI Marketing Agent builds customer segments and generates complete messaging campaigns from a single prompt. The AI Sales Agent recommends products and closes orders inside the conversation. The AI Customer Service Agent resolves enquiries around the clock. A fourth component, OmniClaw, acts as an AI business analyst tracking customer behaviour, chat trends and store-level revenue attribution in real time.

What is OmniClaw?

OmniClaw is Omnichat’s AI business analyst. It tracks customer behaviour, chat trends and store-level revenue attribution in real time, so a business can see which conversations are connected to sales, including purchases completed in physical stores. Omnichat says around 300 brands took it up within two months of its June launch.

Which chat apps does Omnichat work across?

Its agents run across WhatsApp, LINE, Facebook Messenger, Instagram, WeChat, KakaoTalk and TikTok. Marketing, sales and customer service are unified on a single customer profile, rather than kept in separate inboxes, and revenue is attributed back to the conversation that helped produce it.

Should a UK small business switch to chat-based AI agents now?

Not on the strength of one funding round alone. The sensible test is narrower: pick one repeatable job, such as answering common questions out of hours or recovering abandoned baskets, run it for a month with clear measurements for enquiries, conversions and complaints, and only widen the use if the numbers support it. Keep human approval on campaigns and human handover for sensitive or high-value conversations throughout.

CTA: Get a Free AI Consultation