32× sign-up growth in one quarter. Nearly 15× monthly visits in two months. 100+ countries. One small team.
We tested 20+ markets and connected 7+ channels into one GTM system that learned faster than we could guess.
4 principles I’d use again ↓ https://x.com/i/article/2117522476297011200
32× Growth in One Quarter With a Small Team: 4 GTM Lessons I’d Use Again
The Starting Point
When our 2.0 beta launched in July, we shifted global marketing and GTM into full gear.
The entire growth system had to be built from the ground up. I led that build across markets, channels, creator operations, measurement, and the workflows connecting them.
We built Teamily around a clear vision, championed by @AidenChaoyangHe: move AI from personal agents to team agents that share context, memory, and workflows with people, and build an Agentic Internet where intelligence, people, and services are as accessible as friends on a social network.
The GTM challenge was to take that vision global.
Rather than place an early bet on a predetermined winning market or channel, we deliberately designed a system that could learn faster than intuition alone.
Within one quarter, Teamily achieved 32× sign-up growth across more than 100 countries. Monthly visits grew nearly 15× from July to September, then surpassed 1M in October. Teamily also trended on X twice within nine days, generated more than 250M social views in a single month, and reached roughly 10K company teams on the platform.
I don’t think of GTM as a list of channels.
I think of it as a learning system: every market and every channel should teach the next one what to do.
Here are the four principles behind that system.
1. Test Wide Before You Spend Deep
Most startups begin with one beachhead market.
We launched comparable experiments across 20 countries and regions, then let the data tell us where Teamily belonged.
We tested paid acquisition, creator partnerships, and UGC across each market. North America was part of the plan, but so were markets many AI companies tend to overlook, including Taiwan and Latin America.
At first, we made the obvious mistake: we ranked markets by sign-ups.
Then we ranked them again by paid conversion, and the map changed.
Sign-ups showed us where curiosity was. Paid conversion showed us where the product was solving a problem worth paying for.
A month later, we concentrated our budget on the five to ten markets showing the strongest signals.
Taiwan surprised us.
With less than $20K in total spend, it went from outside our top 50 markets to our #3 region by users.
Markets like Taiwan weren’t “easy.” They were just less crowded. Creator CPMs were lower, audiences had less AI-content fatigue, and strong product stories had more room to travel.
Why? Creators weren’t being bid up by a hundred funded AI companies, people still stopped to watch a good technology video, and audiences hadn’t already seen ten AI products making the same promise.
North America still mattered. But some of our most efficient growth came from markets other AI companies were underestimating.
Teamily became our shared operating layer. Market briefs, creator performance, paid results, and decisions stayed connected through shared groups and workflows, so the team could compare markets without rebuilding the context every time.
Test wide. Measure fast. Fund narrow.
2. Every Dollar Needs a Job
During the first month, we launched pilots across creator partnerships, UGC, paid acquisition, SEO, GEO, offline campaigns, events, campus marketing, and referrals.
With a new product, you don’t know which channels will work.
The faster you test them, the faster you stop paying for the wrong ones.
Testing widely did not mean spending evenly.
Offline received a relatively small allocation at first. No budget was permanently locked. We reviewed performance and reallocated spending every week as the signals changed.
We tracked CPM from day one, but CPM alone was never enough.
We also looked at what happened after the impression: site visits, sign-ups, activation, paid conversion, and what each channel taught us about the audience.
Every channel, creator partnership, and event needed a job we could explain in one sentence.
No reason, no spend.
The Bay Area has an event problem. Too many startups host events because hosting events feels like progress.
A logo on a Luma page is not a GTM outcome.
My standard is stricter. If the people in the room are not our ICP, the event is not tied to the product, and we cannot explain the intended outcome, we cut it, even if the attendance number looks impressive.
When we do run an event, we design backward from the outcome:
Who should be in the room?
What should happen afterward?
Which leads should we collect?
What content should come out of it?
How will online distribution extend beyond the people physically attending?
One of our events brought roughly 200 people into the room. Creator content and our own channels extended its online reach to 5M.
The 200 were the event. The 5M were why we did it.
Launch channels quickly. Make every dollar earn the next one.
3. Every Channel Should Make the Next One Smarter
In many companies, paid doesn’t talk to UGC, UGC doesn’t talk to events, and content operates somewhere else entirely.
Every channel can look fine on its own while the overall system never compounds.
We ran ours as one loop.
When a creator or UGC video performed organically, we moved the winning hook into paid acquisition.
It had already been tested with a real audience, so we didn’t need to invent a completely different “ad version” before learning from it.
The loop also moved in the other direction.
Paid performance showed us which hooks, audiences, and use cases were converting. Those signals went back into the next creator brief.
In just one month, a single UGC program led more than 150K people to ask for the product link in the comments.
That wasn’t just engagement. It was a direct signal of demand.
It told us which videos deserved amplification, which creators we should work with again, and which messages should appear in the next campaign.
We also learned not to evaluate creator content like a direct-response ad.
The journey was rarely linear. People watched a video, encountered Teamily again, searched the brand, checked what other people were saying, and only then signed up.
Search or paid search might receive the last-click credit, even when creators and UGC created the original demand.
Attribution should follow how demand was created, not only where the last click landed.
Inside Teamily, creators, scripts, video cuts, daily performance, and paid creative lived in one shared workflow. When a video crossed the performance threshold, the UGC Agent flagged it for the paid team to amplify.
Creator content discovered demand. Paid amplified it. Conversion data improved the next round of content.
The best channels don’t just perform. They make the next channel smarter.
4. AI Should Own Context and Coordination
With 20+ markets and 7+ channels running at once, ideas weren’t our bottleneck.
Context and coordination were.
As a marketer, I don’t need AI to write one more caption.
I need it to remember what we decided, keep an eye on the numbers, and keep things moving when people are offline.
That’s how we used Teamily.
Each week, we walked through the full funnel together: views, CPM, page views, sign-ups, activation, and paid conversion.
The goal wasn’t to celebrate the biggest number. It was to identify the next bottleneck.
With this many channels, collecting and summarizing data can become a full-time job. Teamily brought seven-day funnel reports back into the groups where decisions were being made, so the team did not need to reconstruct the context before every review.
Nobody built that report manually. It simply showed up in the group with the context needed to make the next decision.
Creator operations worked the same way.
Anyone who has run creator marketing knows the work behind it: screen 100 creators to find five, negotiate rates, review scripts, check product accuracy, and repeat the process across markets.
Our UGC Agent posted creator count, daily performance, total spend, total views, and current CPM directly into the group.
Nobody had to open another dashboard to find the answer.
The same applied to content.
We shipped two to three features a week. A workflow could carry a product update, together with our existing visual assets and UI, into a launch video or other campaign asset.
AI could carry the workflow. The team remained responsible for judgment, positioning, and taste.
AI can own coordination. Taste still belongs to people.
This way of working was not limited to our marketing team.
Roughly 10K company teams have connected coworkers on Teamily. Groups process an average of six Agent tasks per day, while more than 300 group-level Agentic Loops are created daily.
We didn’t build a massive growth organization. We expanded the team we already had with Agents that shared the same groups, context, and workflows.
AI becomes valuable when it owns context and coordination, not just content generation.
What I’d Share With Another Marketer
This quarter changed how I think about growth.
Distribution isn’t a list of channels. It’s a learning system.
The faster your markets, creators, paid campaigns, product data, people, and Agents can teach one another, the faster the entire system improves.
If I had to build the GTM system again, I would keep the same four principles:
1. Test wide before you spend deep.
1. Give every dollar a clear job.
1. Make every channel teach the next one.
1. Use AI to own context and coordination, not just content generation.
None of this is a secret.
What made the difference was running all four at once, with a small team, without allowing context to fall apart between markets, channels, and people.
We built Teamily on one bet: humans and Agents work better as one team.
We put that belief to work in our own GTM system.
Every result in this article came from humans and Agents learning and operating together.
Humans + Agents. One Team.
Start with one group, one teammate, and one Agent → teamily.ai
Sign-up growth, company-team adoption, and Agent activity are based on Teamily AI internal data. Monthly visit figures are based on third-party traffic estimates.
Written by Summer Chen (Zhe Chen), Marketing Director & GTM Lead at Teamily AI.