By Matthew McRoberts, VP of Business Development – AI at Sinch
AI-native companies move faster than any software category before them. They ship in days what used to take quarters, and most put a product in front of real users before the market has a name for what it does.
Take Lovable. The Swedish company crossed $400 million in annualized revenue in February with 146 employees, and passed $500 million by June, with users creating a million new projects a week. It’s an extreme case, but it’s not an isolated one. In a market this young, position gets decided in months, and the winners are usually whoever got there first and stayed.
Underneath all of that sits a layer users experience as the product itself. Every signup, every alert, every conversation a customer has runs through the communications infrastructure. It gets chosen when the company is a tenth of the size it will be, against requirements that look nothing like the ones coming. That choice holds for a long time. Then it starts falling short, at the exact moment everything else is going right.
The stress test for comms infrastructure
Every AI-native company has a version of a growth event circled – either one they’ve experienced or one they’re pushing toward. A funding announcement, a launch, a new partnership. Even someone with reach posting on a Friday night. And then, traffic that was meant to arrive over a whole quarter turns up in a single afternoon.
That puts the communications layer under a kind of pressure it was never sized for. Infrastructure that worked at 10,000 users in one market has to now hold at two million across dozens of countries, each with its own rules about who can send what. Even companies that can forecast their own growth find this hard. Sinch research found 42% of enterprises say their current provider isn’t meeting their reliability needs at scale, and those are organizations with a year of history behind them and a reasonably confident estimate of next quarter.
AI-native companies don’t get that, so the risks are higher. Handling the volume is one challenge, but what most teams haven’t considered is that the spike itself can be the problem. A sudden jump in sending is the pattern that degrades deliverability and lands verification emails in the spam folder. It’s what makes carriers throttle. On voice, the added load shows up as delay that makes a human-sounding agent sound like a broken one. All of these – the failed verification attempt, the voice agent that lags – can feel like product problems, when in reality, it’s the communications infrastructure letting you down.
The roadmap pays for the fix
These breakdowns leave a trace, but the trace takes work to read, and the work comes out of the roadmap. Ruling out the other possible causes – the funnel, the copy, the pricing change that shipped around the same time – can take weeks, and that’s time AI natives don’t have.
And if diagnosing takes time, the fix takes even longer.
Fixing customer communications that fail at scale is unfamiliar work for strong engineers used to problems that are documented and reproducible. Failures that only exist at volume won’t reproduce in a sandbox. Regulatory registration runs on someone else’s timeline, with paperwork and compliance knowledge nobody on the team has. And some of it simply takes time: Sender reputation recovers gradually, and throwing more engineers at it doesn’t speed that up.
That time comes out of one place. It’s weeks of the team’s best engineers diagnosing and fixing instead of building the thing that was supposed to keep the company ahead. Meanwhile, the competitor who settled this earlier has spent the same period shipping.
Customer trust takes the second hit
But there’s another cost AI natives can’t afford. Hypergrowth runs on word of mouth: People who like the product tell other people, and that’s what makes the curve steep in the first place.
Which is also why this category punishes disappointment quickly. AI apps lose annual subscribers around 30% faster than non-AI apps, and part of that is simply how easily users move between them. There’s always something newer to try.
Communications failures hit the customer relationship at every stage at once. A first-time user waiting on an authentication code decides the product doesn’t live up to what they’d heard. An established user who has stopped getting the updates they run their week around or that starts experiencing delay when interacting with a voice agent, starts wondering whether the product is actually holding up. And the one who wanted to come back, asked for a password reset and didn’t get it, just doesn’t return.
In a market moving this fast, all of these failures give your competitors an advantage. When customer trust takes a hit, it gives others the edge they need to outgrow you.
What infrastructure that holds up requires
Sinch research found reliability and uptime rank as the single most important criterion when companies choose a communications provider. But what does that look like in practice?
Three things define a communications infrastructure that can hold at this growth rate:
- Headroom, so the choice made at 10,000 users still holds at 10 million, and holds response times with it, because latency is where a voice agent stops sounding human.
- Deliverability managed continuously rather than assumed, so the first sign of trouble isn’t a dip in conversion rate snobody can explain.
- And documentation honest enough about failure modes that an engineer can diagnose something at two in the morning without opening a ticket.
That’s what communications infrastructure looks like when it’s built to hold, and it’s the foundation these companies need to keep growing at the rate they are now.
When comms infrastructure stops being a question
Infrastructure that holds won’t help you win a market on its own. Nobody chooses a product because its verification emails arrive on time.
What it buys is narrower and more useful than that: It buys time and trust. When the growth event arrives, nothing about it becomes an engineering problem. Instead of spending time diagnosing and fixing a failure, the team keeps pushing the product forward.
Which leaves the more interesting question. If your engineers aren’t fixing this, what are they building instead?
Sinch is the communications infrastructure AI-native companies trust, built to hold reliably at scale across email, messaging and voice – so growth events stay growth events rather than becoming engineering problems.
