The next trillion-dollar AI opportunity won't be captured by software vendors.
That might sound strange given the numbers: AI unicorns have surged to roughly 500 globally, worth around $2.7 trillion combined. Nearly half of 2024's new unicorns were AI companies. In 2025, it's the majority.
What's easy to miss if you're only watching valuations is that the AI companies pulling away from the pack have one primary thing in common. They don't ship a login and walk away.
Increasingly, the pattern is services-as-software (SaS): companies deploying engineers and intelligence into organizations, building and running systems on top of their own platforms. Not "here's your dashboard, have fun." More like: we're inside your infrastructure, operating alongside your team, accountable to outcomes you actually care about. The hard part is wiring AI into messy, proprietary data and live workflows without breaking anything important.
Sierra has been the go-to example for a while now. They don't sell customer service software. They deploy AI agents that handle support, built on their platform, tailored to each client's systems and policies.
This isn't a minor product variation but rather a different business model entirely.
One major category has been noticeably slower to transform: go-to-market systems.
That's strange, given the stakes. GTM is where the money moves. Pipeline generation, conversion, expansion, retention. Everything that determines whether those 500 AI unicorns become durable businesses runs through these systems.
If AI is going to eat anything, this should be on the menu. And yet…
The average organization now uses over 110 SaaS applications. Most of them don't talk to each other in any meaningful way. Meanwhile, 64% of B2B marketing leaders don't trust their own analytics enough to make confident decisions.
The symptoms are familiar to anyone who's spent time inside a GTM org. Sales swears a deal came from relationships. Marketing insists it was the campaign. Finance wonders why neither story matches the revenue numbers. Everyone's got dashboards, nobody's got answers.
This doesn't look like a tools problem. It looks like an intelligence problem. If more software were the answer, GTM would already be fixed. We're over-solved on dashboards and under-solved on judgment.
Kyle Poyar's 2025 State of B2B GTM report puts numbers on what most operators already feel deep down in their bones.
The average software company now runs 10.5 GTM channels: 5 they consider "core," another 5.5 they're actively experimenting with. 2025 was defined, in Poyar's framing, by "endless exploration."
Ninety-one percent of teams are using AI tools, but 53% report seeing no meaningful impact or only limited gains. AI SDRs, in particular, have become a punchline. One respondent's summary: "We tried an AI SDR for six months and were unable to generate a single opportunity." Yeesh.
The part that should make every GTM software vendor especially nervous: when asked to name the most impactful tool in their stack, respondents cited ChatGPT more than any purpose-built GTM solution. A general-purpose AI with no native integration into their CRM, no visibility into their pipeline, no understanding of their ICP. And it still rated higher than the specialized tools they're paying for.
The stated priority for 2026: "Ruthless scaling."
The obvious follow-up question: scaling what, exactly?
Teams are trying everything, measuring everything, and still can't agree on what's actually working. The experimentation isn't failing for lack of effort, but because the underlying instrumentation can't connect activity to outcomes.
This is where the SaaS model starts to show its limits.
SaaS is excellent at certain things: standardizing workflows, making capabilities broadly accessible, keeping marginal costs predictable. For problems that look roughly the same across customers and benefit from shared infrastructure, it's a beautiful fit.
But the GTM mess lives in a different category of problem.
Signals are fragmented across dozens of tools that were never designed to work together. The link between those signals and actual revenue is fuzzy at best, hallucinatory at worst. And the context required to interpret any of it (your ICP, your motion mix, your pricing, your competitive position, your internal politics) is deeply organization-specific.
You can't sell a login to a problem that requires human judgment. Adding another dashboard doesn't help when the data flowing into it is scattered, duplicated, misaligned, or simply missing. The more tools you layer on, the more surface area you create for confusion.
So what's working instead?
The companies making progress here tend to combine platform with implementation. Rather than just shipping software, they stay to run it.
What that looks like varies, but the common thread is ownership of outcomes. Less "here's your tool, call us if you have questions" and more ongoing interpretation, adjustment, and accountability. The engagement doesn't end at onboarding. It deepens.
This sits between two familiar models. On one side: agencies and consultancies that sell hours, rebuild similar things over and over, and accumulate limited product leverage. On the other: classic SaaS that sells access but embeds little judgment.
The emerging pattern borrows from both without fully being either. There's a platform underneath: shared infrastructure, reusable intelligence, compounding data. And there's a layer of embedded expertise that stays close to the problem and the customer.
It's not entirely software. It's not entirely services. It's something in between, shaped by the recognition that problems are best resolved with something beyond self-serve.
So if GTM is as broken as the data suggests, and if services-as-software keeps spreading across other domains, what does that imply?
For the future of GTM tools: how many of today's point solutions survive when the winning model requires integration, interpretation, and presence?
For how value gets captured in AI: does the margin structure of software still apply when the hard part is implementation, not distribution?
We're minting AI unicorns at record speed. A decade from now, the interesting divide may not be "AI vs. non-AI," or even "SaaS vs. services."
It may simply be: which companies treated GTM as a problem to solve with their customers, and which ones treated it as a product to sell to them.