Growth hacking was supposed to be dead. The term had a good run. Sean Ellis coined it in 2010, and for a decade it described the scrappy, slightly unholy tricks that built the consumer internet: Airbnb piggybacking on Craigslist, Dropbox handing out free storage for referrals, Hotmail signing every outgoing email with an ad for itself. Then the platforms wised up, the loopholes closed, and “growth hacker” became a job title that’s almost gone.

Now it’s back, and it looks nothing like before. Today, it looks like a research lab: a fleet of AI agents running controlled experiments on the world’s biggest distribution platforms around the clock, publishing the results, and moving on before the platform notices.

The company behind this lab is called enso. And if you spend any time in B2B marketing circles this autumn, you will hear its name.

A Research Lab for growth hacking

The first thing to understand about enso is what it isn’t. It isn’t a SaaS company with a dashboard and a sign up page. That distinction shapes everything. A vendor sells a tool and hopes you get results. A lab forms a hypothesis about how a platform decides what gets seen, tests it with a control group, measures what happened, and writes it up, including when the answer was “nothing.” enso does this for customers, on their behalf, inside the platforms where their buyers spend the day at an extreme scale, and then it publishes the findings for everyone.

Every major platform is drawn as a fortress with walls, and every fortress, the site insists, has a crack. The lab’s job is to find it, walk your brand through it, and document how. The work is organized into five research programs: search and AI answer-engine visibility, outbound sequences that run across email, messaging, SMS and voice on conditions rather than a fixed cadence, community participation in the forums where buyers ask each other for advice, newsletters, and social. The lab calls the discipline agentic growth hacking. Its founder coined the term and is writing the book.

The lab’s own history has a plot twist. It started as an AI agent marketplace and raised a $6 million seed round led by NFX. Then it pivoted, hard, into growth research, and raised $25 million. Pivots usually get buried. enso put this one on the About page, which tells you something about how it operates.

Service as a software

Sequoia Capital posted a blog about this new notion. Software, the firm argued, was about to stop selling tools and start selling the work itself. For two decades the industry sold seats: a login, a dashboard, and the promise that your people would get more done. The next act sells outcomes: the lead, the citation, the meeting, delivered by software that does the labor rather than assisting a human who does it. Service as a software. A market measured not in the $650 billion the world spends on software, but in the trillions it spends on services.

Most companies claiming that shift are still selling seats with an agent bolted on. enso is one of the few actually built on the other side of it. It doesn’t sell a growth tool. It sells growth, produced by agents, and it publishes the receipts.

That last part is the tell. When you sell software, you demo features. When you sell outcomes, you have to show results, and enso shows them with sample sizes, controls, and the experiments that failed, so the entire category can learn from it.

The category enso is walking into

To see why a lab could be worth billions of dollars, you have to look at the market it sits in, because agentic GTM in 2026 is a market eating itself, and it’s eating the agencies too.

Two years ago the pitch was simple: an “AI SDR” that writes cold emails, and a data tool to feed it. Dozens of companies raised on that. Then the arithmetic caught up. Everyone’s agent was writing to the same inboxes, reply rates fell, and the enterprise version of the product now starts at roughly $45,000 a year before you find out whether it works for your buyers. An analysis of 249 Y Combinator GTM startups founded since 2023 found only 2 percent still pitch full SDR replacement. The rest quietly rebranded as copilots.

The old alternative to the tools was an agency, and agencies are having a worse year than the tools. Worldwide ad spending grew 8.6 percent in 2025; holding-company revenues fell 1.2 percent. Read that again. The market for marketing grew by nearly nine percent, and the companies built to serve it shrank. The big holding companies have cut thousands of roles in eighteen months, one post-merger group went from roughly 128,000 staff to about 105,000 in a year, and Forrester, after an average 8 percent headcount cut across agencies in 2025, forecasts another 15 percent of agency jobs gone in 2026. Sixty percent of US marketing leaders say they are spending less on agencies specifically because of AI, and 82 percent of major brands now run an in-house shop.

The mechanics are simple and brutal. Customers use the same AI tools the agencies use. A marketing manager who used to send a brief opens a tab, gets a usable draft in a minute, edits it and ships it. The retainer doesn’t get cancelled; it gets unbundled one line item at a time, and by renewal the decision is already made. Retainers are giving way to project work, project work has thinner margins, and the junior execution layer that used to justify the fee is being replaced by software.

Forrester’s own phrase for where agencies end up is “marketing purveyors“: sellers of products, technology and media rather than hours. A few AI-native shops are reportedly running at 50 to 80 percent margins against the traditional 15 to 20, precisely because they never had that execution layer to lose.

So the two things a company used to buy, a tool or an agency, are both breaking in the same direction. The tool can’t prove it works. The agency can’t justify its hours. What’s left is whoever can show the outcome.

Meanwhile the platforms did what platforms do. The biggest data layer in the space raised at a $3.1 billion valuation, and the incumbents started shopping. In the space of four months this year, one sales-data company bought a signals startup, one CRM giant bought a website-visitor startup, and a video-conferencing giant announced it was buying a community-intelligence company. The independent signal-and-outbound layer is down to a single major player.

And the way this all gets paid for is changing under everyone’s feet. In April, the largest mid-market CRM switched its prospecting agent to $1 per qualified lead and its service agent to 50 cents per resolved conversation, with a 28-day trial. Its enterprise rival bills $2 per conversation. Outcome-based pricing is now the default direction, and once you’re paying for outcomes, the only vendors worth paying are the ones who can prove they produce them.

That is the gap enso walked into: not a tool, not an agency, but a lab that sells the outcome and publishes the proof.

That is the door enso walked through. Every trend in the category points the same way: the inbox is saturated, the tools are consolidating into CRMs, and buyers are demanding proof before payment. A lab that operates everywhere except the inbox, sells methods rather than seats, and publishes its evidence is not competing with the AI SDRs. It’s what comes after them.

Two things are dying at once

Marketing leaders in 2026 are rebuilding go-to-market from a blank page, and the reason is that the two engines that powered B2B growth for twenty years are failing in the same quarter.

The first is paid media. Ads work when the top of the funnel is uncrowded, and it has never been more crowded. Anyone can now vibe-code a product in a weekend, which means the product is no longer the moat; the marketing is, and everyone knows it, so everyone bids.

Cross-industry cost per click rose 12 percent this year, the steepest jump since 2021. Non-brand B2B software clicks are up 29 percent in a year alone and now run $8 to $14 apiece, closer to $18 in security, and $15 to $25 in categories like CRM: several times what a click cost a decade ago.

At the same time AI Overviews have cut paid click-through on informational searches by 68 percent, so the clicks that remain are fewer, later in the journey, and fought over by well-funded AI startups willing to pay $200 to $500 a lead. Paid media hasn’t stopped working. It has stopped being a growth strategy and become a tax.

The second is inbound. Brian Halligan coined the term about twenty years ago, built HubSpot on it, and for two decades the playbook was the same: publish long, useful content, rank for the question, earn the click, nurture the lead. That playbook assumed a human would read the content.

Increasingly, nobody does. Buyers ask a model, the model reads the ten best pages for them, and they get a summary with a name in it or without one. The blog post still exists. The traffic it was built to capture goes to the answer engine, and the answer engine decides which sentence, from which vendor, to quote.

Put the two together and you get the strangest moment in the history of B2B marketing. The paid channel is too expensive to scale and the organic channel no longer delivers readers. Every CMO is holding a plan written for a world where clicks were cheap and people read. The reinvention isn’t optional. It’s already underway, and most teams are doing it without a map.

The future of go-to-market

Three shifts will define the next phase of agentic GTM, and enso is positioned on the right side of all three.

Buyers are leaving traditional search. AI answer engines now sit between prospects and vendor websites, and they increasingly answer the question before the prospect clicks anything. The question for every B2B company stops being “do we rank?” and becomes “are we the sentence the model quotes?” That is inbound’s successor, and almost no GTM firm operates on that surface. enso’s citation and answer-engine research is aimed squarely at it, and it got there early.

Distribution moves off the auction. When clicks cost a tax and content goes unread, growth comes from the places money can’t buy: the forum thread a buyer trusts, the reference page a model cites, the first hour of a post’s life on a professional network. Those surfaces reward participation and timing, not budget, and they are exactly where enso’s agents work.

Growth becomes a research function. Security went through this already. Offensive security became a licensed, budgeted, board-visible discipline because the cost of not testing exceeded the cost of testing. If distribution on major platforms behaves like an attack surface, the same logic applies to growth: firms with documented methodology win the budget, and firms with a slide deck don’t. enso’s published research is the template for what that function looks like.

A unicorn shape

Let’s be clear about what enso is not. It has no self-serve product, no pricing page and no free trial. Engagements are scoped per client, the way a security firm scopes an assessment. On paper that’s a services business, and services businesses don’t get unicorn multiples.

But look at what the research is producing.

First, a category. “Agentic growth hacking” is a phrase enso owns, the founder is writing the book, and the term is already appearing in job titles. Categories are the thing venture investors will pay for when they can’t find a product.

Second, a data asset nobody else has. Every experiment the lab runs, on every platform, for every client, adds to a map of how the world’s largest distribution systems make decisions. That map gets more valuable as the platforms change, because enso monitors every move.

Third, a wedge into the surface that matters most. As the category above shows, the answer engines are where buyers are going and almost nobody in GTM is there yet.

Fourth, the pattern. The founder built a layer once, and a company that wanted that layer bought it. The layer this time sits between every B2B company and the platforms that decide whether anyone sees them. It’s hard to name a more valuable place to stand.

The lab recently hired a VP Creative, Peretz Daniel Markish, a curious appointment for a firm whose product is measurement, and one that hints the next phase pairs the research engine with craft. It also references a capability called Intelligence Mapping that it hasn’t documented yet. Ask, and the answer is that the interesting part is under NDA.

Catch me if you can

Growth hacking always had a catch, and the agentic version has two.

The first is the platforms. Everything enso finds is, by definition, something a platform didn’t intend, and every platform has the power to reclassify a clever tactic as abuse. An agent fleet is a much bigger target than an intern. enso’s answer is the publishing model: disclose, move on, treat every hack as expiring. Whether the platforms see it that way is an open question.

The second is scale. A lab whose credibility rests on one founder’s judgment and a research page has to prove that the method survives without him in the room. The book, the podcast with Forbes, and the open-source skills are all attempts to institutionalize the thinking. They’re also very good marketing, which, given the company, is presumably the point.

Growth hacking died because humans couldn’t keep up with the platforms. enso’s bet is that a research lab full of agents can. If the category keeps moving the way it has this year, that bet is starting to look less like a hack and more like the plan.