Last month I shared data from 300 companies about what’s working in B2B GTM. Now I want to explore exactly what this looks like in practice at some of today’s fastest-growing software companies.

Maja Voje and I brought a dozen operators together to share field reports highlighting one tangible GTM play from 2026 (you can watch the recording here). Today I’m unpacking my 5 favorites featuring growth leaders at Clay, Granola, incident.io, Lovable, and Vercel.

Here’s exactly what’s inside:

Love them or hate them, paid ads will (probably) be in your future if you get big enough (as I wrote earlier).

Tom Wentworth, CMO of incident.io, got sick of spending thousands per month on agency management fees for their paid campaigns. So he built a replacement himself, an autonomous ad agency, and shared exactly how it works. (It took only a few days.)

The build is an autonomous skill in Claude that runs in Claude Cowork. It’s multiplayer, so it can both work autonomously and be accessed by Tom’s marketing team via Slack or through Cowork. The skill knows what good looks like, it’s been trained on B2B best practices, it has benchmarks that it can compare things to, and it follows a set of guardrails so it doesn’t make mistakes. It connects directly to LinkedIn, Google, Meta, and X. Most connect via API; X runs through MCP.

From there, the autonomous agency can create, manage, deploy, and take down campaigns. It can change budgets, launch net-new campaigns, make creatives, and even generate landing pages for ABM campaigns.

It makes itself better. Once a week, it looks at all of its runs, figures out what it could have done better, and opens a pull request for Tom to review.

“The reason all of this works is because it has access to context,” Tom said. “Context is the most important thing in making anything AI work.” The autonomous ad agency pulls fresh context from the following sources:

Tom demo-ed this live, spinning up an ABM campaign for one of incident.io’s target accounts. This used to take an agency 2-3 days to spin up; now it can be done by anyone in marketing in 3 minutes. The example below for Chewy was created entirely by the skill, including ad copy, ad creative, publishing the ad to LinkedIn, and a dedicated landing page.

“If we detect in the morning that a target account is showing high intent, it can spin up a landing page and creative for that target account with no human interaction,” Tom pointed out.

“Not only am I saving a bunch of money not spending with ad agencies, I also now have much faster time to get campaigns out. When we do these ABM campaigns, our outbound BDR team has double the conversion rate compared to when we don’t.”

Most B2B buyers live on LinkedIn. The catch: LinkedIn ads get expensive fast.

Rob Denton, head of marketing at Granola, found a workaround through LinkedIn influencers and thought leader ads (these are the native-looking posts that get boosted and have probably taken over your feed). Granola’s average thought leadership ad on LinkedIn has 10x the engagement of their other LinkedIn ads.

“When I joined Granola, I wanted to find out where our users were coming from,” Rob recalled. “LinkedIn kept popping up. We weren’t doing anything on LinkedIn, a bunch of it was just organic activity and people writing about using Granola.”

Rob didn’t want to rely on organic word-of-mouth alone; he wanted something more controllable and scalable.

Granola’s program focuses on trusted operators rather than typical influencers. They’ve now ramped the program up to 200+ ambassadors including CMOs, founders, and engineers across a range of verticals.

How it works:

Rob’s team finds operator-influencers through two signals. First, they enrich existing Granola users to identify strong social followings and respected companies (Granola only wants people who actually use and understand the product). Second, they run a daily Grok Bot scan that surfaces LinkedIn mentions of Granola in Slack. Each mention is enriched with follower count, company, job title, and more so the team can prioritize the most important people.

While automation handles repeatable tasks, Rob finds that influencer marketing is operationally heavy and still requires a human touch. A freelancer keeps the program moving via outreach, negotiation, briefing, content review, and billing.

Granola is paying for three things with influencer marketing:

Some partners deliver low CPMs and strong engagement, so the spend pays for itself. Others come with (very) high CPMs but earn their keep through standout content or high-value audiences, which makes their posts great candidates for paid boosting.

An overlooked benefit of working with influencers: Granola is a horizontal product so recruiters use it differently than salespeople, marketers, or execs. Influencers create content that’s much more specific to each persona, and thought leader posts can be boosted to reach the exact audience that’s most likely to engage.

This whole piece is about what’s working in GTM. There’s just one small problem: without good attribution, nobody can agree on what’s working in the first place.

Davide Grieco, head of growth at Clay, has strong feelings about this. “I feel like attribution has not changed since 1997. It was a Saturday, and I said ‘What if I try to rebuild it myself?’”

His ideal process looked something like this: go into Salesforce, pull in data from Snowflake for product information, check Gong recordings, investigate ‘how did you hear about us?’ surveys, then consolidate everything and take a view.

So he built it in Clay via CLI, using Wispr Flow (voice dictation) and Claude (model). Clay has the context and tool access to build the audiences, it understands Clay’s pipeline stages, and it both gathers data and uses agents to analyze it. The initial build took about two hours. (Davide initially used Claude instead of the Clay Agent just because the agent did not exist yet.)

The workflow behind the scenes is complex, but the AI agent built it on its own:

Why this matters: a recent $200k deal was credited to a paid Google ad that drove a demo request. The agent discovered that two months earlier the champion signed up for a free trial and answered ‘Network’ in the ‘how did you hear about us?’ field. They also registered for two webinars in the weeks leading up to the demo request.

Last-touch attribution would have handed all the credit to Google. Now Clay can see the full picture of what influenced the buyer.

Related: A definitive guide to marketing attribution

Intimate events like IRL dinners were a major theme in the report. They’re a big deal at Vercel, too, where the field events program generates millions in net-new pipeline per quarter.

While these are among the highest-converting GTM channels, they’re also (very) time- and resource-intensive. Cheryl Wolf, enterprise marketing lead at Vercel, wanted to 3x Vercel’s field events program this year without adding headcount. Her solution: an internal app and AI agent called iris that automates Vercel’s event operations. Her team grew from running 1-2 events a month to 3-6 freeing up time for strategy and craft.

Three event formats anchor Vercel’s field events program:

“Anyone who plans events knows there’s a lot of operations behind it, which is why it’s not exactly easy to scale,” Cheryl said. “A lot of work goes into every single step in planning an event.”

Cheryl started by mapping the event operations process, breaking it down into six manual loops: approval, guest list, invites, prep, follow-up, and reporting. These involve working with reps, BDRs, and just about everyone in GTM, across spreadsheets, Notion docs, Slack DMs, email, calendar, and Salesforce.

Now iris brings all those workflows together, coordinating behind the scenes and giving the team one place to see real-time updates. Its Slack agent gathers approvals and nudges teammates for pre-event context and post-event notes.

It starts with an event intake, which captures the event type, location, date, format, target segments, and exec hosts. Attendee information is all centralized for each event. Emails and calendars are integrated, and iris drafts every email in one click (confirmation email, calendar invite, reminder email, etc.).

Cheryl’s favorite part of the app is seating arrangements. These used to take hours to get right. Now they’re algorithmic. (Last minute attendees can be dragged and dropped onto the tables.)

The app includes a product briefing, event run-of-show, and gameplan doc for internal attendees. On the day of the event, execs see all the attendee summaries and suggested talking points on their phones, then quickly add post-event notes.

Cheryl built iris herself with Claude Code and Vercel, without a dedicated engineering team. 80% of the build was done within a week.

Lovable, the vibe coding startup that recently hit a $13.3 billion valuation, is also scaling its events motion. Because its product is global and can be used for just about anything, Lovable invests in community GTM initiatives (hackathons, campus, ambassadors) led by Mallory Contois.

“When people are thinking about community as a strategy, they’re usually thinking about a place and it’s a really heavy resource lift for a company,” Mallory said. “But a forum is often not the answer.”

Mallory shared one specific motion Lovable is running now and plans to expand next year: large virtual community events. One of their flagship programs is called SheBuilds and it’s now on its 4th season. It’s a virtual hackathon for women who want to learn more about AI or are excited to build a project, and the first three seasons each drew around 2,500 applicants. (SheBuilds has been so successful that Lovable plans to replicate this motion across other personas.)

In the past, Lovable only accepted 200 participants. This season they accepted 1,000, many of whom are either new users of Lovable or users of competitors. They all converge in a dedicated SheBuilds Slack workspace for about 5 days (Slack offers a 30-day free Pro trial every time you spin up a new workspace).

“The main goal of these types of events is to get people talking about them,” she emphasized. “Community is a word-of-mouth motion, it’s an organic brand love motion.”

Mallory focuses on creating shareable moments for the community and making it dead simple to share.

It’s a win-win. Participants want to be known as AI-forward builders, so they’re motivated to share. Lovable gets word-of-mouth at scale and amplifies participants in return.

Tom replaced an expensive agency and can launch new ABM campaigns in minutes. Cheryl 3x-ed field events with no new headcount. Davide rebuilt attribution over the weekend.

Each took a channel that already worked and found a clever way to go bigger, make it better, and automate the manual work. And they did it without losing the secret sauce that made the channel work in the first place.