Google Cloud’s VP of Growth on Building an AI-Native Marketing Team: 8 Takeaways from SaaStr AI 2026

Google Cloud’s VP of Global Demand & Growth, Sarah Kennedy Ellis, joined us for a day-three session at SaaStr AI 2026. She ran marketing at Marketo (sold to Adobe for $4.75B) and led marketing for Adobe’s enterprise software division before Google, so she has lived through more than one platform shift.

Her core argument: Google wants to be “Customer Zero” for every application of AI in marketing. Not the flashy demo. The actual day-to-day org running on its own agents, feeding what breaks back to the product teams.

Most of what she said applies whether you have 5,000 marketers or 5. Here are the takeaways worth stealing.

1. The biggest blocker to adoption is workflow friction, not agent quality

The thing slowing down AI adoption inside Google is not model quality. It is the friction in the existing workflow and the behavioral change required to break it.

Her framing: the greatest friction in a workflow is the biggest inhibitor to adoption, well over agent quality on any given day. Teams that already spend a lot of time on change management and training are the ones getting real productivity out of agents. Teams waiting for a better model are stuck.

If you are a founder wondering why your team “isn’t using the AI tools,” the answer is usually not the tools.

2. Your top adopters are also your top learners

Inside the Google Cloud marketing team, the top 20% of AI adopters are the same people finding the most productivity, and they are the ones who have done the most training and learning.

That correlation matters for how you deploy AI across a team. The instinct is to buy licenses and wait for adoption. What actually drives it is deliberate skill-building, and the people who invest the time pull ahead fast.

3. Training has to fit into 5 minutes a week

Time is the single biggest constraint on adoption. Kennedy said people genuinely only had about 5 minutes a week to spend learning. So Google built training around that reality.

They call it AI Boost Bites: 5-to-7-minute videos, some as short as 2 to 3 minutes, each covering one specific thing you can do. Early ones were basic (“how do you create slides with Gemini”). They have since evolved into multi-agent orchestration across a campaign.

Two things worth copying here:

**They gamified it.**Internal competitions where you had to complete the task and prove you could do it, then earn badges. Kennedy called badges “gamification from 20 years ago,” and it still worked. Everybody loves a badge.**They made it external.**Boost Bites started internal, got adopted fast, and Google eventually published it free on YouTube. It is now past a million views. Customers were asking how Google trains its own teams, so Google just handed them the answer.

The lesson for B2B: if your training does not fit into the gap your team actually has, it does not happen. Design for 5 minutes, not 5 hours.

4. Scaled content is where the real ROI showed up, and quality went up not down

The headline use case was the Gemini in Chrome launch. Google needed thousands of creative assets. They ran scaled asset production and cut production time by about 70%, going from weeks to days.

The 70% is not the interesting part. Everyone quotes time savings. The interesting part: conversion rate lifted. Volume went way up, and quality went up with it.

That is the opposite of what usually happens. Normally when output volume spikes, quality drops. Kennedy’s team saw personalization down to the individual level at a scale that was not possible before, and it converted better. That is the bar to hold yourself to. If you are scaling content and quality is dropping, you are doing the volume without the judgment.

Her rule for where to go hard: high volume plus limited human judgment required to get a high-quality outcome. That is the zone where agents earn their keep.

5. The Google Cloud Next opening video: 3 weeks, mostly AI, impossible a year ago

Three weeks before Google Cloud Next, Kennedy killed the opening video in rehearsals. It was using AI, but not enough of it to actually showcase what the product could do. So the team rebuilt it.

The rebuild: a creative director sketched the concept on a napkin, fed it into Nano Banana for source images, added motion with VO, and stitched it together with a few custom agents built on the Gemini Enterprise agent platform. 138 Easter eggs referencing Google’s history, down to the 1998 origin garage and the original server rack built out of Legos.

The technical wall they hit is the one every event team knows: resolution. Max output was 4K. The screen at Next is the size of a 737. They had to upres from 4K to 12K using a custom Deep Mind model, the same one used for the Wizard of Oz production at the Sphere.

Two points that matter for operators:

The previous version of this project required an agency, a much bigger budget, and a lot more than three weeks. That whole path collapsed into an internal team using their own tools.

And the same project was flat-out impossible a year earlier. Kennedy noted they tried similar work the prior year and it was too early, mostly because of the upres problem. Twelve months later it shipped.

6. Hire for curiosity, and “tell me what you built”

Kennedy’s hiring filter now opens with a simple question: tell me what you built. Not what you managed, what you personally built.

You can feel the energy shift when someone talks about something they actually made themselves versus something they read about. It is a fast read on whether a person is genuinely curious, and curiosity is what she is hiring for.

The bigger idea underneath it: your resume is becoming a collection of the agents you have built. You are not just bringing yourself to a job, you are bringing a team. How that portability works technically is still unsettled, but the candidates already thinking that way are the ones who can build an agent-led team.

7. The sales and marketing line was already blurring. Agents finish the job.

Kennedy’s take, which she called maybe unpopular: the structure matters less than the accountability. Marketing has been moving closer to sales for a decade. RevOps was the start of it.

Agents accelerate the merge because agents do not ask for a job. They start from outcomes. Humans have spent years trying to align on shared outcomes across sales and marketing. Agents begin there by default. That reframes marketing’s job as leading from the front on the technology that produces more effective sellers, rather than defending a functional boundary.

8. Onboard an agent like a human, because the work is the same

The best reframe of the session: stop thinking of it as an agent. Think of it as something you are hiring into your org that needs context and training on day one, same as a human. The difference is it remembers the context faster and more consistently than any person you bring in.

You do not get out of the upfront onboarding. You still have to invest. And the skill of providing that context, training the agent, and designing the workflow is now the premium skill for marketers.

The mindset shift underneath it: marketers used to get their gratification from the output. Now the input is the most valuable thing you create, because the agent is responsible for the output. That is a real change in how you think about your role and where your value comes from.

The governance problem you will eventually have

One more that is worth flagging early. Google let a thousand flowers bloom on agents for about 18 months. They wanted the garden full. Then the sheer number of agents forced a quieter, behind-the-scenes curation process: which teams are actually scaling something versus running a one-hit wonder, and what is reusable across other teams.

The answer was not to stifle building. It was to put infrastructure behind sharing, so a good agent built by the Chrome team becomes available to the Cloud team, all built on the same platform (Gemini)