Hello Reader

I spent yesterday at the Phocuswright AI Marketing Summit in New York City and during my session, I asked the room to raise their hands if they had an agentic AI program in the field.

About half the hands went up. Then I asked how many were seeing real success with it.

One hand stayed up.

Adoption is widespread. Results are rare. And while teams are still working out what produces outcomes, the platforms are accelerating without them.

Microsoft shipped a new version of Copilot. Google gave every user an AI that knows their inbox. Gartner released research on where agentic AI is headed, and Gamma crossed 100 million users by embedding itself inside the tools people already use.

This edition covers what happened, what it means for your marketing program and three questions I heard from readers this month.

March Updates You Need to Know

#1 Microsoft Copilot Goes Agentic

Anthropic’s Cowork is coming to Microsoft.

Claude Cowork—which is fueling the recent hype online—handles long-running, multi-step tasks across Word, Excel, PowerPoint and Outlook, and it uses something Microsoft calls Work IQ to pull context from your files, meetings, chats and calendar.

Most AI tools still treat each request as a one-and-done exchange: you ask, it answers, you copy-paste. Cowork breaks complex requests into visible steps, reasons across multiple documents and keeps working while you do other things. Tasks can run for minutes or hours, coordinating actions and producing outputs along the way.

You can think of Cowork as an AI agent you can delegate tasks to. A single prompt in Copilot Cowork can assemble a customer briefing document from your CRM data, create an Excel overview with usage trends and build a client-ready presentation.

If you’re on the Frontier program, you’ll get access starting in May.

#2 Google Gemini Finally Gets an Upgrade

Google just released Personal Intelligence for every US user on a paid plan. Gemini now connects to Gmail, Photos, Search and YouTube, which gives it access to the context it needs for more personalized answers.

It determines if your custom instructions, past chats, or information from your connected Google apps could be helpful in formulating an answer and pulls that in automatically.

I’ve noticed a big improvement across all my apps: Gmail search actually finds the content I’m looking for, Gemini answers are more on point, Slides can reference relevant data from my Gmail and documents, and Spreadsheets instantly pull in external sales data for comparisons.

#3 Gamma Goes Full Creative Suite

Gamma, the AI-native presentation tool, launched Gamma Imagine just a few days ago and crossed the 100 million user threshold. If you’ve used it’s deck generation feature you’ll love the new graphics and visual creation features. You can generate social images complete with texty overlays, infographics and logos.

The company also rebuilt its template experience and, most notably, embedded directly into ChatGPT and Claude. Ask Claude to build a quarterly review presentation, and it produces a Gamma deck inside the same interface, no tab-switching, no copy-pasting between tools.

If you’re still toggling between six browser tabs to build a campaign asset, you’ll love this update.

#4 AI Agents Are Coming for Marketing

Gartner's latest research shows 40% of enterprise applications will embed AI agents by the end of the year, up from less than 5% in 2025.

That’s an eight-fold increase in a single year.

They also found 60% of brands plan to use agentic AI for one-to-one customer interactions and 42% of organizations plan to design distinct AI agent personalities for different audiences.

Marketing teams have spent decades managing brand voice across channels, campaigns and content types. Now you need to add AI agents to that list.

If your company deploys a customer-facing agent with a tone that clashes with your brand guidelines, or rolls out three agents from three departments with three different personalities, you’re going to have a mess on your hands.

Reader Questions

I heard versions of these three questions in workshops, emails and DMs throughout March. I thought they were worth sharing.

#1. My team is complaining about inconsistent results from our AI tools on more complicated tasks. How do I fix that?

The most common source of inconsistency is a prompt that asks the AI to do too many things at once. When you give a single instruction that requires research, analysis, formatting and a recommendation, the model has to juggle all of those steps simultaneously, and the quality of each one drops.

Claude’s Cowork is a great solution for this. It breaks complex requests into numbered steps, works through them sequentially and shows you its progress.

Don’t have Cowork? You can apply the same logic to any AI tool by breaking your prompt into these stages.

Stage one: Gather and organize your source material, including any examples or templates you want the AI to follow.

Stage two: Analyze the data first to pull out any insights or patterns.

Stage three: Load your inputs and insights along with your prompt.

The other fix is context. If you’re asking an AI to write a campaign brief but you haven’t provided it with your brand guidelines, your audience profile or an example of a brief your team liked, the output will most likely too generic. The more relevant context you provide, the more consistent the results become.

A practical exercise: take the three AI workflows your team uses most often. For each one, write down the exact prompt you're using. If the prompt is a single paragraph asking for a finished product, restructure it into two or three sequential prompts with context documents attached. Run both versions on the same task and compare the outputs.

#2. Some team members are resisting AI adoption, citing concerns about output quality. How do I help them find effective ways to use it?

I hear this in almost every workshop I run. The resistance usually comes from people who tried an AI tool once, got a mediocre result and decided the technology doesn’t (and won’t ever) work.

Nicole Baer at Carta addresses this by mandating workshops, setting expectations for new hires (she asks for tangible examples of how candidates have worked with AI) and leading by example.

She also created AI champions across teams who share effective use cases, help troubleshoot and overall encourage the team to develop their AI skills.

Here’s a quick tactic you can employ to unstick your team:

Start with one high-frequency task per team member. Pick a task they do every week that takes at least an hour. Build the AI workflow with them (if you ask them to figure it out on their own they’ll likely get lost and give up). Show them the time savings in a side-by-side comparison. When the result is a specific, measurable win on a task they personally care about, the resistance tends to soften.

#3. How can my marketing team start implementing AI without it feeling like we’re boiling the ocean?

The instinct is to buy enterprise licenses and add AI to everything all at once. That approach stalls more often than it works, because it creates a gap between the promise ("AI will change everything") and the day-to-day reality ("I still can't figure out how to get a decent first draft out of this thing").

Jasper's 2026 State of AI in Marketing report shows 91% of marketers now use AI, up from 63% last year, but the share who say they can prove ROI fell from 49% to 41%. More people are using AI tools; fewer can show what the investment is returning.

That gap narrows when you start small and measure everything. I give similar advice in this scenario as for the resistant team members.

Pick one repeatable, time-consuming task per team member: weekly reporting, competitive monitoring, brief writing, first-draft content production. Build the AI workflow for that single task. Measure how long the task took before and how long it takes now. Share those numbers with the team.

Have a question about AI and marketing? Send it my way.

I read every reply and your questions shape what I write about here.

End Notes

I spend a good part of every week reading research, watching demos and testing tools. Most of what I find ends up in my notes. Some of it ends up in these editions. But every month there are articles, threads and resources that deserve more than a bookmark.

End Notes is a new section at the end of each monthly roundup where I pass along the best of what I consumed that month.

The March roundup:

Ann Handley, "A Ladybug in Winter"​

Ann Handley on why judgment, taste and knowing when to pause will matter more than speed in a year when AI makes output cheap and volume easy. Also, who doesn't love ladybugs?

Section, "Why You Can't Get Your AI Automations to Work"

Edmundo Ortega on why most AI automations stall (your systems were built for humans, and agents need infrastructure rebuilt for machines).

The NYTs: "Who's a Better Writer"

The New York Times ran a blind test asking readers to choose between AI- and human-written passages, and a slight majority preferred the AI versions. The result highlights how quickly AI writing has reached parity in clarity and usefulness, even as humans still stand out for voice and nuance.

Don McMillan, "Font Choice Matters"

Nothing to do with AI, but too funny not to include. This 60-second video shows how your font choice says more about you than you think (worth the laugh).

​

Want to Level Up Your AI Game?

If your team is ready for a hands-on AI strategy session, my custom-designed workshops are built to uncover the workflows that can save you hours every week.

Prefer to start small? My YouTube channel is packed with quick, practical “how-to” videos that show you exactly how I use AI tools for marketing, content, and automation.

Planning an event or conference? I deliver high-energy AI sessions that engage audiences and leave them with actionable strategies they’ll talk about long after the event. Book me for your event here.​