Humans vs AI Agents

This Week

The conversation around AI has shifted from tools that help us move faster, to agents that might not need us at all. This week, I’m digging into the tension between humans and AI agents: what’s changing, what it means for your role, and how to adapt.

Back in January, I shared a primer on AI agent fundamentals (you can revisit that one here). That covered the basics. This week, we’re looking ahead—where AI agents are evolving into autonomous, decision-making systems. 

Ready to dive in? Here’s what we’ve got this week:

• From Tool to Teammate: How AI is Changing Work

• Google Workspace Is Getting an AI Glow-Up

• What’s Powering the Next Wave of AI Agents

• What People Leaders Trust AI With (And What They Don’t)

• Meet the AI That Knows What Your Customers Want

Read on.

From Tool to Teammate: How AI is Changing Work

What happens when AI stops waiting for your prompt and starts acting on its own?

Until recently, most of us used AI like a calculator. You fed it a prompt, it gave you an output, and that was the end of the story. The responsibility to act, analyze, or decide still sat squarely on your shoulders.

But now? We're seeing AI agents that work across apps, remember objectives, and make decisions on your behalf. They can manage projects, optimize outcomes, and sometimes spot opportunities you’d miss. 

That’s a big leap, and it’s making a lot of people uneasy. And for good reason.

If your value as a worker is tied to task execution (drafting emails, managing spreadsheets, prepping decks) AI agents can feel like they’re stepping directly into your job description. 

But don’t panic (yet). There is an answer. 

The smartest people I know are changing how they work. They’re becoming editors instead of writers, strategists instead of executors, and financial analysts instead of bookkeepers. They’re designing processes, training AI agents, overseeing quality, and rethinking what work looks like for them.

It’s time to start asking yourself these questions:

• What am I uniquely good at that AI can’t replicate?

• How can I design a workflow where the agent handles the heavy lifting, and I focus on judgment, creativity, and direction?

• How can I learn to work with AI agents, rather than just use AI tools?

If you’re still treating AI like a better autocomplete, now’s the time to level up. Build workflows where the agent handles the routine, and you focus on judgment, creativity, and strategy. Not just what’s being done, but why, when, and what it means.

Want to know what this looks like in action? Section’s Become an AI Leader in a Day breaks it down in a way that’s clear, tactical, and immediately useful. I took pages of notes.

You can watch the replay at Section.

Google Workspace Is Getting an AI Agent Glow-Up

At Cloud Next—Google Cloud’s annual showcase for AI and productivity—Google unveiled a new vision for Workspace. The goal? Turn your everyday tools into intelligent, proactive teammates.

Here’s what’s coming:

Audio in Docs: You’ll soon be able to turn any document into an audio version or generate a podcast-style summary of key points. Great for catching up on the go—or for people who prefer listening over skimming.

Help Me Refine: A new writing assistant in Docs that gives suggestions to improve clarity, tone, and flow. 

Vids: The new Vids tool now taps into Google’s latest image generation model, Veo 2, to help create more polished video content.

Chat: Google Chat is getting a smarter, more integrated Gemini assistant to help summarize threads, answer questions, and surface key info without context switching.

Sheets: Automatically analyze your data and highlight insights—no formulas required.

Workspace Flows: Probably the most interesting of the bunch. Flows is a new feature that lets you build custom agentic workflows across Workspace apps. It's designed to automate multi-step processes using Gems to complete tasks like research, analysis, and content generation.

Google wants Workspace to feel less like a folder of tools and more like a connected, intelligent platform. As great as these updates sound, most of them won’t be available until Q3 or beyond.

So yes, the future looks great. You just might not get to use it until later this year.

Check out all of the Google Next announcements here.

What’s Powering the Next Wave of AI Agents

There are now more than 170 companies building AI agents across 26 categories—everything from customer support to vertical tools in finance, healthcare, and retail.

Most agents today still operate with training wheels. They’re more capable than basic copilots, but not yet fully autonomous. That’s partly by design. Guardrails are still necessary, especially with current limits around reasoning, reliability, and access.

That said, the pace of development is accelerating. Advances in memory, planning, tool use, and reasoning are pushing the space forward.

Here’s what that looks like under the hood:

• Reasoning gives agents their cognitive engine. It’s how they understand language, evaluate information, and make decisions.

• Memory allows them to hold onto both short-term context and long-term knowledge, so they don’t start from zero every time.

• Tool use lets them actually take action—triggering APIs, pulling data, updating dashboards, sending messages.

• Planning is how they break big tasks into smaller steps, track progress, reflect, and adjust.

As these capabilities improve, we’ll start to see more startups moving up the autonomy curve—agents that can manage complexity, adapt on the fly, and operate with less human oversight.

And with this much activity in the space, it won’t stay early-stage for long.

Get your hands on the CB Insight’s AI Agent Market Map at their website.

What People Leaders Trust AI With (And What They Don’t)

Last week, I hosted an AI debate at a people leader summit in Silicon Valley. The question on the table: Where does AI belong in the people function—and where should we draw the line?

To ground the conversation, I ran a quick poll. I asked: 

Where are you using AI? And what do you insist on keeping human?

The responses showed that most leaders are sitting in the middle—AI is being used, but it’s far from deeply embedded. But attitudes and approaches are changing. 

Here are the highlights from the session:

Where AI is gaining ground

• Resume screening

• Interview scheduling

• Drafting job descriptions and follow-up emails

Tasks that are high volume, rule-based, and time-consuming are fair game. One leader said, “If it saves me five hours a week and the risk is low, I’ll take it.”

Where humans still take the lead

• Coaching and development

• Conflict resolution

• Final hiring decisions—especially at the leadership level

And not a single leader said they’d use AI to deliver performance feedback (although a few admitted they use it to help write the initial report).

The deeper concern wasn’t capability—it was fairness. Several leaders voiced worries about bias being baked into their AI tools. Others flagged the risk of using AI to make people decisions without full transparency or context.

As one CHRO said: “If I can’t explain how a decision was made, I’m not comfortable making it, especially when it affects someone’s career.”

If you’re thinking about introducing AI into your people stack, start by asking: Will this tool help us move faster without losing trust?

Because in the end, speed is great. But fairness matters more.

Meet the AI That Knows What Your Customers Want

There’s a shift happening right now—and it’s not about who can write the catchiest subject line or automate a welcome sequence. It’s about AI that makes marketing decisions on its own.

AI Decisioning is what powers the tailored experiences you see on Netflix, TikTok, and Amazon. What used to take a floor of data scientists is now in the hands of marketers using platforms like Hightouch, Snowflake, and BigQuery.

Here’s how it works:

• AI agents are the workhorses. They analyze user behavior and choose what to show, when, and how at a 1:1 level.

• Reinforcement learning makes those agents smarter over time. Every click, open, or bounce becomes feedback that the system uses to improve.

• Large Language Models fine-tune the experience by analyzing content and matching tone, structure, and style to each individual.

Together, this trio replaces guesswork with self learning experiments.

The era of one-size-fits-all marketing is over. With AI Decisioning, you can deliver personalized experiences at scale, replacing outdated methods like A/B testing with always-on optimization. If you’re looking to drive customer engagement, increase lifetime value, or improve operational efficiency, AI Decisioning should be in your toolkit.

Want to learn more? Head over to Hightouch.