Almost Timely News: 🗞️ The Next Generation of AI Agents (2025-07-27)

3 new agents hit the market this week. Do they matter?

Christopher S. Penn

Jul 27, 2025

Almost Timely News: 🗞️ The Next Generation of AI Agents (2025-07-27) :: View in Browser

The Big Plug

👉 My new book, Almost Timeless: 48 Foundation Principles of Generative AI is now available!

Content Authenticity Statement

100% of this week's newsletter was generated by me, the human. You will see bountiful AI outputs in the video. Learn why this kind of disclosure is a good idea and might be required for anyone doing business in any capacity with the EU in the near future.

Watch This Newsletter On YouTube 📺

Click here for the video 📺 version of this newsletter on YouTube »

Click here for an MP3 audio 🎧 only version »

What's On My Mind: The Next Generation of AI Agents

This week, let's dig into something that's been in the news a lot - agentic AI.

With new tools like Claude Code's subagents, OpenAI's ChatGPT Agent, and Google's Opal, there are so many agent services and concepts to choose from. Let's take a quick tour and then talk about how these agentic servicse might be useful to you.

Part 1: What is an AI Agent?

We've covered this in the past but it's always appropriate to do a bit of a review. First, an agent is just a piece of software that acts autonomously. You've been using agents for as long as you've been using computers.

That first antivirus software you installed way back in the day? That's an agent. The sophisticated negotiation on your smartphone that happens invisibly when you connect to wifi? That's four different agents.

Agents are self-driving apps. They behave autonomously and typically fulfill highly-repetitive, specialized functions that you and I, the users, don't need to do.

An AI agent is an agent that has AI capabilities somewhere in their infrastructure. Most often this means some kind of large language model, the same kind that powers tools like ChatGPT, Gemini, Claude, and many others.

It's worth pointing out that the amount of an agent's operations that are generative AI can vary wildly, from very litle involvement to most of the work.

For example, if you had an agent that read and summarized your inbox, a substantial part of that agent wouldn't be AI at all - it'd be very old school automation, connecting to APIs, reading and writing to email protocols, working with data storage, etc. The AI portion might be 25% of the total agent's workload.

Part 2: What's Special About This Week's Agents?

As I said at the top, we had a lot of new stuff drop this week. Let's do a quick roundup of what it is.

OpenAI's ChatGPT Agent: ChatGPT Agent is what we call a browser-use tool. Inside of a virtual machine, ChatGPT can fire up its own browser and do browser-based tasks autonomously, as though it had its own iPad.

This tool allows you to give ChatGPT web surfing tasks of nearly any kind, and it goes and does whatever browser-based task you might ask of it, from checking your email to buying a pizza.

Google Opal: Opal is agent design software that allows non-technical users to create their own agent flows in a browser. Using components from the Google ecosystem, like YouTube, Google Docs, etc., Opal allows non-technical users to build fairly sophisticated agents.

For example, I demonstrated at the AFT TEACH conference this week an Opal agent flow that read and transcribed a YouTube video, did some Deep Research to get supplementary knowledge, then built a teacher's lesson plan, worksheets, quizzes, audio overview, and a short video from the data. The system, once kicked off, just goes and does the thing. It took about 40 minutes for it to complete that workflow.

Claude Code Sub-agents: Claude Code released sub-agents this week, agents within their coding environment that allows you to build mini-programs within Claude Code. These sub-agents are quite skillful at executing highly repetitive tasks, like doing quality assurance or debugging on code.

What's cool about Claude Code's sub-agents is that they can be invoked as a set of commands inside Claude Code, that then go and do their thing. Sometimes, depending on the context of the task you're working on, they can run in parallel, like team members each working on their own assignments.

What I find especially interesting is that these agents are all so different. ChatGPT is grabbing the keyboard and mouse and trying to behave like a human. Opal is having the human design very complex workflows that then become self-contained apps. Claude Code is creating lots of little agents to support a master process, using agentic tools to scale up your efforts and also help you stop typing the same things over and over again.

Part 3: Okay, But So What?

These agents are very abstract when described. It's very difficult to picture the use cases for them, the way you might use them, until you start thinking about problems you might want to solve.

The key with AI agents, like all agents, is to answer this simple question: what task might you want AI to do, start to finish, without intervention on your part?

If the answer is "absolutely nothing", then you don't need to dig too deep on agents. However, I'd wager you a small pastry (retail value 5 Canadian dollars or less) that there's at least ONE task, especially with generative AI, that you'd love to hand off to AI to do by itself.

Here's how you identify that task: what prompt do you keep copy pasting over and over again?

If you're copy pasting the prompt, there's a very good chance what you're doing would be a great agent. In general, copy pasting repeatedly is a sign that automation should be happening instead.

Here's what I'd use each agent for.

ChatGPT Agent: If there's a specific task where the agent would be good at browsing a website and gathering information over and over again, that's a great use for this tool. I could definitely see using it with something like timesheet software - if you have to fill out a timesheet each week at work, you could just dictate your hours and then have ChatGPT navigate the software and type in your time.

Google Opal: Within the Google ecosystem, I'd use this a ton for automated content transformation and creation. Because it can connect to YouTube, to anything in your Google Drive, and connect to output generators like audio and video as well as Google Docs, this is a maker's power engine.

Claude Code Subagents: As a very technical nerd, Subagents are the most useful to me. The ability to take defined, documented processes and use an AI system to apply them is a huge deal to me, for things like automated code QA as well as modifying written content.

Part 4: Demonstration

In the video version of the newsletter, watch me set up a Claude Code Subagent system from scratch, including making the agents and running through an example. The rest of this section, in fact, was produced by the Almost Timely Agent I built.

Now, before you run screaming because I said "code agents," stick with me here. We're not actually going to write any code. Zero Python, zero HTML, zero anything that looks remotely like programming. Instead, we're going to use a coding tool for non-coding purposes, which might sound counterintuitive but makes perfect sense once you understand what's happening under the hood.

The secret? These AI coding agents are just language models with better infrastructure wrapped around them. They're still LLMs at their core, but they're designed to handle multiple tasks sequentially and autonomously, coming back to you only when they need specific approvals or guidance. That's incredibly valuable for any multi-step workflow, whether it involves code or not.

Step 1: Infrastructure First (Or: How to Not Mess This Up)

As with any project worth doing, we start with infrastructure. This isn't glamorous work, but it's the foundation that makes everything else possible. Think of it like mise en place in cooking – you want your ingredients organized and accessible before you start the actual work.

I'm working with a blank Visual Studio Code project today, but you can use any folder structure that makes sense to you. The key is establishing consistent patterns that both you and the AI agents can follow.

Here's the folder structure I recommend:

docs/ – Your instruction manuals and style guides
inputs/ – Raw materials that need processing
outputs/ – Finished products
logs/ – Progress tracking and debugging information

This might seem overly structured for what we're doing, but trust me on this. These conventions exist for good reasons, and they'll save you significant headaches as your projects scale.

Step 2: Building Your Agent Arsenal

The real power of this approach lies in creating specialized agents that handle specific tasks exceptionally well. Think of these as your digital assistants, each with their own job description and expertise.

Agent #1: The Transcript Cleaner

The first agent I'm building focuses on transcript cleaning. I already have a robust set of system instructions for this task – essentially a detailed manual that tells an AI tool exactly how to transform messy, raw transcripts into clean, readable text.

Rather than embedding these instructions directly into the agent, I'm storing them as a separate document (transcript.md in my docs folder). This approach offers flexibility – I can update the instructions without rebuilding the agent, and I can reference the same instructions across multiple projects.

The agent's role is straightforward: "You're a world-class transcriptionist skilled in editing and cleaning transcripts. Use the system instructions in @docs/transcript.md and follow directions precisely to clean the transcript specified by the user."

Agent #2: The Style Specialist

The second agent handles writing style mimicry. This one's particularly interesting because it demonstrates how you can codify something as subjective as writing voice into a repeatable process.

I've created two supporting documents for this agent:

csp-writing-style.md – A comprehensive analysis of how I write, including tone, structure, and stylistic preferences
writing-samples.md – Actual examples of my writing that the agent can reference for style matching

The agent instructions read: "You're a world-class ghostwriter capable of mimicking any writing style with precision and accuracy. When directed by the user, reference @docs/csp-writing-style.md to understand how to apply the writing style of Christopher Penn in any context. Use @docs/writing-samples.md if you need to see verbatim examples of how Christopher Penn writes."

Step 3: The Orchestration Process

Here's where the magic happens. Rather than manually running each agent separately, we can create a workflow that chains them together automatically.

The prompt I use looks like this:

"Read @inputs/[filename]. Then apply the transcript cleaner agent to clean transcript. Produce the cleaned transcript in @outputs as a markdown file, and then use the Christopher Penn ghostwriter to check the writing style and make changes in the transcript output file produced by transcript to exactly mimic writing style of Christopher Penn. Christopher Penn ghostwriter should write its results to @output/final-csp-transcript.md."

This creates a sequential workflow: raw transcript → cleaned transcript → style-matched final version.

Step 4: What Actually Happens (And Why It Matters)

When you kick off this process, you're essentially watching a sophisticated delegation system in action. The AI reads your instructions, understands the workflow, and executes each step while maintaining context about what it's supposed to accomplish.

In my test run, the system took about 15 minutes to process a substantial transcript. It not only cleaned up the text but intelligently added speaker attribution where context made it clear who was speaking – something I hadn't explicitly instructed it to do, but which demonstrated sophisticated contextual understanding.

The style matching agent then took that cleaned transcript and rewrote it to match my particular voice and structural preferences. The key insight here is that this isn't just find-and-replace editing; it's genuine style transformation based on pattern recognition from the training materials.

Step 5: Quality Assurance and Fine-Tuning

No automated system works perfectly on the first try, and this is no exception. In my initial run, the style agent pulled some content directly from the writing samples rather than using them purely as style references. This taught me an important lesson about prompt precision – I need to be more explicit about using samples for style guidance only, not as source material.

This is part of the iterative improvement process. Each run teaches you something about how to refine your instructions, your supporting documents, or your workflow structure. The goal isn't perfection on day one; it's building a system that gets consistently better with each iteration.

Step 6: The Broader Implications

What we've built here extends far beyond transcript processing. This same methodology applies to any multi-step content creation process where you have:

Specific quality standards
Repeatable processes
Clear input/output requirements
Style or format specifications

Think about other applications: technical documentation from meeting notes, social media content from long-form articles, executive summaries from detailed reports, or even content strategy development from market research data.

The key insight is that we're not asking AI to be creative or strategic (though it can do both). We're asking it to execute well-defined processes consistently and at scale.

Step 7: Why This Approach Works

This methodology succeeds because it leverages what AI does exceptionally well while avoiding what it struggles with. AI excels at:

Following detailed instructions consistently
Maintaining context across long documents
Pattern matching and style replication
Sequential task execution
Handling repetitive processes without fatigue

What makes this particularly powerful is that we're essentially creating custom AI assistants with specialized skills, documented processes, and clear performance standards. Once you get this dialed in, you can process content at scale while maintaining quality standards that would be difficult to achieve manually.

Step 8: Getting Started

If you want to experiment with this approach, start simple. Pick one content transformation task you do regularly – maybe cleaning up meeting notes or standardizing report formats. Build the supporting documentation first (this is crucial), then create a single agent to handle that specific task.

Test it thoroughly with sample content before running it on anything important. Refine your instructions based on what works and what doesn't. Only after you've got one agent working reliably should you start thinking about chaining multiple agents together.

The real power emerges from the second and subsequent runs. Once you've worked out the details, updating your content pipeline becomes as simple as dropping new source material into the inputs folder, letting the system run, and coming back to polished, on-brand content that's ready for your review and publication.

Not bad, huh? That was good enough that with just a couple of minor edits, it's verbatim what Claude cleaned from the video itself, plus some nice polish based on previous editions of my newsletter.

Part 5: Wrapping Up

To be clear, because we've covered agents before, this is not new. What's new and different about this week's agents is that they're getting more capable and easier to use.

Agents are a prime focus because in many cases, they solve issues that generative AI has had for a while - the ability to connect to the systems that already have our data (ChatGPT Agent, Opal) or the ability to do sequential processes by themselves with minimal supervision (Claude Code Subagents).

These are still the earliest of early days for agents. They're still nascent, still working out the bugs, but they are a strong preview of the way things are headed with AI.

How Was This Issue?

Rate this week's newsletter issue with a single click/tap. Your feedback over time helps me figure out what content to create for you.

Here's The Unsubscribe

It took me a while to find a convenient way to link it up, but here's how to get to the unsubscribe.

If you don't see anything, here's the text link to copy and paste:

https://almosttimely.substack.com/action/disable_email

Share With a Friend or Colleague

If you enjoy this newsletter and want to share it with a friend/colleague, please do. Send this URL to your friend/colleague:

https://www.christopherspenn.com/newsletter

For enrolled subscribers on Substack, there are referral rewards if you refer 100, 200, or 300 other readers. Visit the Leaderboard here.

Advertisement: Bring Me In To Speak At Your Event

Elevate your next conference or corporate retreat with a customized keynote on the practical applications of AI. I deliver fresh insights tailored to your audience's industry and challenges, equipping your attendees with actionable resources and real-world knowledge to navigate the evolving AI landscape.

👉 If this sounds good to you, click/tap here to grab 15 minutes with the team to talk over your event's specific needs.

If you'd like to see more, here are:

ICYMI: In Case You Missed It

This week, I've been on the road, but there's still tons to catch up on.

Skill Up With Classes

These are just a few of the classes I have available over at the Trust Insights website that you can take.

Premium

Free

Advertisement: New AI Book!

In Almost Timeless, generative AI expert Christopher Penn provides the definitive playbook. Drawing on 18 months of in-the-trenches work and insights from thousands of real-world questions, Penn distills the noise into 48 foundational principles—durable mental models that give you a more permanent, strategic understanding of this transformative technology.

In this book, you will learn to:

Master the Machine: Finally understand why AI acts like a "brilliant but forgetful intern" and turn its quirks into your greatest strength.
Deploy the Playbook: Move from theory to practice with frameworks for driving real, measurable business value with AI.
Secure Your Human Advantage: Discover why your creativity, judgment, and ethics are more valuable than ever—and how to leverage them to win.

Stop feeling overwhelmed. Start leading with confidence. By the time you finish Almost Timeless, you won’t just know what to do; you will understand why you are doing it. And in an age of constant change, that understanding is the only real competitive advantage.

👉 Order your copy of Almost Timeless: 48 Foundation Principles of Generative AI today!

Get Back to Work

Folks who post jobs in the free Analytics for Marketers Slack community may have those jobs shared here, too. If you're looking for work, check out these recent open positions, and check out the Slack group for the comprehensive list.

Advertisement: Free AI Strategy Kit

Grab the Trust Insights AI-Ready Marketing Strategy Kit! It's the culmination of almost a decade of experience deploying AI (yes, classical AI pre-ChatGPT is still AI), and the lessons we've earned and learned along the way.

In the kit, you'll find:

TRIPS AI Use Case Identifier
AI Marketing Goal Alignment Worksheet
AI Readiness Self-Assessment (5P & 6Cs)
12-Month AI Marketing Roadmap Template
Basic AI ROI Projection Calculator
AI Initiative Performance Tracker

If you want to earn a black belt, the first step is mastering the basics as a white belt, and that's what this kit is. Get your house in order, master the basics of preparing for AI, and you'll be better positioned than 99% of the folks chasing buzzwords.

👉 Grab your kit for free at TrustInsights.ai/aikit today.

How to Stay in Touch

Let's make sure we're connected in the places it suits you best. Here's where you can find different content:

My blog - daily videos, blog posts, and podcast episodes
My YouTube channel - daily videos, conference talks, and all things video
My company, Trust Insights - marketing analytics help
My podcast, Marketing over Coffee - weekly episodes of what's worth noting in marketing
My second podcast, In-Ear Insights - the Trust Insights weekly podcast focused on data and analytics
On Bluesky - random personal stuff and chaos
On LinkedIn - daily videos and news
On Instagram - personal photos and travels
My free Slack discussion forum, Analytics for Marketers - open conversations about marketing and analytics

Listen to my theme song as a new single:

Advertisement: Ukraine 🇺🇦 Humanitarian Fund

The war to free Ukraine continues. If you'd like to support humanitarian efforts in Ukraine, the Ukrainian government has set up a special portal, United24, to help make contributing easy. The effort to free Ukraine from Russia's illegal invasion needs your ongoing support.

👉 Donate today to the Ukraine Humanitarian Relief Fund »

Events I'll Be At

Here are the public events where I'm speaking and attending. Say hi if you're at an event also:

AMA Pennsylvania, York, August 2025
SMPS, Denver, October 2025
Marketing AI Conference, Cleveland, October 2025
MarketingProfs B2B Forum, Boston, November 2025

There are also private events that aren't open to the public.

If you're an event organizer, let me help your event shine. Visit my speaking page for more details.

Can't be at an event? Stop by my private Slack group instead, Analytics for Marketers.

Required Disclosures

Events with links have purchased sponsorships in this newsletter and as a result, I receive direct financial compensation for promoting them.

Advertisements in this newsletter have paid to be promoted, and as a result, I receive direct financial compensation for promoting them.

My company, Trust Insights, maintains business partnerships with companies including, but not limited to, IBM, Cisco Systems, Amazon, Talkwalker, MarketingProfs, MarketMuse, Agorapulse, Hubspot, Informa, Demandbase, The Marketing AI Institute, and others. While links shared from partners are not explicit endorsements, nor do they directly financially benefit Trust Insights, a commercial relationship exists for which Trust Insights may receive indirect financial benefit, and thus I may receive indirect financial benefit from them as well.

Thank You

Thanks for subscribing and reading this far. I appreciate it. As always, thank you for your support, your attention, and your kindness.

See you next week,

Christopher S. Penn