# Strategic Shift to Agentic AI Workflows

**Podcast:** The AI Daily Brief (Formerly The AI Breakdown): Artificial Intelligence News and Analysis
**Published:** 2026-07-26

## Transcript

Today on the AI Daily Brief, we're talking about your AI summer adventure.
The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.
All right, friends, quick announcements before we dive in.
First of all, thank you to today's sponsors, Robots and Pencils, Rackspace, Blitzy, and Airtable.
To get an ad-free version of the show, go to patreon.com slash AI Daily Brief, or you can subscribe on Apple Podcasts.
To learn more about sponsoring the show, send us a note at sponsors at aidailybrief.ai.
All right, it is a weekend, a summer weekend, no less, and we are going to go action-oriented today.
For this week's Big Thinks slash Long Reads episode, we're actually going to do two things, both of which are focused on the very practical.
The first is that Professor Ethan Mollick has published the latest version of a blog post, which he's done a number of times, that's all about which tools he recommends for different tasks.
He called it an opinionated guide to which AI to use to do stuff.
And so we're going to look at all his recommendations in case they might influence the way that you want to experiment and change your workflows.
Secondly, though, we're going to be introducing the latest training program from AIDB.
This one is at summeradventure.ai and is a choose your own adventure to expanding your AI skills.
We'll talk about how it works and give you a preview of some of the projects.
But let's start over with Ethan.
The first thing that he notes in this new opinionated guide.
is that a lot has changed since the last time he did this.
In fact, he writes, what it means to use AI to do stuff encompasses so much more stuff than it used to.
Until recently, using AI meant talking to a model through a chatbot in a constant back-and-forth conversation.
Now, it means using an agentic system where the AI is capable of doing the equivalent of many hours of real human work in one go, by combining the brains of an AI model with a set of tools that let it plan and act for you.
Basically, an agentic system gives an AI a computer to use.
So Reddit Top even divides his advice into these two categories.
And effectively, he says, if you are just looking for some low stakes answer, such as asking a chatbot for a recipe or even asking it to help you write a letter or something, pretty much any model that you might use, including any of the free models, is going to do well enough.
On the other hand, for anything where you really care about the answer, you're going to want to be using one of the premier models like Fable or 5-6 Soul set to a reasonably high thinking level.
And when it comes to doing real intensive work, Ethan argues that there are effectively only two choices right now, which is ChatGPT or Claude.
Now, it is worth noting that that means that Gemini is officially out of the rankings, at least at this point, which I don't think will come as a surprise to anyone, but still is pretty notable about the state of things.
Now, Ethan's next section is called giving your AI a computer.
And although he isn't using the word harness, he's effectively talking about the systems that you're going to put in place, including controls, permissions, etc., that are going to shape what the models can do.
He says, for example, I have the systems connected to my email, a non-private part of my Google Drive, and lots of other applications, but you have to decide what you are comfortable with.
And it's clear that the nudge that Ethan is giving is to actually use these sort of connectors.
He writes, once you are set up, you can do pretty powerful things.
For example, I told both systems, connect to my Gmail and help me prep for the MBA seminar I'm giving on Monday the 21st, including building some presentation and demos as inspiration.
Answer any outstanding messages on the topic.
Both systems got to work.
They connected to my email and figured out the task, including correctly figuring out that the next Monday the 21st was in September, not August.
And after that, they just started working, which is what agents do.
They did research on the web, decided on a presentation demo, thought about how I might want to respond to the colleague who emailed me, and more.
About 10 minutes later, both returned answers, having created a range of teaching materials and writing an email to the colleague.
This is impressive stuff that would have taken a couple hours of human work.
But he says, while Claude only prepared a draft, ChatGPT actually sent an email to my colleagues.
What happened?
Well, it was my fault.
I had previously given ChatGPT permission to send email on my behalf, and Claude was told to ask me first.
So he says in one of his first big lessons, when you use these systems for real work, the permissions matter a lot.
Both companies let you decide whether the AI must check with you before acting, such as sending an email, buying something, or changing a file, until you trust the system and understand its mistakes.
leave everything to ask for approval first, which is the default.
As you get more comfortable, Ethan suggests that you can expand how much of your computer the AIs have access to and get more complicated work as a result.
Indeed, Ethan writes, probably the most interesting trick of these apps is that they can just use your computer the way you would.
If you turn on the computer use option in code or codex, the AI can literally take over your mouse, browser, and computer.
Yes, this is a security concern, so you should proceed carefully.
Yet the results can be amazing.
I asked ChatGPT 5.6 Sol and Codex to download a 3D modeling program and use it to create a very particular design.
Download Blender and make an otter using a laptop on an airplane.
He then shares a video of the AI doing exactly that.
If you put this all together, he says, you will find the AI can do almost anything that a person with access to your computer can do.
Sometimes much better, i.e.
I have no idea how Blender works.
Although sometimes worse.
I'd rather make my own slides and write my own emails, thank you.
But the AI keeps getting better so the capabilities keep improving.
Now, clearly responding to some of the complaints that he still hears, in one of the examples he gives, he points out where people maybe have mistaken assumptions based on previous iterations of AI.
Ethan says, I have a new book coming out in October.
It has been through rounds of professional editing and proofreading, but I gave GPT 5.6 Sol and Codex the full PDF anyway and asked it to check it all over.
The AI worked for 30 minutes, chased down 195 references, and gave me pages of notes that would have taken a team of researchers many hours.
One sign of how far AIs have come.
is that every one of the AI's notes was accurate and there were no hallucinated page numbers, no invented text, no errors I could spot at all.
In fact, I had the opposite issue.
The AI was incredibly nitpicky.
Fortunately, I used my human judgment to reject these sort of complaints, which fits the theme that working with these systems is more like managing than it is chatting.
You can almost think of the AI agents as a team you delegate work to.
Now, lastly, Ethan does come back to Google.
He writes, Google, which led on benchmarks not that long ago, has fallen behind where it now counts.
It has no leading frontier model and has nothing close to codex and code.
That is why I don't suggest Gemini as your primary system right now, though this could change quickly.
However, he says that doesn't mean that Google has nothing to add.
He points to Gemini Notebook, formerly known as Notebook LM, also some of their more rich media models like Gemini Omni, which works with video.
Now, overall, this is fairly simplistic.
In fact, with absolutely no shade to Ethan, you might be sitting here thinking, wait, is that it?
But you have to remember that Ethan is writing to a much wider audience and indeed probably a much less mature AI using audience than the folks who are listening to a daily AI podcast.
And what's interesting to me about this piece, especially as opposed to previous editions of this same idea, is that rather than going use case by use case and point by point and pointing to which model he likes, really what Ethan is doing here is drawing a bright line between two entirely different interaction patterns.
On the one side, there is the old chat-based interaction for which pretty much anything will suffice.
And on the other is this totally new managing agents pattern.
It is very clear to me that Ethan is trying to make the agents management interaction pattern feel approachable and accessible for people who, if they see words like harness, are just going to have their eyes rolled to the back of their heads.
But it really is a big shift in working.
One that requires not only learning new skills, but reprogramming your brain.
Which is where now we get to the AI summer adventure from the AI Daily Brief and Super Intelligent.
I have had a ton of fun over here releasing these various educational type experiences throughout the year.
The first was AIDB New Year, which was a 10-week challenge that was meant to give you 10 basic skills of using AI in a way that could level you up over the course of the first period of the year.
Tons of people engaged with that, including many in teams.
And given the context of what we were just discussing, it's almost quaint looking back how many of the skills from that were really in that pre-agentic paradigm.
Now, a couple months later, as we all got familiar with agents and harnesses and OpenClaw took over our brains, next up was Claw Camp.
This was a crawl through glass experience designed to help you build agents in a way that really led to the learners understanding the guts of these systems and the type of opportunities that they represented.
However, OpenClaw was only one platform and pretty quickly people stopped thinking about can I build an agent?
to how do I use ClaudeCode or Codex or whatever harness I'm working within to build an actual agentic operating system that can help me do lots of different types of work.
That's what led to AgentOS, which is still, I would argue, a totally valuable program to go check out if you haven't yet.
That once again is a free self-directed program for building an agentic system that can help you do lots of types of work.
AgentOS was designed by Nufar Gaspar, who also now leads our training programs at Superintelligent, including our Executive Catch-Up program.
and our Executive Agent Leadership Program, which is effectively the enterprise-grade version of AgentOS.
Now, when it comes to the AI Summer Adventure, this one is purposely designed to be more dynamic and fun and, well, choose your own adventury.
Building on the adventure and travel theme, it's organized into different destinations.
Each of the 20-plus destinations has a lab or a project that allows you to learn some new skill.
Some of them are for beginners, some of them are for intermediate users, and some of them are even advanced.
You can filter them by level or by topic, with the topics being things like tools, agents and automation, building and creating or setting up knowledge and context.
Once you complete a project, you can stamp your passport.
The passport can be just for you, or you can set it to public and share all the cool things that you've done.
The more stamps you get, the more street cred you get.
One stamp and we call you a day tripper.
Three stamps and you become a certified AI explorer.
At six stamps, you're a trailblazer.
And at 10 stamps, you're a globetrotter.
Is that the height of millennial cringe?
Yes.
Do I care?
Not even a little.
One thing I keep seeing in enterprise AI, companies hedging across every cloud, every model, every framework, or paying a GSI for a pilot that never ends.
The team's actually shipping, they've picked a lane, and they move fast.
That's one of the reasons I like today's sponsor, Robots and Pencils.
They've gone all in on AWS.
They're an advanced tier and AWS pattern partner, and they ship production AI co-workers in 45 days.
That's led to them doing some of the more interesting work I've seen on AI co-workers.
And by that, I'm not talking about chatbots.
I'm talking about actual agentic systems that sit inside a business architecture and do real work.
That kind of focus matters if you're an enterprise leader trying to get something real into production or an AWS rep trying to move a customer from interested to deployed.
Request an AI briefing at robotsandpencils.com.
One conversation with robots and pencils and you'll know.
One of the more interesting shifts in enterprise AI right now is how quickly the conversation is moving towards infrastructure and operations.
As AI moves into core workflows, regulated data environments, and agentic systems, enterprises need governed infrastructure and inference that can operate reliably day-to-day, with clear operational accountability built in from the start.
As those systems scale, the operating model increasingly becomes part of the AI strategy itself.
Rackspace Technology is the operator of the full enterprise AI stack, from agents to infrastructure, across private cloud, hybrid cloud, and edge environments.
Rackspace builds and operates governed AI infrastructure, inference, and production AI systems, for organizations where sovereignty, compliance, and uptime are non-negotiable.
Therefore, deployed engineers stay embedded beyond deployment to help operationalize and run AI in live environments.
To learn more about where enterprise AI runs and outcomes scale, go to rackspace.com.
Weekends are for vibe coding.
It has never been easier to bring a passion project to life, so go ahead and fire up your favorite vibe coding tool.
But Monday is coming, and before you know it, you'll be staring down a maze of microservices, a legacy COBOL system from the 1970s, and an engineering roadmap that will exist well past your retirement party.
That's why you need Blitzy, the first autonomous software development platform designed for enterprise-scale codebases.
Deploy the beginning of every sprint and tackle your roadmap 500% faster.
Blitzy's agents ingest your entire codebase, plan the work, and deliver over 80% autonomously.
Validated, end-to-end tested, premium-quality code at the speed of compute.
Months of engineering compressed into days.
Vibecode your passion projects on the weekend.
Bring Blitzy to work on Monday.
See why Fortune 500s trust Blitzy for the code that matters at blitzy.com.
That's B-L-I-T-Z-Y dot com.
This episode of the AI Daily Brief is brought to you by HyperAgent, where you run fleets of agents your team can manage together.
New users get $1,000 in inference.
Forget local agents and chat workflows waiting on your laptop to be prompted.
Hyperagent deploys always-on agents in the cloud, doing real work across the tools your team already uses.
Marketing's agent turns competitor moves into landing pages.
Sales's agent enriches leads, drafts emails, and updates the CRM.
Ops agent chases the paperwork and tracks the budget.
Every agent has access to shared context and follows your rules about scope and approvals.
It's time you add agents that feel like teammates.
Hire yours at Hyperagent, built by the team at Airtable.
Claim your $1,000 in inference at hyperagent.com slash aidailybrief.
Part of what I'm excited about for the AI adventure is that as you saw with both Clawcamp and AgentOS, a lot of the stuff that we were doing there was very advanced.
Whereas when it came to AIDB New Year, a lot of it was really accessible.
It was meant to be for people who were still just getting their feet under them when it came to AI.
We've brought more of that back with AI adventure.
And so if you have friends or colleagues or even yourself have just areas where you feel really behind, even some of the beginner quick trips might be something that you want to check out.
One example of a quick trip project we're calling Pack Your ID.
This is, of course, a context project that helps you build a profile of yourself that you can give any AI you're interacting with so that it has better context about you and can provide you better, more contextual results.
The way that these projects or labs work is that they're each going to come with a set of instructions and some set of core tasks, including prompts that you can copy.
So, for example, This is a complex prompt that asks the AI to, quote, help me write and install a lightweight global identity for this AI product so future chats start knowing who I am.
The prompt involves figuring out where standing personal context should live based on the current system.
It instructs the AI to interview the user to draft a short brief.
Then finally, it asks the AI to produce, one, a paste-ready global ID block of roughly 150 to 300 words written as instructions to a future AI assistant, two, exact install steps for this product's current UI, And three, one test question I should ask in a fresh chat to verify it loaded, i.e.
something only my ID would answer well.
Now for each lab or project, we're also going to provide some additional layers, i.e.
project extensions.
In this case, it's add a second block, how to push back on me, when to challenge, when to ask a clarifying question, when to refuse a vague request.
And finally, we're also going to give you a stretch goal.
Now when it comes to the pack your ID, there is actually an even more advanced version of this context project called personal brain.
i.e.
a small set of files that teach any AI who you are, what you're working on, and how you like things done.
This is basically a context pack of the type that I shared at contextportfolio.ai in that portfolio builder, but organized as a personal learning experiment instead.
Now, if these quick trips are easy ways to get more from the AI you already have, excursions are a type of project where you are going to build something that actually outlasts the current session.
The personal brain that I was just talking about is an example of that.
But we also have excursions for directing a complete and finished creative piece, i.e.
not just one shot prompted, but actually a full, complete AI-generated creative project.
There is an excursion for vibe coding something.
I know that as much as we talk about these tools, there are still plenty of people that just haven't had the time or bandwidth to actually go build their first application, which I believe is genuinely one of the biggest unlocks that you can have using AI right now.
And more than just about anything else you can do, we'll shift you from thinking of AI as an assistant that can help you do your current work faster, cheaper, or better, to something that unlocks totally new capabilities.
And then for those who want to be really ambitious, we have what we call expeditions, which are meant to totally change how you work with AI.
There are two expeditions that are live right now.
The first is called Lemonade Stand, and the goal of it is to actually create an AI-staffed micro-business with a real ready-to-run demand test.
Now, this one is not just for entrepreneurs or solopreneurs, even for folks who are completely happy in their day job and sure they'll never leave.
Going through the process of designing a complete micro business is going to have incredible learning effects for how you use AI in your normal context as well.
Now, this particular expedition is organized into three sprints.
In the discovery sprint, you'll extract non-trivial ideas from your history, skills, access, and external idea labs, leading to a short list of five to eight ideas.
and a ranked top three with founder fit notes, i.e.
which of them seem most likely to be a good fit for you.
Sprint two is about creating a business plan, taking one of those top three concepts and mapping it as a real micro business with an AI org chart, including a week one task list, a one pager plan, and an AI staff map.
The third sprint is the validation sprint, where you learn to test demand before you overbuild, including a thin demo for software, an experiment design, and getting real signal from a human.
Now again, Each of these sprints is going to come with a core task or a set of tasks and a copyable prompt that you can drop into whatever AI you're using.
Now, importantly, especially when it comes to the more extensive excursions like this, the copyable prompts may be more about getting you started.
And indeed, many of the prompts actually open up an entire set of conversations and interactions between you and the AI.
For example, in the idea mining sprint, the copyable prompt actually organizes things into a set of phases that include things like a deep interview.
by the AI of you to help it figure out what business ideas might be a good fit.
Now, especially for some of these more complex ideas, the AI adventure platform is also going to come with a set of resources for where you can go deeper.
One more expedition that I want to sell you on, we're calling the loop.
Now, you have heard me talk endlessly about loops on this show, where you take a well-defined task with tests, builds, and diffs the agent can recheck, but applying it in non-technical work is in many ways more difficult or at least less intuitive than it is in applying it to software engineering.
The goal of the loop expedition is to help you walk through building an actual agentic loop in an area of work that is useful for you.
Now before this one even gets into the prompts and steps, there is a bunch of background learning that's going to get you more familiar with the concepts as well.
Some of the sections include what a loop is and isn't, costs, models, and how loops fail, and how to fire a loop in major tools like Claude Code, Cursor, and Codex.
Now, once you listen to this show, summeradventure.ai is going to be fully live.
You can sign up for free and start taking in.
So far, about half of the destinations are unlocked, and each week here through the end of the summer in the US, i.e.
basically the beginning of September, we'll be unlocking even more projects.
Some of those projects we already have planned, but some of them we're anticipating building on the fly based on what's happening in the industry.
Who knows, maybe we'll go build ourselves a model router, given that it seems like everyone else on the planet has at this point.
We'll also set up a new sub-community for people who are doing the summer adventure on the AIDB operator circle, and I'll include links to sign into that in the show notes as well.
As we wrap up here, one of the concepts that has been most important throughout 2026 especially is this idea of the capability overhang, the gap between what AI can do and what we're actually using it for.
Now, I would contend that there is almost no one on the planet who doesn't have some capability overhang that they are dealing with.
Even researchers inside the labs, even content creators whose entire job is to pay attention to what is happening in AI and try out everything new.
There is simply not time enough to do everything that we would like, and AI capabilities keep racing ahead.
The goal of the summer adventure is to find fun ways for you to close your own personal capability gap around whatever part of that gap matters most to you and actually have a good time doing it.
As always with programs like this, the summer adventure is completely free.
You can sign up at summeradventure.ai, and I'm excited to see you there.
For now, that's going to do it for today's episode.
Appreciate you listening or watching as always.
And until next time, peace.