The Last Mile of Marketing Just Got Automated With ChatGPT 6 Astra

I don’t have access to ChatGPT 6 Astra yet, but the early demonstrations are enough to make one thing clear: computer use has crossed an important threshold.

Previous AI agents could reason, call APIs, and move information between systems. That was useful, but it left a stubborn gap. They couldn’t reliably open the applications marketers actually use, navigate unfamiliar interfaces, click through settings, build workflows, and coordinate work across a dozen browser tabs.

Astra appears to close much of that gap. It can work across applications, browsers, office tools, APIs, and MCP connections. In other words, it can perform far more of the work a human does on a computer.

That doesn’t mean entry-level marketing jobs vanish tomorrow. It does mean the work that used to protect those roles—coordination, setup, data movement, and execution across multiple tools—is becoming increasingly automatable.

The practical response is straightforward: don’t compete with AI on the work it can already do. Learn how to direct it, check it, improve it, and build the systems around it.

The last mile was what kept marketers in the loop

For a while, AI could do impressive things in theory but struggled with the last mile of execution.

It could write a campaign brief. It could suggest an email sequence. It could plan a workflow or explain how to configure one. But someone still had to log into the marketing platform, find the right settings, connect the steps, check the conditions, and publish everything.

That work is easy to underestimate. A campaign might involve:

• An email platform
• A CRM
• A project-management tool
• Slack or another internal communication system
• A social media scheduler
• A spreadsheet with campaign details
• An analytics dashboard
• Several people who need to review and approve the work

The context is scattered across all of them. APIs and MCP connections help, but they don’t always expose every function. Some tools have incomplete integrations. Others require a human to navigate the interface manually.

That cross-application work has been a major reason marketing teams still need junior people. It’s not always strategic, but it’s necessary. Someone has to make sure the email matches the social post, the links are correct, the audience is configured, the workflow fires at the right time, and the right people know what’s happening.

That’s the last mile. And computer-use models are starting to automate it.

Why Astra feels different

The improvement isn’t just that Astra can interact with a browser. Earlier computer-use systems could do that too, but they were slow, brittle, and prone to mistakes. You had to watch them closely. You also had to worry about what might happen if an agent misunderstood its instructions and took an unsafe action.

What stands out about Astra is the combination of speed, reliability, and compliance.

If you give an agent a goal, it needs to pursue that goal without going rogue. It shouldn’t decide that the best way to solve a marketing problem is to access systems it wasn’t authorized to touch or take actions that create security and legal risks.

That control matters as much as raw capability. An AI that can complete a task but can’t be trusted is not ready for serious business use. An AI that can work inside the browser, coordinate with desktop applications, use available APIs, and stay within its permissions is much more useful.

The result is an agent that can take on a meaningful project instead of merely answering a question.

A concrete example: building the workflow

I use HighLevel often because it combines a CRM with a broad set of marketing functions. One of its strengths is the ability to build deterministic workflows—automation that always follows a specific sequence under specific conditions.

That kind of automation isn’t something I want AI improvising every time. Once the logic is right, I want the workflow to run consistently. Classic automation is still excellent for that.

The frustrating part has been setting those workflows up. An AI system could plan the sequence, but I still had to go into HighLevel and build it manually. That meant clicking through the interface, choosing triggers, adding actions, configuring fields, and testing the result.

Now imagine telling a capable computer-use agent:

Build a lead-nurture workflow for contacts who download this guide. Send the first email immediately, wait three days, check whether they booked a call, and route the unbooked contacts into a different sequence. Test the workflow without activating it.

If the agent can navigate HighLevel reliably, inspect the available options, build the workflow, and report back on what it changed, a meaningful piece of junior marketing work has been compressed into a review task.

That’s the shift. The human may still define the objective, provide the context, approve the logic, and check the result. But the manual construction work no longer requires the same amount of labor.

This won’t happen overnight

I’ve predicted that computer use would reach this point, but I don’t think the transition will be immediate. There are several reasons.

First, the technology is expensive when you run it continuously. If every click, screen interpretation, and action consumes model resources, the bill can add up quickly. But model prices have been falling rapidly. What’s costly today will likely become much more accessible over the next few years.

Second, organizations are slow to change. Businesses have approval processes, security reviews, legacy systems, and people who are understandably cautious about giving an AI access to production tools.

Third, the technology still needs supervision. A model can complete a task and still make a subtle mistake. It might select the wrong audience, misunderstand a field, use an outdated asset, or configure a condition incorrectly. Marketing systems are full of small errors with large consequences.

So this is the beginning of the end for some entry-level marketing tasks—not the immediate end of entry-level marketers.

The curve will be gradual, but the direction is clear. As computer use gets faster, cheaper, and more dependable, companies will delegate larger projects to AI. Work that once took a junior marketer several hours may become a 20-minute review. Work that once needed a small team may be handled by one experienced marketer with a set of well-designed agents.

You can see the broader pattern in the difference between classic automation and AI agents. Deterministic systems remain valuable, but agents can now handle the messy coordination required to create and operate those systems.

What happens to entry-level marketers?

There will still be entry-level marketers. But their baseline will change.

In the past, someone could begin a marketing career by learning a collection of simple tasks: formatting emails, updating lists, assembling reports, scheduling social posts, and configuring basic campaigns. Those tasks provided a gradual path into more strategic work.

AI is already taking over much of the simple work. That creates a difficult transition. New marketers may need to jump from knowing very little to executing at a much higher level because the tools will handle the beginner tasks for them.

That sounds intimidating, but it also creates an opportunity. A capable new marketer can use AI to learn faster and accomplish more than a beginner could before. The entry-level person of the near future may be expected to produce work that would have looked closer to mid-level or even senior execution a few years earlier.

The people who thrive won’t necessarily be the ones who know the most about a single platform. They’ll be the ones who can:

• Understand the business goal behind a campaign
• Translate that goal into clear instructions and constraints
• Provide an AI system with the right context and data
• Coordinate multiple tools and agents
• Review outputs for accuracy, quality, and risk
• Diagnose failures and improve the process
• Recognize when automation should stop and a human should decide

That’s why the most useful career advice is no longer just “learn a marketing platform.” Learn how the work flows through the platform, and learn how to make AI more useful inside that workflow.

The skills marketers need to build now

1. Agent orchestration

Prompting is still useful, but prompt engineering by itself is no longer enough. The important question is becoming: how do I structure a system that can pursue a goal across several steps and tools?

That involves defining the objective, supplying context, setting boundaries, choosing the right tools, and deciding where approval is required. It’s closer to directing a small digital team than asking a chatbot for a paragraph.

For a deeper look at this shift, I’ve written about how agents and custom GPTs are changing marketing workflows.

2. Building loops instead of one-off prompts

A useful AI system rarely ends after one response. It gathers information, produces an output, checks that output against a standard, revises it, and asks for approval when needed.

Marketers need to learn how to create those loops. For example, an agent might draft a campaign, compare it against brand guidelines, verify every link, check the audience criteria, and return a list of issues before anything is published.

The goal isn’t to remove judgment. It’s to make judgment more efficient by letting the system catch predictable problems first.

3. Thinking like a software builder

Marketers don’t all need to become professional programmers. But they do need to understand how software works well enough to design useful tools and workflows.

That includes breaking a process into inputs, decisions, actions, and outputs. It means recognizing which parts should be deterministic and which parts require AI. It also means learning enough about data, permissions, testing, and failure modes to avoid building a system that quietly creates problems.

I expect more marketers to build small internal tools rather than paying for a separate subscription for every narrow use case. The tools don’t have to be complicated. A campaign QA assistant, reporting dashboard, lead-routing utility, or content-repurposing workflow can create real leverage.

4. Human judgment and taste

As execution becomes cheaper, judgment becomes more valuable.

AI can generate options. It can assemble assets. It can move data between systems. It can even produce work that looks polished. But someone still needs to decide whether the message is worth sending, whether the offer makes sense, whether the audience will trust it, and whether the campaign reflects the company’s actual point of view.

That human edge is not a sentimental extra. It’s part of the job. I’ve explored that distinction in the Human Edge Framework for marketers.

How I’m thinking about my own tool stack

One consequence of better computer use is that we may need fewer disconnected tools. Instead of subscribing to a separate application for every small function, I’m increasingly interested in consolidating around a smaller set of capable platforms and building on top of them.

For my own work, the core tools I’m focused on are Riverside, HighLevel, and ChatGPT. HighLevel is especially interesting because it covers much more than a CRM. It includes the workflows, communications, and operational pieces that make it useful as a base for a larger marketing system.

With a capable browser agent coordinating between those tools, I can imagine delegating more of the setup and maintenance without constantly jumping between applications myself. I’d still want to define the strategy and review the work, but the repetitive movement between systems could largely disappear.

If you want to test that kind of consolidated marketing stack, you can try HighLevel with a free trial. The point isn’t that one platform will replace every tool. The point is to start asking which tools are essential and which ones exist mainly because no one has had the ability to connect the work yet.

A practical test for your marketing team

Don’t wait for AI to become perfect before experimenting. Choose one difficult but bounded process and test it.

1. Pick a cross-application task. Choose something that currently requires several tools, such as launching a campaign, building a nurture sequence, or preparing a weekly report.
2. Document the desired outcome. Write down what success looks like, including the business goal, required inputs, constraints, and approval points.
3. Give the agent real context. Provide the relevant documents, brand guidelines, data, account access, and examples. Vague instructions produce vague results.
4. Keep the first test in a safe environment. Use a draft, sandbox, test list, or approval-only mode. Don’t let a new agent send messages or change production data without review.
5. Measure the result. Track time saved, errors caught, quality issues, and how much supervision the process required.
6. Improve the system. Add checks, clarify instructions, and decide which steps should remain deterministic. Then run the process again.

This is a better starting point than collecting shiny AI tools. Begin with a painful workflow and see whether an agent can reduce the friction without reducing the quality.

You can also follow a more deliberate approach to choosing which marketing workflows to automate first.

The next advantage belongs to people who keep learning

The uncomfortable truth is that the skills that made someone effective with AI a year or two ago may not be enough for the next generation of tools. Prompt writing matters less when the system can plan, browse, operate software, and execute multi-step tasks. New skills take its place.

That doesn’t mean marketers need to chase every product release. It means we need a habit of testing what the tools can actually do. Spend 30 minutes watching a serious demonstration. Give the model a difficult problem from your own company. See where it succeeds, where it fails, and what kind of supervision it needs.

The people who get ahead won’t be the ones who predict every model release correctly. They’ll be the ones who continuously update their understanding of what’s possible.

ChatGPT 6 Astra may be expensive today and imperfect in practice. But the direction matters. Once AI can reliably operate the software where marketing work happens, a large amount of coordination and execution becomes available for delegation.

Entry-level marketing won’t disappear in one dramatic event. It will be reshaped task by task. The best preparation is to move up the value chain now: learn the systems, design the workflows, supervise the agents, and bring the judgment that automation still can’t provide.

Get a free trial of High Level at https://danchez.com/highlevel The release of ChatGPT 6 Astra marks a significant shift in the marketing landscape, with the potential to automate a wide range of marketing tasks. This advancement has implications for entry-level marketing jobs and the future of marketing work. Takeaways ChatGPT 6 Astra is a game-changer for marketing automation Entry-level marketing jobs will be impacted by the automation capabilities of Astra

Dan Sanchez: ChatGPT 6 Astra just dropped. I don't even have access to it yet, but for the people who had early access, it is blowing their minds with what it's now capable of. It is far more powerful than anything on the market. Even beating out Claude's just released Fable 5.1. This is the best model on the market. It's better at just about everything. But there's one particular thing that's so good. That I think this is with this model is the beginning of the end for entry level marketing jobs. Let me explain. Welcome back to the AI Driven Marketer. I'm Dan Sanchez. My friends call me Danchez. And again, I think this is the end. Not it's just the beginning of the end. It's not going to happen overnight. It's going to take place over a few years. But Astra is a completely different model in one regard. It has one feature that's going to change marketing, and that it can do computer use faster and more reliably than ever before. And it's at a point now that I'm like, mm. This is gonna make a difference. Now it's more expensive than it's ever been before, so it's not gonna be fast, but you know how fast the price on these models are dropping. Every six months they drop down by half, it seems like. And before long, we're gonna be using this feature within Astra and models like Astra to do a lot of what was protecting a lot of marketing work. Cause if you remember, back in December of last year, in 2025, I posted a whole episode called. Future proof your job, six new marketing career paths. And I remember talking to my co-host Travis about what I saw coming in 2026, this year. So this was a prediction that I made. It wasn't like a super like way out there prediction. It was kind of obvious this this was coming because we had computer use then. We saw computer use launch in summer of 2025. It just wasn't any good. It was slow, it made mistakes. you had to worry about all kinds of cybersecurity issues. It was it it just wasn't there yet. We knew it would get better. And I predicted sometime in 2026 computer use would get to a point where it was fast, reliable, and most importantly, it's a big deal with Astra, it's much more compliant. That if you gave it a goal and a mission to go and accomplish, it won't go like rogue on you and start hacking everybody in order to get the goal done. That's one of the things they they fixed with this model. And that's good news. But it also means that a lot of marketing work, a lot of things that was protecting marketers from a AI takeover, was eliminated. Because the one thing, like I said in December, protecting especially entry-level marketers was cross-tab work. The kind of work going on in marketing a lot is the stuff where we're like having to coordinate between a few emails, the Slack message. this MailChimp campaign, coordinating it with social and getting it all cohesive and working with all the people we need to work with in order to get the campaign out, right? It I mean you're talking about work across multiple applic a dozen different applications. And it's the context is spread out all over the place. And as good as all the like harnesses were with the agents, it's like it can only do so much through APIs and MCPs. And maybe the the the knowledge doc the the knowledge it had in the folder or in your Google Drive or something. It was still missing this ability to log into an application, move the mouse around, click on things, and move things around. For example, I love the application high level. I use it all the time. In fact, I always I I promote it because it's like one of my favorite pieces of software. It's the one that I can't seem to get rid of. I would vibe code it if I could, but I can't because it does too much that I can't vibecode. so I'm like, yeah, I'll just build on top of it. but there's a wonderful. feature in it where you can build, you know, automated workflows. Cause sometimes you don't want AI to come up and automate everything. You want to have a deterministic or a set like, it always happens in this way, in this order at this time. Right. It's where classic automation is still great. But you can't automate building the workflows. AI could plan them for you, but it couldn't use the MCP or the API, go into high level and build all those workflows. You still had to do that painfully manually. And it's a pain. It's still one of the big pain in the butts that I have to do today. it's it's weird to call it a pain in the butt because you I mean we're outsourcing more and more to AI, but that was one of the things that, you know, protected marketers. I s someone still had to set that stuff up. And not anymore. Not with Astra. Astra is so good that it can go into Blender and build 3D models of cake characters you could draw a pen pencil sketch of and it'll build it in Blender. If it can do that, it can build hot workflows in high level. It's pretty dang powerful. And it's it's beyond even just Chrome tabs. So It's not just Chrome tabs, it can work in your computer, in your your your office suite. It can use the different applications on your computer and coordinate with the browser and coordinate with the information and use the API and the MCP and now it's got everything. Anything a human can do on a computer, AI can now do. So that's gonna make a big difference for a lot of entry level marketing work. And I called it in December that once this got released, and I predicted that it was gonna come this month, or not this month, but like sometime this year, here it is. but it's still just the beginning of the end, like I said, it's expensive right now. If you were running this all the time, you'd be stacking up quite a bill for AI. But it's going to drop, it's going to get easier. It's just the beginning. But as this ability continues to improve. To get cheaper. More and more of these marketing tasks that require working across the browsers and the different apps you use are going to be automated. We're going to delegate larger tasks and larger projects to AI that it would have taken a junior level marketer to do before. Shoot, there's some senior level things that market that AI can do now well now. So it's like, wow, this is a big change for marketing, starting with ChatGPT six Astra. I'm excited for it, but I'm also kind of like, dang. entry level marketing is going to change. There will still be entry level marketers. Don't get me wrong. There will be some hotshots that like come up through AI and figure it out and get learn faster and kind of get in there. What that work will look like, it's almost like they have to do like a big jump from knowing nothing to being able to execute a lot. But I I think they will. I think with AI they can actually get there faster and farther. that can get them still pretty good, meaningful entry level marketing jobs. But there'll be if you would compare the entry level marketer of two years from now to the entry level marketer from a year, a year ago, one would be considered almost more of a senior marketer. That's just the skill difference between the entry level marketer of maybe twenty twenty four and the entry level marketer of twenty twenty eight. The difference between those two people, drastically different. Cause they're gonna have to be different because AI can do all the simple work already. And I think people will rise to the occasion. And who knows what this means for the outcome of all marketing work. Maybe there's less marketing work, maybe there's more marketing work. We don't know. Not until we actually get there will anybody know. But what I do know is that the tech is going to continue to move forward. We have to learn to master it, in order to stay ahead of our own careers. Because if one thing I'm sure of that will slow the curve a little bit in a good way is The speed of change is pretty slow when it comes to organizational change. When it comes to human change, we don't like change. So that's gonna help. That'll slow things down a little bit for those who are listening to podcasts like this to get ahead, to stay up to date. And you do have to stay up to date because your prompt engineering skills from a year or two ago are no longer relevant today. And now there's new skills. There's agent orchestration. There is building loops, setting goals, thinking through how to build software. Is a big one that marketers need to really start taking up today. The cool thing is Astra and AI continue to make it easier, but then there's always a new hard thing to learn. So we have to continue to upskill with AI, not just how we use AI, but the skills we need in order to make AI more useful for us and for the companies we work for. So go out there, learn more about Chat GPT, learn more about what Astra is capable of. And if you haven't started coding with it, go and code something. I think in the future, in the near future, I'm gonna do a a whole episode on different tools you should be considering building in-house and letting go of that subscription you've been holding on to for a long time. It's getting better. And now you can fine-tune it to be whatever you want it to be. And we can start consolidating our tools. I could tell you The two tools that I'm consolidating around are as Riverside, this thing that I'm recording with right now, and High Level. high level something that I've thought about vibe coding because it's just a CRM, but it's so much more than just the CRM. So more and more, and I'm excited about Astra and its browser use because it can actually take over more of high level for me. but high level is just a fantastic tool. So I highly recommend checking that out at danchez.com slash high level to get a free demo of it. And an extension on the 14-day trial to the 30-day trial, because again, I'm building everything around these two tools now. Riverside and high level and chat GPT. And now with Astra's browser mode, I can orchestrate more and faster between just these three tools than ever before. So future episode coming up on how to use things like Astra to build and orchestrate your own custom tool set. But until then. Go and watch some of the videos around Astra that dive deep into the specs to get a get your mind around what Astra is capable of. Spend spend at thirty minutes to an hour this weekend watching a few tutorials on people who had early access to it. And when it lands into your account, ask it some hard questions. In fact, if you can, give it a lot of data for your company and have it analyze it for this problem or this opportunity or whatever it is. Give it something hard. Test it and start to learn how to incorporate it in your marketing department.