You may not need 10 agents
When I put this list together, my goal was not to recommend installing all ten AI agents at once. I see each play as one focused part of the go-to-market strategy.
The temptation is to turn every repetitive task into a separate agent, but I don't think that creates a better system. You will see that as your AI workflows mature, related tasks can often be combined into fewer, more capable agents with clearer responsibilities.
The goal is not to use as many AI agents as possible. It is to automate repetitive work while keeping in your control the strategy, messaging, and customer experience.
In fact, if your marketing is already scattered, ten disconnected agents can make it even more scattered, only faster.
An AI agent is not a strategy
I have written before about why scattered marketing never works and the same principle applies to AI.
An agent can complete a job, but it cannot decide why that job deserves to exist. It does not automatically understand your positioning, your customers, or the experience you want them to have with your brand.
The most useful way I have found to think about an agent is as a repeatable play with three parts:
- Input: the strategy, customer data, brand rules, and performance signals it needs
- Process: the focused decisions and actions it is allowed to complete
- Output: a concrete result that a person can review, measure, or act on
The output should then feed the next cycle: strategy informs execution, execution produces reporting, and reporting sharpens the strategy. An agent becomes more useful when it supports that loop instead of firing in isolation.
Before automating anything, I ask myself three questions:
- What decision or action is this workflow supposed to improve?
- Where does it sit in the customer journey?
- What should remain under human control?
If you cannot answer these questions yet, your workflow is not ready to be automated.
This is also why I do not believe in handing your entire marketing strategy to AI. AI can support research, execution, analysis, and iteration. The strategy still needs a point of view. It needs priorities and someone willing to say, “this does not help our users, so we are not doing it.”
The 10 AI GTM workflows we discussed
In the episode, we look at ten agents that support different parts of the go-to-market journey, from finding and nurturing leads to creating content and learning from its performance.
1. LinkedIn intent detection and outbound
For this first workflow, an AI agent can identify people on LinkedIn who are publicly describing a problem you solve, check whether they match the ICP, and prepare outreach anchored in that real signal.
However, for this specific example, even though the agent can help you notice opportunities, a human still needs to decide whether there is a real reason to start the conversation.
2. Lead nurturing based on real behavior
Most nurturing is built around a schedule: wait two days, send an email, wait three more days, send another one, etc.
Behavior-based nurturing reacts to what someone actually does. Did they return to a specific page, read several articles around the same problem, or engage with one format while ignoring another?
An agent can connect those signals to a defined persona and choose a more relevant next step. Someone who visits a pricing page should not receive the same follow-up as someone reading an introductory article. This brings automation closer to the principles of UX Marketing: respond to the user’s journey instead of forcing everyone through the same funnel.
3. Automated newsletters
Newsletters include plenty of repetitive work: collecting material, categorizing it, drafting summaries, formatting sections, and preparing distribution.
AI can make that process faster and segment the newsletter by persona, so a strategic decision-maker and a hands-on practitioner do not receive the same version. Additionally, performance from each newsletter send can improve the next round.
4. Social media planning and distribution
For social media, AI agents can help build a calendar, adapt a core idea to different formats, and schedule approved posts, but the process should always start by looking at the previous month.
What did people respond to? What did they ignore? When were they most engaged? Those answers should help your agent shape what to publish next.
The risk here is confusing consistency with volume. Publishing more often does not help when every post sounds generic. As I explained in How AI slop is quietly damaging your marketing, content that is fast but forgettable still costs your brand attention and trust.
Use automation to protect the rhythm of your content, not to remove its personality.
5. Paid-ad creative and optimization
Agents can produce channel-specific variations, rotate approved creative, monitor patterns, and cut underperformers. This said, I would still keep a human involved because the best-performing ad is not automatically the best expression of your brand.
Short-term clicks can become a long-term problem when the message creates the wrong expectation, which AI won't be able to notice.
6. Turning videos into blogs, clips, and other content
For me, this is one of the most practical uses of AI in a content system.
A long conversation can become an article, several platform-specific posts, an email, short clips, and quote assets. Repurposing should translate an idea for a new context, not paste the same transcript everywhere. The blog can carry the complete argument, while a LinkedIn post leads with the professional insight and an email becomes a more direct outbound.
This article is a good example of repurposing content by turning our long podcast conversation into a short blog :)
7. Programmatic SEO and AI-search visibility
Agents can classify keywords by informational, comparative, or transactional intent and structure repeatable pages around what a person at each stage needs. Ranking results can then feed back into a broader keyword plan.
Scale still needs a quality threshold. Creating hundreds of pages around thin keyword variations is not a visibility strategy. Useful programmatic content starts with a real information need.
8. Glossaries and topic clusters
Glossary pages often get treated as passive SEO content, but they can become useful entry points to your website. Think about someone who finds your page after searching for a term they don't understand. They probably don't know your brand yet, and they aren't ready for a sales pitch. The page needs to answer their question first.
An agent can monitor which definitions attract traffic and identify the related questions people are asking. If a page starts performing well, it can suggest a more detailed explanation or connect the reader to a relevant article or guide. It can also track which links people actually follow.
I would keep these pages educational and use the agent to make them more helpful over time.
9. Google Search Console and analytics-driven optimization
One useful play is to find pages with high impressions but low click-through rates, inspect the queries creating those impressions, and propose better titles, descriptions, introductions, and internal links. The workflow can then measure the change before recommending another edit.
After the update, leave the page alone long enough to see whether the change worked. If the agent keeps rewriting it before there is enough data, you are only automating activity, not improving performance.
10. Cross-channel content repurposing
This last one is easy. An agent monitors your blog, podcast, newsletter, and YouTube feeds. When new content is published, it adapts and schedules it for the right channels.
Someone might discover an idea through search, see it again on LinkedIn, hear the full explanation in a podcast, and finally subscribe through a newsletter. Each version should feel connected without being identical.
The agent can handle the distribution, but you still decide how the original idea should be adapted. Otherwise, repurposing becomes copying and pasting the same content everywhere.
What didn't work particularly well
The conversation also covers what created more noise, cost, or risk than value. That included generic cold email generated at scale, social media automation without thoughtful review, broad “AI employees,” low-quality content, and agents that cost more than the work they replaced.
The common problem was not the technology. It was giving an agent a job that was too broad without clearly defining where its responsibility ended.
For example, if one agent finds leads, writes the outreach, sends it, follows up, updates the CRM, and books meetings, one mistake at the beginning can affect every step that follows. It also becomes difficult to understand what went wrong and when a person should have stepped in.
Boundaries define what the agent can do, what it should never do, and when it needs human approval. Without them, you may spend more time monitoring and correcting the automation than you actually save.
Why I think fewer, more capable agents can work better
I understand why teams separate every function into its own agent when they begin experimenting. But as the workflows mature, more agents can also mean more handoffs, duplicated context, maintenance, and places where something can quietly break.
I think a stronger approach is to consolidate related jobs into fewer AI workflows. Each one still needs a clear purpose, reliable inputs, specific triggers, defined approval points, a measurable output, and a human owner.
The foundation matters more than the model. A clear marketing strategy and a reliable reporting loop give the agent something useful to work from. Without them, even a technically impressive agent is mostly guessing.
How to decide what is worth automating
When I decide whether something should be automated, I begin with the friction instead of the tools. I ask whether the task happens often, whether the inputs are reliable, whether I can recognize a good result, and where human review belongs. Most importantly, I ask whether it will improve the customer experience or only increase the volume.
Automation should make the operation faster and the experience more coherent for the people on the other side of it. If it only produces more activity, it is probably not the right thing to automate.
Conclusion and next steps
In conclusion, AI agents can create real leverage in marketing. They can surface intent, connect behavioral signals, reduce repetitive work, and help teams learn from their data faster.
But leverage works in both directions. When the strategy is clear, an agent can extend it. When the strategy is scattered, an agent can multiply the confusion. For me, the lesson is not that every company needs ten agents. Each agent needs a clear job inside a system designed around the customer journey.
The next step is simple: start with one useful problem, give the agent good context, decide where human judgment belongs, and measure whether your new AI agent workflow creates value. Improve it before adding another one.
This reflects the UX Marketing methodology, which connects strategy, execution, and reporting around the customer journey, and it is the better way to build a go-to-market system that lasts.