Marketing with AI Agents:
How to Build a 10x Marketing Engine

Summary

In this webinar, Shiv Narayanan, Founder and CEO of How To SaaS, explains how marketing teams can use AI and agents to produce far more with smaller teams, and what that means for how marketing organizations are structured. He covers the 10-80-10 rule for working with AI, which marketing work to hand to agents and which to keep human, and use cases across the marketing value chain. He also explains why the foundations need to be in place before you deploy agents. In the Q&A, Ardy from the HTS team explains the tools behind an agent-built reporting dashboard.

Key takeaways

• AI and agents won’t replace your marketing team. The value is in scaling how much a core team can produce.
• Use the 10-80-10 rule. Humans define the work (the first 10%), AI and agents handle execution (the middle 80%), and humans apply judgment and editing (the final 10%).
• Keep anything that drives differentiation human: positioning, messaging, thought leadership, creative work, budget and strategic decisions. Hand agents research, analysis, first drafts, monitoring and reporting.
• Build agents from three components: a skill file (the playbook), connectors to your data, and a schedule to run on. Start with the repeatable tasks your team spends the most hours on.
• Expect to front-load time training the AI to produce the output you want. Teams that give up early fall behind.
• Agents can help across the whole marketing value chain:
  • live TAM and segmentation
  • ongoing product marketing audits
  • pre-call sales enablement
  • one-to-one ABM content
  • repurposing one original asset into many
  • daily paid media optimization
  • pipeline forecasting
• Shiv expects many marketing organizations to become four or five people directing a layer of agents.
• Get the foundations right first: a business strategy, a go-to-market strategy and programs that work. Then use AI as a multiplier.

Transcript

Speakers: Shiv Narayanan, Founder and CEO, How To SaaS, and Artie Shlykov, How To SaaS. Lightly edited for readability.

Why are marketing teams asking about AI agents?

Okay, so we did a webinar about a month or a month and a half ago when my new book launched. It’s called The AI Marketing Blueprint. If you haven’t gotten it yet, please go buy yourselves a copy. When we launched that, we did a webinar on the framework of how to stay relevant and visible in a world where AI platforms are becoming a huge part of the buyer journey.

Subsequently or simultaneously, we’ve been doing a ton of work with clients as part of this business. We have hundreds of PE folks that we do work with, and we work on their portfolio companies. We have active clients that we’re doing work for, and as we engage with those companies, we spot a bunch of opportunities. One of the big opportunities that we are seeing pertains to agents in particular. I can’t tell you how many times a week I get asked about agents. Even with our clients, there are things that we are rolling out with them on an ongoing basis that help us execute on engagements faster. We’re able to deliver on certain types of engagements. We do a lot of agency work for our clients as well when it comes to paid media, content, website work, or ABM work. AI and agentic work is really helping us do those things faster.

Internally, marketing teams are often understaffed. I’m sure a lot of attendees know this and are experiencing this in real time. Because of that, we wanted to do a webinar to walk you through what we are seeing. Obviously, the landscape is changing constantly. I’m sure you guys have experienced this where, six months ago, it was all ChatGPT, and then suddenly it shifted to Claude, and there are more tools and more capabilities. There was a new release this week as well, and everybody’s talking about that now. This will keep changing.

What I want to focus on today is evergreen stuff that’s going to apply to any business that you’re in and things that you can start applying to any area of the marketing organization. What we won’t talk about are tips, tricks, and specific prompts. If you want a list of prompts, that’s not what this webinar is. There are plenty of people that have published best practices on how to prompt AI or how to leverage certain tools in different ways, so we won’t focus on that. What I want to get you thinking about is how you can leverage these tools to work through things faster.

What are the two big AI opportunities for marketing?

I’d say there are two big AI opportunities. The first is what we talked about with the book, which is the AI-enabled buyer journey, being visible, and making sure that—since a lot of companies are losing organic traffic or seeing paid media performance decline—you are visible, creating more content, and becoming more prominent as an expert so that AI starts using you in citations.

The second part of it is a question I’m getting asked on sales calls or catch-up calls with investors and CEOs all the time: with AI and agents, how do we reimagine our marketing organization? I actually had one crazy call last week where an investor was thinking about overhauling the entire marketing organization of 25 people. They were like, “If we were to wipe the slate clean and start over, what would the team look like?” As crazy as that question sounds, it is actually a good question because you want to figure out, with what’s technologically possible, how we actually structure a company to deliver as much enterprise value from marketing activities as possible.

Just as a refresher, I want to talk about the buyer journey and some foundational elements before we fully answer that question. One of the things we’ve talked about is that buyers are self-educating by leveraging AI platforms. I’m sure a lot of you are doing this. On the generative search side, this is super relevant, especially if you’re buying a product. There are all kinds of buying keywords and queries that people are asking, and they’re bypassing things like talking to BDRs and SDRs, attending a demo, or going to your website before they actually make a decision. Because of that, you’ve got to be more prominent to build expertise, and you have to let people self-educate on all the different platforms where they can interact with you. That’s the basis of the new book, The AI Marketing Blueprint.

Why do marketing teams need to move faster?

At the same time, there’s this issue of still having to do the foundational marketing work. This is a framework that we’ve been talking about for years: everybody wants the outcome of revenue growth, but you have to build these underlying layers in order to get there. When we meet companies, they look more like this. This could be whether they’re doing $5 million or sometimes even $200 million, where they have a marketing team that is often understaffed and underbudgeted in terms of how many resources and how much program spend they have, and they’re just not able to work through all these areas. We’re always educating companies to work through this pyramid as quickly as possible, brick by brick.

Five years ago, working through all of this could have taken 6 to 12 months for a company. Our value prop is that we will come in and do it. But even then, it’s still four months. The problem is that time is of the essence. You need to move through this stuff as quickly as possible.

I’m going to jump around here because one of the things we’re seeing inside companies is that more companies are missing targets than ever before. There’s less pipeline. We’re seeing lower margins, price resistance, and lower EBITDA margins. CEOs, CMOs, and even CROs are being changed over more than before, and boards are pushing companies to figure out how to leverage AI so that they are more efficient, increase margins, and find EBITDA even if there isn’t as much revenue growth. Because of that, working through this pyramid as quickly as possible becomes essential. This is where AI or agentic work can really add a ton of value. If you think about these four layers here—foundational stuff, programmatic elements, resourcing, and ultimately the outcome—you want AI to be helping you throughout all of those stages as much as possible so that you can do more with less.

How much more can a marketing team do with AI and agents?

Historically, if you had a three-person marketing team, that would be considered under-resourced in a lot of companies. But now, with a three-person marketing team, you can do the damage of a team that previously had 10 to 12 people, as long as you’re using these tools in the right way. That’s what I want to go further into.

Another way to think about this is that without AI, when it was purely humans, we were entirely responsible for execution, so the output capacity was 1x whatever team you had. With AI working side-by-side with you, you can 3x your capacity, and in some ways, maybe get to 5x. On the agentic side, if you can leverage agents to do a lot more of the execution work, there’s 10x capacity to be found inside a lot of companies.

One way we are talking about this and educating our clients is the 10-80-10 rule, where the first 10% is the definition of the task, project, or area that you want to work on; that is human-oriented, or the input is from the human side. Then you have the middle 80%, which is execution. Then you have the last 10%, which is the judgment, discernment, editing, and fine-tuning of whatever has been created. If you start to deploy this across all different areas inside the marketing organization, your ultimate output from what you’re investing or the core team can go up drastically.

Another way to visualize this—and we’ll spend some time on this just to expand on it. I see people are raising their hands. If you have a question, can you just put it in the chat, please? At the end, I’ll go through the webinar chat and try to answer as much as possible. We’ve also got some HTS team members here because they do client work directly, and I’ll have some of them jump in to answer some of your questions as well.

Another way to think about this is: the first 10% is defining the actual deliverables or what you’re trying to create in as much detail as possible by giving as much context as possible. This could be about your ICP, your business, your revenue numbers, or your average deal size. The more context you give, the more contextual the output will be. The middle part is where you’re leveraging AI platforms and agents to actually do the work—analyzing data, coming up with outputs, or even building creative in some cases—to provide you with scale. The final 10% is based on the outputs you get. You actually have to work with the model. A lot of people give up here because the outputs they’re getting aren’t exactly what they would want. AI platforms are still learning what you want an output to look like. You have to invest the time to save things to memory, and as the platform learns what you’re looking for, the output will look better, and then you’ll need to spend less time on the back end.

A big thing to level set on: because we do so many of these calls, I get calls from investors asking, “How do I get more AI to do all of this so that I barely need any team members?” It’s a bit of a different question than reimagining the org side because it’s almost as if there is a belief that you can replace the work humans are doing with AI, and that’s not true. What you actually want to do is scale how much your core team can produce with the help of these platforms so that you’re creating more enterprise value. That is a far better value proposition than cutting a team and thinking that you could keep running things the way that you’re running them. Critical human inputs are still required.

Which marketing work should you hand to AI, and which should stay human?

Another way to think about this is as the marketing value chain. If we think about the four foundational steps: anytime you’re doing any type of marketing project, there’s the research, analysis, and brainstorming phase; then there’s the actual strategy—where we’re going to focus, the road map, or the planning; then there’s the actual work of doing the creative, running the campaign, and creating the content; and finally, we have reporting and infrastructure. There is no way to fully outsource any one of these things if you actually want to have quality outputs and quality marketing being done for your business. However, you can outsource a big chunk of some of these boxes if you do it the right way.

One very important consideration: AI does fall short all the time, even if you configure it in certain ways. If we think of it as training data: anything in green here is above the mean; anything in pink here is not a great source of training data, but it’s out there. A lot of AI outputs will regress to the mean. It would be very hard to get an AI platform to produce, for example, a new movie like The Odyssey that’s coming out, because it’s completely new. For things that are creative, thought leadership-oriented, and foundational to what your positioning and messaging are—the things that drive distinction and competitive advantage—those should not be outsourced to AI. But what you can outsource is heavy directional research and analysis, ideation, and brainstorming. You could use it to speed up your execution. You could even get it to make assets for you under your direction. But you need to be the director. You need to channel where this thing is focusing its energy rather than using it as the ultimate decision-maker.

Coming back to this: if you create poor content, AI can—if you’re below the mean—help you get to the mean. This is one of the reasons why, on social media platforms right now, you’re seeing a lot of noise from mediocre content; everybody can get mediocre content created in about five minutes with a few prompts, and that creates a lot of noise. In a world like that, differentiation and originality—which we talked about in the last webinar—become a huge advantage. We want to lean into those things, but we want to leverage AI to speed things up so that the quality of our outputs is much better.

Another way to look at this is: what are agents doing versus what are humans doing? An agent could handle competitor analysis or monitoring for you. It could create a first draft or come up with a list of potential content ideas. It can summarize reports for you. On the human side, you want to be the one interpreting the data it brings back. You want to decide how to allocate budget and strategically which areas to focus on. Those are decisions we don’t want to outsource.

If we take this a level down to actual execution and things that you encounter constantly: if we look at the main functional areas like TAM and segmentation, product marketing, sales enablement, website, paid media, SEO, GTM, and ABM, there are some things you want to use AI for and some things that you don’t. As an example, for your website key solutions pages: if you give AI enough context about your business and product, and give it examples of other pages you’ve created, you can create all kinds of great web pages and then work back and forth with the platform. That type of content is not a thought leadership piece. It’s more like product marketing or sales enablement content. You are providing the context, but it’s just doing the execution of creating a web page. That web page may have taken one team member’s entire week to put up a proper landing page with the right messaging. What you don’t want to outsource: while you’re building that page with AI, you don’t want it to decide what the core content and messaging should be.

Similarly, you can leverage AI to build battle cards or sales playbooks for you, given that you are giving it the answers on things like what makes you better than competitors, what distinguishes you, and where you win and lose. It can do the win-loss analysis for you, but you have to analyze it and say, “We may win these kinds of deals, but these are deals we actually don’t want because it’s not our ICP.” That is your judgment and discernment, but on the battle card side, you can produce them 10 times faster than you previously could.

Looking at all these different areas, I would say that when we come into companies, almost none of this is currently being done by AI. People are trying to figure basic things out, but they’re not leveraging AI or agentic workflows to create this stuff at scale.

Another great example is ABM. One of the biggest challenges with ABM—especially if you’re doing one-to-one ABM—is that you have to produce one-to-one content for all your named accounts. If you have intent signals from a large account, you should start targeting them today, but you need people to build the content. If you train an agent correctly to build the right assets for you, in the type of output the sales team can leverage, you can create one-to-one content at scale in a way that you previously could not. Yet, when we go into companies and they’re doing ABM, it’s not one-to-one or customized, even though this technology is at their fingertips. They’re still either manually doing it or not doing it at all. Not doing it at all is more common than any version of that.

What does a reimagined marketing organization look like?

All of this eventually comes back to that question I was asked: “What does a reimagined marketing organization look like?” On the left is what I would call a traditional marketing org. You have a CMO and functional leaders for content, demand gen, RevOps, or corporate marketing; then you have managers and specialists underneath them. What we’re moving towards is far fewer specialists, managers, and agencies, and just a larger agentic layer that these people are pulling upon to generate even more output than before.

This is an important concept, especially if you’re not in the room with your board or your CEO. These are the conversations that are actually happening, whether you like it or not, or whether you’re worried about your future career or job redundancy. If this is the conversation, the question is: what does this agentic layer need to do? Who do the people above it need to be, and what do they need to focus on to add value to the organization? If you’re not doing those things, you will eventually be made redundant.

Where this is eventually going: I think the future marketing organization is four or five people in most companies, where you have true leaders leveraging agents for all kinds of tasks. Because they have the expertise, they’re acting like a director and managing all of that in the right way. We’re not close to this yet, but this is where it’s going. Maybe this will be the case in two years, or you’re going to see more companies look like this. If this is where things are going, you have to start reimagining roles inside the organization. What it means for you as a marketing leader, or if you’re a CEO on this webinar thinking about your team: what does your CMO need to do to build this type of organization?

This scales to different types of models. If you have a $3 million business, maybe you have one marketing leader or one or two people, and for the rest, you’re leveraging AI and agents, which brings cost efficiencies. If you’re a 200-person marketing organization, it might just be that you need 50 people, not 200, to do the same amount of work. That seems like a drastic reduction in force, but you have to ask yourself: can the tasks of the 150 people not at that company be replaced in the execution layer by agents? There are a lot of tasks like that. I’m going to show you some examples shortly.

The last piece on the financial or conceptual side: the financial impact. Imagine this column is the current state of a business: team costs are 1.8 million, and they have a tech stack of 500,000, for an overall marketing budget of 3.5 million. Their return on marketing spend is about 71 cents per dollar spent. If they go this way, the team can theoretically shrink—let’s say that goes down to 1.1 million—they potentially can spend more on programs, and they’re obviously going to spend more on tokens and usage with AI platforms. Let’s say their tech stack increases by 500,000, but their budget remains the same. Because this group is way more productive, they end up producing more pipeline. Even if this group produces similar pipeline, there’s a potential to produce way more down the road because of the efficiencies found by this approach. Again, these are the conversations happening at the board level.

Rethink what it means for every single role inside the marketing organization and then bring it back to say: “This is our current state. This is what we’re moving towards as version one or version two, and eventually we want to be here. Here are all the projects between these different phases and what people are going to be working on.” At the same time, we’re going to work through this entire pyramid as quickly as possible because we figured out how to leverage agents as quickly as possible. The key thing here is to incorporate AI and agents into all of your work. If people aren’t doing this in their specific roles, it means that, more likely than not, that role or person will get phased out. That’s what you need to be pushing as a leader.

How do you build a marketing agent?

Talking about tactical things: I’m going to give you an overview, then we’ll get to Q&A. As more details are needed, we’ll have people from my team jump in to answer questions.

On the setup: the way we have been finding success with this is with three components. One is your actual skill file, which is the playbook of what you need the AI platform to do. Then you have your connectors. This is what gets the AI access to your files or data, which is critical context for it to actually be able to give you the types of outputs you need. Then you have some sort of schedule it runs on to automate and produce things for you. Depending on what type of agent you’re building, the variations will look different, but these are roughly the three components. I’m sure some of you are already doing this, so I’m sure it’ll resonate.

You can do this for any type of problem. You can do it for identifying content, running paid media campaigns, building a sales playbook, battle cards, or data aggregation. As long as it’s a repeatable process that you or someone on your team is doing on an ongoing basis, it can be trained as a skill, saving your team a bunch of time. A good way to think about this is: what are the repeatable activities on your team? If you were to stack-rank them by the number of hours spent per week, the highest-ranking things are likely what you should turn into skills. That’s going to save your team time that can then be allocated to other areas.

As you’re building this, you’re describing the skill or what you want to create, you want to see the end outputs it’s producing, and then you need to refine and repeat the process until you get the output you want. I mentioned this earlier, but a lot of teams are giving up around stage four or five because the output is not at the level they expect or want, or they’re not happy with the quality. The reality is that you actually have to work with the AI platform to get to the type of output you want. There is a bit of front-loading of time. There’s resistance inside teams because they also need to finish things yesterday. There’s a bit of, “Should we front-load this investment to get to an eventual place?” The answer is yes, because if you don’t do it, eventually, you’re going to be much slower than your competitors. You have to somehow front-load this time in whatever role you’re doing. Again, it comes back to the organization side. The people that do this are the ones that are going to survive as roles are slowly phased out, and agents will be leveraged for a lot more of these things.

As you build through all the different skills, there’s a bunch of tactical things you can reference. You can build in strong anchor files or have different components on the AI side making sure that you’re evolving the skills over time. A big part of this is having the context inside the platform to help you with different things. I’ll share this afterward to avoid jumping too much into the tactical. Again, the idea is that you’re trying to teach the AI platform to produce the types of outputs you want. Teaching and training it to be effective is a way better investment than trying to pick and choose the best tactic or tool today. If you create a custom connector today, it might be that, in two months, there’s an actual connector out there, and you never needed to produce it. You could have used an AI platform to produce a connector, but you wasted that time because it’s already on the roadmap. On the flip side, the more you’re teaching it, interacting with it, giving it these plays, getting outputs, and configuring things correctly, the more it’s going to serve you as you continue to build your business because it’s going to have all that context saved in its memory.

Where can agents help across the marketing value chain?

I’m going to give you a few more use cases, and then we’re going to jump into specifics.

Reporting and paid media

Here’s an example with some of our clients to bring it to life. For one client, we’ve been able to stitch together ad platforms, CRM data, data from Search Console and GA4, and connect it all into a dashboard that the agent is constantly updating. Historically, this would have taken us weeks. It would have also required more experts, and now we’re able to do it significantly faster for all the clients we’re working on. This type of dashboard is what you need to ensure you’re spending on the right things, optimizing the right campaigns, and for board-level reporting. Often we see when we get inside companies that there isn’t a RevOps expert; they need to hire one, or it’s a major skill gap. I see somebody asked, “What’s the stack that enables the dashboard?” I’ll get Ardy to answer that shortly, but you can do this fairly simply with Claude and some basic tools.

Another example is budget projections and pacing. We have Google Ads budget pacing for multiple client accounts that we’re managing. You can see how much is being spent daily and what’s being spent on specific campaigns—whether we should scale something up or down, or if there are conversion rate issues. All of these things are being pulled in, and we can make adjustments as we go. There’s also ongoing data analysis being pushed into Slack for different accounts. We can see how much was spent yesterday, or areas where there is a gap, or things that are stuck, or keywords not working. We’re just interacting with the agent on Slack to optimize spend. We’re doing it in a much faster loop. If you work with an agency or do it in-house, most companies do this weekly. With an agent, you can do this daily. The feedback loop is just much faster. You can save more money and be way more efficient with your existing marketing spend.

These are just examples. Paid media or data are easier because they jump off the page. There’s money and data involved, so leveraging agents for that becomes use case number one. For most of the clients where we are deploying stuff, paid media is one of the first places we look because so much of it can be automated. The reality is that you can use it across the value chain, yet we’re not seeing teams do that, and there are opportunities across the board. I’m going to work through all these boxes and then shift to Q&A.

TAM and segmentation

On the TAM and segmentation side: the way we’ve historically done TAM and segmentation is companies or PE firms hire someone to do a large market mapping exercise to define TAM, SAM, and SOM. They’ll say, “Here’s where you’re more likely to win, or here’s where you have a right to win. Here’s how many of those people are out there.” Then you go into ZoomInfo, upload the people, and start an outbound cadence or set of campaigns. That’s good, but the problem is that it’s static and you don’t get data to continuously improve things. You could leverage agents to update your ICP and segmentation on an ongoing basis. You can pull in intent signals and competitive monitoring of any changes to pricing on an ongoing basis. That can change who you’re segmenting or focusing on on a weekly, monthly, or quarterly cadence. Those studies cost tens of thousands of dollars, but in six months, they’re not relevant anymore. If you can leverage AI platforms to do these things, you can move faster.

Another way to think about this is the basic product marketing idea of a “job to be done.” You could theoretically have a bunch of agents doing the job that a market mapping exercise would do. You could look at call recordings, sales call transcripts, and win-loss analysis. You could look at people changing jobs, or companies’ specific activities or press releases regarding acquisitions. All of those data points can be automated through an agent to update the segment of folks you’re focused on.

Product marketing

Similarly, on the product marketing side, a lot of companies look at it as a one-time exercise, but it’s a very live thing. We’ve seen it especially where every company now says “AI” above the fold on their website. It’s natural because that’s top of mind inside many companies. Instead of a one-time product marketing exercise, you can have ongoing agents and activities that audit your website against your messaging and architecture. You can look at what your competitors are doing and talking about. You can make sure your URLs and technical SEO are aligned with how you want content on your website to live. At the same time, you can ensure content on your pages is aligned with the messaging and positioning you want.

There are many other use cases, but as you’re building your product marketing house, there are many steps in between. One example: we look at survey answers for NPS ratings, or actual sales call transcripts to see which ones lead to more closed-won deals. There are actual interviews with customers and the feedback they give us. You can automate a lot of that analysis and get through product marketing work much faster because that stuff can be researched, analyzed, and summarized for you. Then, you as a human can take that information and turn it into product marketing language on a much more rapid basis.

Sales enablement

On the sales enablement side, we talked about things like battle cards and core sales assets that help us win more deals, like sales scripts. You can analyze closed-lost deals and reasons, then look at the calls associated with those. Some of these don’t need to be done in ChatGPT or Claude; some can live in the actual tools you’re using. If you’re using Gong, there are AI components that can get you these answers, and you should be feeding that into any agent you leverage to help you create those battle cards. The system of record or whatever you’re connecting to the skill file could be living in another system. But as you connect those things, the quality of the outputs goes a lot further.

Similarly, if there’s a sales call coming up with a specific company or deal, a bunch of contextual content can be fed to sales reps so they’re better prepared and have higher quality sales calls to increase close rates. That’s a really big use case, and it requires work. In a large sales organization, there are sales enablement people doing this work. But in most companies, there isn’t someone doing pre-call sales enablement. You can do that with an agent and improve your close rates significantly. That’s one of the lowest-hanging fruits or biggest opportunity areas.

Go-to-market strategy and demand gen

You want to do the same for your overall go-to-market strategy and demand gen. You could do this for content and SEO. You can generate board reports and look at the overall pipeline and revenue targets. In this case, you can see the agent summary saying we didn’t hit our revenue target or benchmarks for cost of acquisition. Since there’s a gap in the mid-funnel, that identifies an opportunity for us to direct our go-to-market strategy accordingly. Now we can focus on middle-of-funnel content, nurture campaigns, and so on to increase conversion rates. This is adding AI or agentic workflows into things a human would have done anyway—this is what a CMO would do. They would pull up a dashboard, see the problem areas, and review it to make sure the team is focused in the right direction. With reporting, you can improve team focus with much better feedback cycles.

On the demand gen side, we covered what you can automate regarding reporting, campaign structures, and optimizing spend. The big one beyond paid media is content for ABM, because ABM requires so much content to be done correctly. Leveraging agents can help you build a robust playbook—we’re doing this with some of our clients right now so they can build more content at scale. Coming back to this matrix: on the ABM side, you should leverage agents for the account-by-account playbook and whitespace analysis, but the orchestration with sales is something you might work on as the marketing leader.

Content

Regarding content: don’t outsource creative things to AI platforms or agents, because it’s so easy for the content you’re producing to become commoditized. What you can do is produce a core piece of original, differentiated content, then leverage an agent to build assets against that core asset. You maybe produce one video podcast episode, and the agent turns it into 25 pieces of content to leverage across all social media platforms. That is still your original content, but now, instead of paying a social media manager 80k to 120k, the agent is doing that job, likely much faster. You could do this for anything: a webinar like this, a full video, or turning a piece of content into a nurture sequence or a sales script that connects to an ABM campaign. Multi-purpose the content you’re producing.

Customer marketing

On customer marketing: similar to the TAM and ABM work, as you do whitespace analysis, you can figure out which contacts to focus on based on their likelihood of purchasing, which other products of yours they don’t have, or their likelihood to spend. You can look at “voice of customer” data from different sources and connect that to where sales is focused. There are often intent signals there regarding what you could cross-sell or upsell. Extra content can also be produced. One thing we talk about often is how we leverage customer stories more. In customer marketing, that’s a huge part of the value for upselling and cross-selling. You can have an agent pull customer stories or case studies to be leveraged for those campaigns.

RevOps

We talked about RevOps regarding weekly or daily reporting, but you can also do pipeline forecasting. As long as it’s connected to Salesforce or HubSpot, you can have a forecast telling you where you’re trending. You’ll have better reporting and understanding of where you’re headed so that everyone, including your CEO and the board, knows what to expect. Normally, companies have a once-a-quarter cadence and maybe a monthly meeting, but for everyone to stay on top of pipeline and forecasting goes a long way. On the actual data side, there are many hygiene and connector issues. Tracking things correctly can be fixed in many ways by leveraging AI platforms, or at least identifying where the issues are. Where is there duplication? Where is the tracking breaking down? That gives you a list to work against.

Team and budget

Finally, on team and budget: everything we’ve talked about so far comes together here. Where can we deploy these agents, and how does that change our team? Overall, what does that look like for our spend? A big part is also looking at your priorities for the year or quarter and how fast you’re able to move. We talked about the new org structure, but on the budgeting side, how do you set the budget in a world where you potentially have fewer team members and more program or AI spend? How do you tie that to the priorities and overall business metrics like pipeline, revenue, and EBITDA margins?

There’s a lot more here, but the idea is that you should be leveraging as much of this as possible at every stage. Five years ago, it would have been acceptable to say, “Marketing doesn’t have enough resources.” Now, that excuse is not there as much. I’m pushing our own team to work much faster in how we syndicate our podcast, update our website, or run campaigns. The same is true for our clients. Think about this in your own way. Use this as a matrix to think about where you can leverage AI versus not. What is the “secret sauce” that your team brings to the table?

Should you deploy agents before your foundations are in place?

One thing I want to leave you with is that we’ve met companies lately that are very gung-ho on deploying agents. Everything I walked through is important, and I’ve been stressing the urgency. But we have seen companies asking about this when they don’t have anything foundational in place. They don’t even have a proper business or basic go-to-market strategy in place. While you can use AI to move faster, step one is getting the foundation right. If you have to move slower on that side, that’s okay. I am giving you permission for that, and to have that back-and-forth with your CEO: “No, we need to get this piece right before we layer in a paid media agent. We should figure out if paid media is even the right thing for us.” Maybe the best thing is to focus on partners, events, or anything else available. While these agents exist, you have to think about which area the business needs to deploy its limited budget into to drive the most pipeline. Once you figure that out and have programs live where you know the direction, you can start using AI as a multiplier on top of what you’re already doing.

That is pretty much it. We can jump into some Q&A. We’ll send you the slides and the recording as well. I see somebody just asked that.

Questions from the audience

Where do you see agents replacing roles at a regional versus HQ level?

Shiv: Question one: for a reimagined org chart, where do you see AI agentics replacing roles at a regional versus HQ level? It’s a good question. Let’s say you have a company with headquarters in the US, a European org, and an APAC organization. There are local boots on the ground. You’re not going to replace boots on the ground for events or local partners. But you can definitely leverage agentic workflows to scale content and demand gen campaigns to all these regions. One great example is language. Multinational organizations have a tough time in different regions because language is a barrier. ABM at scale also becomes a problem because if you’re targeting a company in Germany, maybe you need to speak German. With agentic workflows, you can do that at scale significantly faster without needing a local German content writer. That changes how you structure your organization. Figure out what is an essential human activity—like boots on the ground or running a local event—versus the many centers of expertise or shared services areas where agents can help you find more scale than if you were to hire a local presence.

This is also true in the social media realm. I see some influencers taking their content and translating it into other languages, actually speaking in that language, to expand into other markets. You’re going to see a lot more of this going forward.

What is the stack that enables the dashboard?

Shiv: Joseph asked, “What is the stack that enables that dashboard?” Maybe, Ardy, you want to jump in on this?

Ardy: Yeah, for sure. The base dashboard is built on Claude. The majority of marketing tools can actually just natively plug into Claude these days. They’re pretty good at this and build their own connectors pretty much daily. For some of the more complicated stuff like ad platforms, we use a tool called Ryze.ai (no affiliation). They’re just a good tool. At some point, they’re going to go out of style as Claude or other folks just start making their own. On the CRM side, you’ll have a bit of a challenge. Salesforce, for example, doesn’t let Chrome tools crawl or use them. Some folks actually have built custom MCP connectors. That’s a little daunting, but it’s not as hard as you think, especially if you’re just extracting data. If you want to make changes, that’s a whole other game. But if you’re just extracting data, go into Claude Code and say, “I need an MCP connector to Salesforce,” and keep pushing the snowball down the hill. It’s not as tough a lift, and the majority of these platforms can natively connect already.

Should you use the AI built into the tools you already have?

Shiv: Yeah. I see Nisa made a statement: she said she’s been using HubSpot’s AI to segment her customer base. That makes total sense. Certain ecosystems are almost like walled gardens; you can’t really take data out of it, or you have to use a workaround. Many software companies are building AI natively into their platforms. Whatever your stack is, spend time exploring that.

A big part of where we are on this journey is that marketers need to spend time playing with all these tools. We don’t play with stuff anymore because, about 3 to 5 years ago, a lot of it was figured out—HubSpot and Salesforce were what they were. It’s like rinse and repeat: you do things in Google Ads and Google Analytics. Now, for the first time, there are all of these tools that enable a bunch of different capabilities. That’s why I didn’t spend as much time on prompts, as they keep changing. I could give you a list of prompts that we currently use, and they would probably be irrelevant in 4 to 5 months. What’s more important: think about where you can leverage AI across this value chain and which systems you have access to—Gong, Salesforce, or HubSpot. Inside those tools, there’s also AI capabilities. Spend time playing with that, and have your team play with it as well. A culture shift needs to happen to actually capitalize on the value that sits here.

I’m looking at the questions in the chat. I think those are all. If there are any more, jump in. We’ll try to address them as we have a few more minutes. Otherwise, we’ll share the recording and the slides. Okay, thanks, guys. This was great. We’ll get you all the resources after this. We appreciate you attending. Please rate the session, and we’ll be in touch soon. Thank you.