Marketers, Your Job Description Just Changed: Building AI Systems That Work for Your Team

Using AI occasionally isn’t the same as having a system. For marketing and sales practitioners, the emerging competitive gap is now between people who use AI reactively and those building repeatable, human-input-driven machines.

Many marketing and sales professionals are using AI on the job. But far fewer have built a working system. That’s a widening gap we can’t afford to ignore.

In Part 2 of my conversation with Mike Kaput, Chief Content Officer at Marketing AI Institute and SmarterX, we got into the execution layer: what it actually takes to build dependable AI-assisted workflows for content and prospecting, how to think about a tool stack without getting distracted and derailed by shiny objects, and what the role of marketer or salesperson begins to look like as AI agents start handling more of the work.

If you missed Part 1, where we covered where most B2B teams actually are on AI adoption and the structural moves organizations need to make to go from piloting to scaling, you’ll want to check that out, too.

Watch episode 40 of The ChangeOver, then subscribe on Apple, YouTube, Spotify, Weidert.com, or your favorite podcast app. You'll be notified when part two of our conversation drops.

AI Flipped Content’s Supply and Demand Equation Upside-Down

Before getting into why it’s so crucial to move from “we use AI” to “we’ve built an AI system,” it’s key to understand how AI has changed the constraints around content production. Just a couple of years ago, much of content work was heavily manual: research, note-taking in interviews with subject matter experts, drafting, refining, repurposing for use across channels and platforms.

Still important work, but now, an AI assist speeds the work and reduces the manual effort. That time constraint around content production is all but gone — at least on the side of volume (quality is another matter).

“You can literally click a button and depending on the input, which is the key here, you can get basically unlimited output. So your limiting factor is not time to produce content.” — Mike Kaput, SmarterX

That shift has a direct implication for how practitioners should budget their time. If production is no longer the bottleneck, it’s now the value and originality of the inputs that go into your AI-enabled content engine.

That’s at the core of the framework Mike described for a two-sided content machine: human-gathered inputs on one end, AI-scaled output on the other.

As generative AI’s capabilities (and our ability to make it do good work) improve, our work becomes all about mining our own data and providing the inputs no AI or competitor can replicate. Think original data, proprietary frameworks, interview transcripts, subject-matter expert thinking and storytelling. It’s about understanding what makes knowledge, experience, and insights valuable to our audiences, and what we can offer that AI doesn’t already have in its model, and can’t synthesize on its own.

A thoughtfully crafted AI system can handle the work of drafting that material into whatever format and voice the distribution strategy requires.

This means the actual job of a content marketer has already shifted. Less time at the keyboard producing from scratch. More time finding, capturing, evaluating, and structuring the original inputs that give the machine something worth amplifying. And more time creating and refining the AI system’s skills, so the outputs it creates are distinctive, valuable, and resonant with the humans (and bots) reading it.

For the writers who leaned heavily on production and volume as their points of pride, this can be a bewildering place to arrive. The stakes have changed. On the bright side, for small and scrappy teams with a deep understanding of their customers, an AI system changes what’s possible. But output quality depends on input quality — so the work of gleaning great inputs is now a critical part of the job.

RELATED: AI + Audience Research: The New Formula for Marketing Advantage

What “Gathering Inputs” Looks Like for a Lean B2B Team

Imagine a small and scrappy sales and marketing team of just a few people running the whole function at a mid-market industrial company. A few years ago, the work of production was nonstop, and time constraints limited the volume of what they could produce.

Now, with an AI system in place, the focus moves to recognizing that language is data — and content fuel. Mike’s advice is practical: stop treating language as something you write, and start treating it as something you capture.

“I don't think you’re served on that lean team to sit in front of a blank page and type out your thoughts. I’d be moving very fast towards having people interview me or interviewing subject matter experts or capturing all the meetings and calls we have in audio and text form, because then this all just becomes raw material that we can endlessly use, reuse, and remix.” — Mike Kaput, SmarterX

Think about what that mindset shift changes:

• Sales calls, discovery conversations, and customer interviews are much more than pipeline activity; now, they’re also raw content material
• Internal expertise locked inside people’s heads has real marketing value, if you capture it and feed it to your system
• The tools you already have (call recording software, AI meeting summaries, a decent frontier AI model) are likely enough to start

Collecting that raw material is now the start of the workflow. It takes discipline and new ways of working to mine that stuff. It takes human discernment to decide what’s good — finding the best quotes, extracting the recurring questions customers ask, identifying the frameworks your best people use to explain complex concepts.

That’s a skill set most marketing practitioners haven’t been trained to develop. It wasn’t the way things worked before. It wasn’t a necessary ingredient.

But it is now, and mastering it starts with embracing a growth mindset, taking initiative, and creating the mechanisms for collecting and evaluating those inputs. That means a new level of organization and governance around assets like call transcripts, video interviews, primary research, and proprietary data.

AI-Assisted Prospecting: What’s Really Useful, and What Can Go Sideways?

The same input-first logic applies to prospecting. Mike was honest about his own learning curve here, and that’s the kind of candor I can appreciate as someone who’s also learning by doing.

The single most important question before building any prospecting workflow is not “what tool should I use?” The better question is whether the outreach is going to have any value to the recipient.

“I don’t mind if someone emails me. I mind when it's completely irrelevant to anything I do. It'd be a joy to be cold prospected for something I was interested in that I didn't know about. Genuinely.” — Mike Kaput, SmarterX

That’s the failure mode AI can amplify. If your relevance question isn’t answered first, AI won’t fix it. But it will annoy at scale. The tool that lets you send hundreds of personalized emails in minutes is a liability if the targeting is wrong.

Assuming you’ve ensured that relevance, AI tools — including agentic tools like HubSpot’s Breeze Prospecting Agent or Claude Cowork — are becoming genuinely useful for the scale problem.

In Mike’s own example, he had a list of 250 contacts, a well-targeted email template, and a goal of personalizing every single message. What once would have taken hours of copy-paste work (and almost guaranteed errors) was done in roughly 20 minutes by asking Claude Code to build an HTML email hub with individual send buttons pre-populated for each recipient.

It’s not magic; it’s more a matter of understanding what’s possible and making it a point of leverage. Once you have the right list, with the right data inputs and the right messaging sets, AI removes the friction of executing personalization at scale. Plus it eliminates the errors that typically creep in after a hundred or so manual emails.

In what Mike refers to as “vibe prospecting,” sales uses AI to crawl publicly available data for intent signals to profile which handful of people on a broader list are most likely to be ready for a specific outreach. When you do this before the send, your segment is better targeted. That can mean real time savings and better outcomes for sales teams that historically relied on spray-and-pray volume.

But watch out: the quality of filtering depends on the assumptions the AI is making, and those assumptions aren’t always visible. Mike advises staying skeptical of outputs, especially when a tool is making inferences about intent from public data. The same critical thinking that applies to AI content outputs applies here — verify before you act on it.

Searching for the Right AI Tool Stack? You May Already Have It

The tool question is where a lot of teams can lose time and momentum to distractions. Mike’s framework for getting started is pretty blunt: maximize what you already have before adding anything new.

Most organizations paying for Microsoft 365 have Copilot. Most teams using Google Workspace have Gemini. Many CRM and marketing automation platforms have embedded AI features that go almost entirely unused. Before evaluating a third-party AI tool, ask: have we tried using what we’re already paying for?

After that comes the frontier model tier — like ChatGPT, Claude, or Gemini — as a team-wide baseline. These tools offer the most versatility for different types of users to develop workflows that support their goals, encourage self-education through exploration, and require the least specialized training to start delivering value.

And the effort involved in learning to use them well also builds the AI literacy that makes every subsequent tool adoption easier.

What about specialized third-party tools? They’re everywhere, touting their capabilities to make the work easier for the user. But the third-party tool conversation is only worth having once the first two tiers are genuinely in use. Here’s why: most of these tools are wrappers or workflow layers built on the same underlying models that power frontier AI products — with a specific UI, a data integration, or a workflow automation on top.

So before subscribing, Mike would ask, do you actually know what specific use case you’re solving for.

• Have you exhausted what you can do with the first two tiers of tools?

• Do you have baseline AI literacy on the team to use it well?

• How will you measure success 90 days from now?

• How much onboarding is required, and do you have the bandwidth for it?

“If you’re talking to me about trying to buy third-party tools and you haven’t built your own GPT yet or started to experiment with the agentic capabilities in ChatGPT or Gemini or Claude, then I don’t know what we’re doing here.” — Mike Kaput, SmarterX

The failure mode Mike describes is familiar to anyone who’s watched a technology rollout die on the vine: millions spent on licenses, handed to people without context or training, and the most sophisticated thing anyone achieves with it is rewriting their emails. That’s a change management failure, not a tool issue.

AI Agents Are Here. What You Need to Know Before They Show Up in Your Workflow

Agentic AI isn’t coming. It’s here. The question is whether your team is ready to work with it in any organized way.

Mike recommends treating agent readiness as a distinct preparation effort, separate from general AI adoption. And for very good reason: AI agents are non-deterministic. They interpret instructions in ways you don’t always anticipate and that aren’t always visible to you. They act autonomously, and they need access to systems and data to do anything useful. That combination requires a different level of organizational thinking and governance.

“An agent in a vacuum is not going to do anything useful for you. So you need to think about — and this is really hard — what are and aren’t we comfortable with when it comes to agents? It’s a coworker. You are bringing on a new hire. It just happens to not be a human being.” — Mike Kaput, SmarterX

The practical preparation steps, in order:

1. Develop an agent-specific policy before anything is deployed. What can agents access? What requires human approval? What systems are off-limits?
2. Educate the team on what agents are — not in vendor sales language, but in the real mechanics: autonomous, non-deterministic, prone to unexpected behavior in both directions.
3. Map your existing workflows before trying to hand them off. If you can’t explain clearly how a process works and teach it to a colleague, you can’t effectively train an agent to do it. AI can help you document those workflows, but someone still has to know how the work is done, and own that knowledge.

Teams that skip directly to agents without building the foundation on knowledge and governance can end up with processes that run unattended in the wrong direction.

RELATED: Understanding AI Agents & Their Potential: Real-World Examples for Marketers

Staying Relevant Demands a Growth Mindset

Mike’s closing point in our conversation deserves special attention, because it’s not about tools, capabilities, or shiny objects. It’s about the people — you, me, and our teams. Our roles are changing from doing the work to orchestrating the systems that do the work. That’s an uncomfortable place for many people who chose marketing and sales because they liked the craft of it.

“You are probably no longer going to predominantly be the one doing the work. You’re going to be orchestrating and managing and overseeing, whether it’s systems, a bunch of different chat windows, or probably agents eventually. The best marketers and salespeople are going to inherently be managing these systems.” — Mike Kaput

This doesn’t mean craft disappears. It’s human judgment, taste, strategic thinking, and insight that make great content and great sales conversations. But the distribution of time and effort is changing fast.

Marketing and sales professionals who position themselves as exceptional at finding and structuring original inputs, who understand how to direct and evaluate AI output, and who are comfortable managing systems rather than just working inside them will continue building durable value in this environment.

And those colleagues waiting for AI to “settle down” before engaging with it seriously are making a choice that will be much harder to reverse 18 months from now.

If you’re looking for a deeper framework on building AI fluency, check out the SmarterX AI Academy, where you can find coursework covering everything from baseline literacy to evaluating tools to mapping AI's impact across specific roles in your organization. You can also follow his weekly analysis on The Artificial Intelligence Show podcast.