AI-First Evolution: Killing the Product Marketing Battlecard

In the previous episodes of this series, we explored how different teams at SafeBreach are adopting AI-First methodologies:

• Development teams use:
  • Product requirements documents (PRDs) to control intent
  • Test-driven development (TDD) to validate implementation
• TAM teams use structured protocols to control truth and reliability

But there’s another dimension to AI-First transformation that is just as important: Sometimes, the right answer is not to improve a workflow—but to replace it entirely.

To explore this, I sat down with Tova Dvorin, Senior Product Marketing Manager at SafeBreach for our sixth installment of the series. Tova leads competitive intelligence and sales enablement—a space traditionally dominated by battlecards, static documents, and… let’s just say, questionable usability.

The Moment It Broke

Yossi

Tova, let’s start at the beginning. When you joined SafeBreach, what did competitive intelligence actually look like?

Tova

Honestly? It looked like… a collection of things. Some spreadsheets. Some battlecards. A few scattered documents.

All I knew is that I was in charge of competitive intelligence now, and that meant also learning about breach and attack simulation (BAS) while catching up on major industry players. It was a lot. 

Yossi

That sounds like a well-structured onboarding experience.

Tova

Very. I opened the battlecards expecting them to guide me. Instead, I found myself asking:

• Is this accurate?
• What’s missing?
• Why is it structured like this?

I had no baseline. No way to tell if anything I was reading was correct. At some point I just sat there thinking: “What am I even looking at?”

Yossi

And at that point you assumed it was just onboarding friction?

Tova

Exactly. I thought maybe I just needed more time. But after a week of going over the same material, I wasn’t getting closer to clarity. I was getting further from it. That’s when it stopped feeling like a “me problem.”

The Reality Check

Tova

So, I went to the people who were supposed to use this every day—the sales team. I asked a simple question: “How useful are the battlecards?”

Yossi

Let me guess—mixed feedback?

Tova

Not exactly. Every single person gave me the same answer:

“Wait…we have battlecards?”

Yossi

That’s…impressively consistent.

Tova

I have to give the sales team credit for that, by the way. Product Marketing Managers work with many different teams. It was great to receive the same honest appraisal of the situation—even if that presented a challenge. 

And that was the moment I realized this was not a “fix the battlecards” project. This was a “kill the battlecards” project.

The Deeper Problem

Yossi

So what was actually broken?

Tova

Everything that matters for something to be usable. Battlecards were:

• Static
• Inconsistent
• Hard to interpret
• Not aligned to real sales conversations
• And most importantly—not used

There was also no single source of truth. Knowledge was scattered across decks, buried in people’s heads, and spread across disconnected documents. This meant onboarding was hard, updates were painful, and scaling was impossible.

Yossi

So basically—there was a system that existed, but didn’t function. How did you plan to tackle this challenge?

The First Attempt (a.k.a. “Vibe CI”)

Tova

I started with the classic AI-first approach:

• Upload documents into ChatGPT
• Ask questions
• Generate outputs on the fly

Yossi

Also known as “Prompting and praying.”

Tova

Exactly. And the results were… inconsistent. Sometimes great. Sometimes wrong. But always unpredictable.

There was also no standardized structure, no prioritization of sources, and a high hallucination risk. And, at some point, I realized this wouldn’t scale.

The Shift: From Content to System

Yossi

So what changed?

Tova

I stopped thinking about battlecards as documents and started thinking about competitive intelligence as a system.

Yossi

That’s a big shift.

Tova

Yes—and it led to three core components.

1. A Structured Source of Truth

A comprehensive, standardized dataset that included:

• Market data
• Product capabilities
• Messaging
• Pricing
• Customer perception

2. Curated Inputs

Instead of “throw everything into AI,” we strategically fed it:

• Competitor websites
• Product team insights
• Sales feedback
• Customer reviews

3. AI Interface

We utilized an LLM layer that:

• Ingests structured data
• Provides contextual answers
• Generates outputs on demand

That’s when everything changed. We moved from static documents to a dynamic, queryable intelligence system

The Reality of Non-Developer AI Adoption

Yossi

This is interesting because your users are not developers. How is AI adoption different in your world?

Tova

Completely different. Developers are comfortable experimenting. Non-technical users are often not. The main challenges are:

• Skepticism about accuracy
• Lack of prompting skills
• Preference for familiar formats

Yossi

So what actually works?

Tova

I tried not to necessarily sell this as something AI-specific. Instead, I presented it as a practical assistant with clear use cases, predefined prompts, and structured outputs.

Yossi

So less “look at this cool model” and more “this will save you time tomorrow.”

Tova

Exactly.

The Method Matters

Yossi

This sounds very similar to what we’ve seen in other teams. Organic AI usage works…until it doesn’t.

Tova

Yes. The turning point was becoming methodical. Specifically, making sure there was:

• A clear problem scope
• Structured inputs
• Controlled sources
• An iteration loop: test → tweak → maintain

The result was more accurate outputs, more consistent results, and higher trust.

Yossi

So the same principle again: Garbage in, garbage out.

Tova

Right. AI didn’t fix the system. It forced us to build a better one.

The Hot Take

Yossi

Let’s talk about your boldest claim: AI killed the battlecard. That’s a strong statement.

Tova

Battlecards are static. That alone makes them obsolete.

AI is dynamic but, more importantly, AI shifts product marketing from:

• content creation → insight orchestration
• static assets → dynamic intelligence systems

The New Role of Product Marketing

Tova

PMMs are no longer just creating content. They are:

• designing systems
• curating data
• training internal AI tools

Yossi

That sounds… suspiciously like engineering. So, what’s still hard about the process for you?

Tova

A few things:

• Trust in AI outputs
• Maintaining data quality
• Over-reliance on automation
• Skill gaps in prompting

I’m done with providing one-off updates. Any team that needs access to competitive intelligence needs to:

• Understand what they need
• Engage with the system
• Take ownership of insights

It’s not the easiest transition, especially for sales teams used to leaning on us for intel. It’s still an ongoing process, but I want to stress that the idea is not to pass the work back to those teams; it’s to reduce the back-and-forth so everyone can get to the information that matters to them faster.

Closing

With each blog in our series, a pattern has emerged. AI-First is not just about adopting new tools. It is about:

• Structuring intent (PRD)
• Validating behavior (TDD)
• Controlling truth (Anti-Hallucination Protocol)
• And now, replacing broken systems entirely

Tova’s story highlights something important:

AI doesn’t just make you faster. It forces you to confront whether your current way of working makes sense at all.

And sometimes, the right answer is not to improve it. Sometimes, the right answer is to kill it and build something new.

Stay tuned for future installments as we continue to explore our journey and the ways in which we adapt our methodologies, tools, and environments based on lessons learned. Specifically, we’ll share some of the in-house innovations we have created to support our progress and will continue to explore the impact of AI on other disciplines within our organization.