This week for the AI for GTM newsletter, we talked with Dan Cotet, Chief Product Officer at ZeroBounce, who leads the platform's strategy and development, as well as currently driving AI adoption across ZeroBounce's GTM teams to help each function work better. |
|
5 steps to build a competitive intelligence AI agent |
|
You know you should be tracking competitors more closely. Latest pricing changes, new messaging, feature launches that directly affect your pipeline in the upcoming quarters. But somehow it never happens consistently. We spoke with Dan Cotet, Chief Product Officer at ZeroBounce, whose team had exactly the same problem. But instead of investing in another tool or making it someone's weekly responsibility, his team built an AI agent that automatically tracks competitors and delivers a summary every Monday. In this edition, we'll break down exactly how they built it so you can set up something similar for your team. |
|
First, build the internal context files |
|
Dan starts by creating a set of resource files that the agent reads before every run. It is basically to give the agent the business context it needs to evaluate competitors the way the ZeroBounce team would. This included: - A company overview
- Product descriptions
- The list of competitors to track
- What matters most to the business
- How to judge impact
- Examples of good and bad competitive updates
|
|
Here is what the build actually looks like |
|
1. Collect competitor dataThe first automation in the system is collection. It scrapes the web, socials, and Reddit every week for each competitor. This collection data is fixed across four dimensions: - Pricing and packaging changes
- Product and feature changes
- Funding, M&A, and leadership updates
- Customer sentiment from places like G2 and Reddit
|
|
2. Score the data The second layer of the agent is to score the data scraped from online. For example: - Low impact changes like minor website copy changes or product announcements that don't affect positioning.
- Medium impact, such as new features that are useful but don't directly compete with your core offering.
- High impact, such as pricing changes, major launches, or positioning updates that directly challenge Zero Bounce’s offerings.
|
|
3. Generate the weekly briefOnce the high-impact changes have been identified, the agent turns them into a concise Slack memo, which is delivered automatically every Monday. For each competitor, the brief follows the same four-part structure: - What changed
- Why it matters
- Suggested action
- Impact rate
|
|
4. Improving the agentOnce the tool was built, the primary focus shifted to output quality. One thing Dan noticed was that a lot of the team's judgment about competitors was nuanced and rarely showed up in logic files. So instead of asking people to document their thinking upfront, he sent the first version of the system directly to the team and let them use it. As they reviewed the output, they naturally pointed out what was missing. |
|
One thing to keep an eye out for: ownership clarity Most GTM teams are made up of people with no technical background. The product or engineering team, meanwhile, has its own workload sitting on top of whatever it's asked to build. That combination is exactly where ownership tends to tip too far in one direction. Either the technical team ends up owning the whole thing or the non-technical team gets pushed to the sidelines. Dan's workaround for now: "You do not need a marketer to be technical; you need them to be specific." Marketing's job is to define what good output looks like, provide real examples of right and wrong, and review what comes out the other side. That removes a lot of work from the technical team as they don’t have to figure out what good looks like and can focus on building the system. At the same time, Dan doesn't think this is the long-term answer. The current split keeps the work balanced, but it still routes everything through a back-and-forth between two teams. What he wants to do next is a cultural shift: instead of one centralized AI team building everything, each person owns their own setup. |
|
AI news: What's happening right now? | - Adobe just rolled out even more GenStudio updates, this time adding Commerce Media Networks support, so brands can advertise across retail media networks using their existing content, plus a "Simulate Skill" that tests how audiences might react to creative before any ad spend goes out. For GTM teams: you can now pressure-test campaigns on synthetic audiences before launch.
- Aurasell just launched Agent Builder, letting GTM teams build and run AI agent workflows using plain natural language, no code required. For GTM teams, this means complex, multi-step workflows that once needed engineering support can now be built and deployed by anyone on the team.
- CiteLens just launched its GEO platform, tracking how often brands get cited and recommended across ChatGPT, Claude, Perplexity, and Google AI Overviews, not just where they rank in traditional search. Marketing teams can now easily see which sources AI models actually trust and get prioritized fixes to earn more of those citations.
|
|
|
| That's all, folks! I'll see you soon with more tips and ideas. If there is something we can do better for you, please let me know by replying to this email. |
|
Until next time,
Aquib
CEO, Mailmodo |
|
|
|
