Hello Reader
Your team just spent two weeks building an outbound sequence that generated four replies, two of which were out-of-office messages.
If you’re experiencing this scenario then it will come as no surprise that reply rates on B2B email have dropped from 7% to 3.43%, according to Instantly’s 2026 Benchmark Report.
At the same time, an estimated 376 billion emails are sent every day, a number expected to climb past 408 billion by 2027.
With the math working against you, what are you supposed to do?
Chris Arden, a Fractional CMO and GTM engineer at Arden CMO Advisory in Dallas, decided to build a different kind of outreach.
When Everyone Has the Same Trigger
To understand what Chris built, it helps to know how most B2B outreach works today. Tools like Clay and Apollo are sales intelligence platforms that monitor business signals such as funding announcements and new role appointments and automatically pull contact data for decision-makers at the relevant companies.
When a signal fires, the platform launches a pre-built email sequence.
The problem is that every B2B vendor with a similar ICP is running the exact same playbook.
As Chris put it: "Everybody's already got their Clay signal set up for companies that just raised funding and then they're blasting them with generic messages."
Which means the CEO of every newly funded startup gets 60 near-identical pitches all at the same time.
Chris’s answer was to build a system where every outreach asset is specific to the company receiving it, built on demand, by an AI agent working autonomously.
The Build, Step by Step
Chris has 20 years of demand gen experience and a background heavy in data analysis. He used that analytical instinct to learn how to build in Claude Code, starting from a single prompt and iterating from there.
He had the full system running in under six hours over one weekend.
“I try not to do one huge prompt,” Chris says. “I give Claude code the context and then direct it step by step. If I give it too much information at once I find that it can get confused.”
Each step below reflects that method.
Step 1: The Research Agent
Chris subscribes to multiple newsletter sources, including Venture Daily Digest, and Crunchbase. Each day, he receives the emails that outline new funding announcements across a variety of industries.
He built an agent using Claude Code that reads his email, parses the details, and identifies companies that raised between $10M and $100M and filter by sector: Vertical SaaS, Blue Collar Tech, and AI companies.
The agent runs the process automatically, pulling qualified companies from each newsletter as it arrives.
Step 2: The Scoring Agent
Before writing the outreach, Chris created an ideal customer profile (ICP) brief so the AI Agent understands his target audience. This one page brief defines exactly wh
o he targets, what disqualifies a lead and which signals indicate a strong fit.
He uploaded the document to Claude and built a separate scoring agent that runs each company from the research step against his ICP criteria and grades the match. Only accounts that clear the threshold move to the next stage.
Everything else drops out.
Step 3: The Staging Database
After the agent pulls the details from the newsletters and scores the accounts, only those that meet the threshold score flow into the Google Sheet.
This is the staging database for the system.
Each row captures company name, domain, funding amount, the CEO and CMO’s names and their LinkedIn profile URLs.
A second research pass then layers in value propositions, who the company sells to and what differentiates them from competitors. By the time an account reaches this stage, it has a full brief.
Step 4: Landing Page Generation
Next, Claude builds a dedicated landing page for each start-up and hosts it on the Netlify platform. Each page utilizes background research, mirrors the company’s branding, including colors pulled from their website and background imagery matched to their industry.
The page also maps Chris’s services to the company’s ICP, lists a custom 90-day growth strategy and routes the CTA to his Calendly so there’s no manual follow-up..
For R1 Therapeutics, a pharma company that raised $77M, the page (view here) includes specific target markets, maps Chris’s playbook to their growth stage and outlines sector-specific strategies.
As Chris described it: “Custom landing pages at scale are becoming a reality now, and because you’re using research and account information you can create each page with a high level of personalization.”
If your site runs on Webflow or WordPress, the landing page layer can be piped directly into your existing CMS. Netlify is the fastest path to a working version, says Chris but not a requirement.
Step 5: The Outreach Layer
Now that the landing pages are live, the next step is outreach.
Chris uses Instantly or SmartLead for email and Dripify for LinkedIn outreach. Highly personalized messages, connection requests and follow-up sequences are ready to deploy, with a human approval gate so Chris can review everything before it goes out.
What used to require a developer, a data team and several weeks now runs automatically, triggered by a VC funding newsletter hitting an inbox.
"I want to be the first one to reach out to these companies," Chris says. "But I want to do it in a way where they're impressed with the amount of research and customization for their brand. I haven't seen many people do it at this level yet."
What to Watch Out For
Chris built a working system that lets him spin up landing pages at scale, but it wasn’t a smooth ride the entire way.
Here are four issues worth knowing before you attempt it yourself:
Gotcha #1: Claude will send the email if you let it.
In Chris’s words: “I told it to write drafts and it sent an email. I was like, oh God, what are you doing? I hit the undo button really quick.”
If your agent is connected to an email client, “draft” and “send” need to be explicit, separate instructions in your prompt. Assume nothing.
Gotcha #2: Things break after updates.
A build that works on the first run can fail on the second. This is a recurring pattern in Claude when the underlying model or tool updates.
Budget debugging time separately from build time. A working system needs ongoing maintenance, not just an initial setup. Chris is now moving away from Google sheets to Notion databases for better scale and consistency.
Gotcha #3: Hosting limits arrive faster than expected.
137 pages on a free Netlify tier triggered credit limits. Factor infrastructure costs into your planning before you deploy at scale, not after.
Gotcha #4: LinkedIn doesn’t like raw automation.
Chris uses Dripify because it operates within LinkedIn’s guidelines. Automating connection requests directly through Claude risks account suspension.
His rule: checking profiles is acceptable, automating connection requests is not. When you connect the outreach layer, set the approval gate before anything else.