A strong product, an AI "marketing office", and an outbound engine that wasn't producing replies.
ContentBuilder is a US AI training platform. It turns a company's existing knowledge (docs, SOPs, recordings, policies) into structured, AI-narrated, multilingual training with full LMS delivery (SCORM/xAPI, SSO/SCIM). Its sister module, Lumi, handles HR work: writing job descriptions, posting roles, screening applicants and running first-round interviews.
I came in as Marketing Manager on contract. The company had an AI "marketing office" with departments and agents already set up, but none of it was connected to real platforms like Search Console, GA4, social publishing or Canva. Cold email was going out, but replies weren't coming back.
The copy wasn't the problem. Microsoft was sending every email to spam.
The easy assumption was bad copy. Before rewriting anything, I ran a full inbox-placement test across every sending mailbox and read all five sub-reports:
- Google → Google: 96% inbox. Content, SPF, DKIM and DMARC all passed. Spam-filter score 0, no blacklists.
- Google → Microsoft: 0% inbox, 100% spam. Every email to an Outlook or Microsoft 365 address was landing in junk.
- That isolated the failure to sender reputation with Microsoft. Content, authentication and blacklisting were all ruled out. For a B2B audience, where a large share of companies run on Microsoft 365, that alone explains most of the missing replies.
The fix plan: register with Microsoft SNDS/JMRP, add 2–3 sending domains on a second provider, cap volume at 25–40 emails per mailbox per day, segment sends by the recipient's mail provider, and turn off open tracking. I also sized what the client's high-volume target would take in reality: about 200 warmed mailboxes across many domains, with a 3–4 week warm-up that can't be skipped or bought.
Rebuilt the ideal customer, and found one trigger that beats industry or headcount.
- ICP rebuilt from scratch when the focus moved to the HR product: from HR/L&D directors at 100–2,000-person companies to owners, founders and COOs at companies with no dedicated HR person, who feel the hiring pain directly
- The best targeting signal: 5+ open roles posted right now. It's public, dated and free to find, and it predicts buying intent better than industry or company size
- 15 verticals researched, narrowed to 3 for parallel testing: home & field services, home health & senior care, and multi-site salon/restaurant groups with 10+ locations
- Five full ICP + offer pairs for the core training product, plus a 40-company manual lead-fit review to check that real leads matched the ICP before any budget went into sending
Rewrote the sequences to match what's working in cold email right now.
- Fixed structural issues in the original emails: they were signed by "the team" instead of a real person, asked for a 30-minute demo on the first touch, and carried heavy footers. I replaced this with person-signed emails and a low-commitment 15-minute ask
- Wrote a five-campaign launch using current benchmark data: under 80 words per email, 4–7 touches per sequence, and most of the effort on email one, since that's where most replies come from
- Delivered the complete launch spec: audiences, campaign structures, and a reply-handling agent configuration with a response table for every reply type
- Designed a new offer (a "next-hire practice pack") and a founder-ready approval document for the US outbound push
Checked that the product matched the pitch, and flagged legal risk before any email went out.
- Re-audited the live product against the outbound promise and found it narrower than "runs your HR team". I flagged the gap before it could reach prospects and hurt trust
- Compliance research on AI hiring tools: NYC Local Law 144 (annual independent bias audits, published results, 10 business days' candidate notice) and similar rules in Illinois and Colorado, which affect who can safely be pitched a hiring-screening product
- Reviewed a proposed review-incentive offer against the FTC rule on consumer reviews and testimonials
Grounded everything in real data, and reported outcomes rather than activity.
- Full-site SEO & CRO audit of contentbuilder.ai, plus a live Google Search Console audit using real indexation and query data instead of third-party estimates
- 20-day research & execution plan with a working tracker and a one-page summary for the founder
- Built a demo course in the product and scripted a 90-second product walkthrough for sales use
- Set up outcome-based reporting: delivered, bounce %, inbox placement, replies, positive replies, calls booked, calls held, clients won, MRR added. Daily updates followed a fixed format: decision, why, numbers, lessons, tomorrow
