Thanks, Michael. This is actually a much better starting point than it might feel.
Your top-of-funnel numbers are not terrible. A 20% acceptance rate and 2.5% meeting rate shows you can generate interest. The bigger issue is that only 10% of those meetings are qualified.
My guess is that the curiosity-led messaging is doing exactly what it is supposed to do. It gets people interested enough to chat, but not necessarily because they have an urgent problem or any intention to buy.
Before scaling email again, I would focus on a few things.
First, review the last 20 to 30 meetings and write down exactly why each person was or was not qualified. You will probably start to see patterns around whether the agent is actually in production, the volume of usage, the size of the engineering team, who owns the problem and how painful it is.
Second, I would narrow “enterprises building AI agents” quite significantly. For example, companies with customer-facing agents already in production, enough usage for behavioural issues to matter and an engineering leader who is accountable for reliability or performance.
Third, I would move away from asking for general feedback and towards testing a clear problem hypothesis.
Something like:
“We are speaking to engineering teams running AI agents in production who can see individual traces, but still struggle to understand recurring behavioural patterns across thousands of interactions. Is that something your team has experienced, or do you already have it covered?”
I would definitely test email again, but I would not jump straight into another large automated campaign. I would start with 50 to 100 carefully selected accounts, identify two or three relevant people within each and test a few different problem hypotheses.
I would also judge it on qualified conversations and confirmed problems, not just opens, replies and meetings.
The fact that you are still in the design-partner phase is also useful. You do not need to pretend the offer is fully standardised. You can be direct that you are looking for a small number of teams with the right environment and problem to help shape the product.
Happy to work through this publicly. Share one of the messages you are currently using, plus the most common reasons people turn out not to be qualified, and we can pull it apart.
