The Lead Generation Tool I Actually Want Is Not an AI SDR
Useful AI lead generation should surface evidence, context, and human-reviewed opportunities, not blast cold messages at scale.
I have been building a small internal lead generation tool inside Hermes Agent.
The goal is simple: find potential opportunities, surface the evidence, score the fit, and keep the final decision human-reviewed.
No spam automation. No mass outreach. No “AI SDR blasting 10,000 people” nonsense.
That whole category gives me hives.
The useful part is not the sending
Most AI sales tooling seems obsessed with volume. More contacts, more emails, more sequences, more “personalisation” that somehow still reads like it was assembled in a panic by a spreadsheet.
That is not the bit I care about.
The useful bit is qualification.
Can the system notice a company that might actually have a problem worth solving? Can it show me why? Can it separate weak signals from genuine pain? Can it avoid wasting my time without pretending it should make the final call?
What the system looks for
The current version looks at signals like AI automation need, workflow or system pain, manual operational processes, scale, funding, founder or operator fit, and evidence from public sources.
Each lead gets scored, summarised, and queued for review.
From there, the tool suggests an angle, highlights the pain signals, stores source URLs, and drafts a possible comment or outreach note.
The important bit is that it does not decide for me.
It gives me enough context to decide whether something is worth researching further.
Relationship first, automation second
The version I want feels more like: “Here are some relevant opportunities you may want to look at.”
Not: “Here are 500 cold emails we fired into the void.”
That distinction matters because outreach is not just a mechanics problem. It is a trust problem.
If the system pushes me toward bad-fit prospects, robotic messages, or fake urgency, it is not helping. It is just making me faster at being annoying.
AI can be useful in prospecting, but only if it keeps the human judgement where it belongs.
Evidence first. Context second. Suggested action third. Human decision before anything public.
That is the workflow I actually trust.