Answer inbound leads in 60 seconds without hiring anyone
Nothing in a small business has a shorter path from automation to revenue than response time, and the fix is a twenty-minute build anyone can copy.
Harvard Business Review audited the response times of 2,241 US companies in 2011 and found something that has never really been contradicted since. Firms that managed to contact a prospect within an hour were nearly seven times as likely to qualify that lead as firms that waited even one hour longer, and sixty times more likely than the ones that took a day or more.
The study is old enough to have a beard, and the mechanism behind it hasn't changed: the person who filled in your form is at their desk thinking about the problem right now, and in four hours they won't be.
This is a queue, which is why it's worth automating
Most automation advice would have you start with whatever feels most tedious. The better test is where work piles up, and inbound leads pile up in exactly one place: the inbox of an owner or a single salesperson who is in a meeting until three. The delay has nothing to do with how hard the reply is to write; it comes from sitting in a queue behind one person.
That makes speed to lead unusually well suited to a first build. The constraint is obvious, the fix is cheap, and the payback shows up in the same month rather than two quarters later. It also passes the test Nick Saraev applies to automations worth building: a tangible deliverable, a real revenue pain, and something templatable rather than bespoke.
Twenty minutes in Make.com or n8n
Either tool does this. Assume twenty minutes if your form and your CRM already talk to each other.
Twenty minutes to build in Make.com or n8n. A person still sends the reply; the drafting and the triage are done before they open Slack.
- Trigger. Your web form submits, or an email lands in the shared inbox. Both are standard modules in Make and n8n.
- Score. Claude reads the enquiry and scores it against your ideal customer profile, returning structured JSON rather than prose.
- Route. The scenario posts to Slack: who they are, what they asked for, the score with its reasoning, and a suggested first reply written from what they actually wrote.
- Log. The lead and its score are written to HubSpot so the scoring is visible next to the deal rather than trapped in Slack.
- Escalate. Anything above your threshold pings a person directly. Everything else waits in the normal queue.
The Slack step is the one that earns its place. A human still sends the reply, so nothing goes out in your name that you haven't read, and the drafting work that used to make replying feel like a task is already done.
Scoring the enquiry before a human sees it
You are scoring an inbound enquiry for a business that sells [what] to [who]. Our ideal customer looks like: [3 to 5 concrete characteristics, including size, sector and the trigger event that usually precedes a purchase].
Here is the enquiry:
[paste the message, the form fields, and the page it came from]
Return only JSON in this shape:
{
"fit": 1-5,
"urgency": 1-5,
"budget_signal": 1-5,
"reasoning": "two sentences",
"suggested_reply": "under 120 words, referencing what they actually wrote",
"missing_info": ["the questions we still need answered"]
}
Score fit on how closely they match the ideal customer, urgency on evidence of a deadline or trigger event, and budget signal on any indication they can afford us. Do not infer anything that is not present in the message. If a dimension cannot be judged from what they wrote, score it 1 and say so in reasoning.
That last instruction is doing real work. Without it, a model will happily award a four out of five for budget signal to someone who mentioned nothing about money at all.
What five percentage points is actually worth
Run the arithmetic for your own numbers rather than trusting a vendor's case study. Here it is for a business with 200 inbound leads a month, a 12% conversion rate and a $4,000 average deal.
Five percentage points is the assumption to interrogate, because that is where the honest and the promotional numbers part company. The lift figures you see quoted in the 25% to 50% range almost all come from lead-generation vendors reporting their own results, so treat them as what's possible for somebody, not as what you should plan around. Anchor on the HBR audit, and treat the vendor numbers as illustration rather than a plan.
Auto-replies that read like auto-replies
This is the one way the build backfires. Speed only helps if the response is worth reading, and a templated "thanks for your enquiry, someone will be in touch" arriving in nine seconds tells the prospect precisely how much attention they can expect later.
This is the reason to put a model in the loop rather than a Zapier template with merge fields. The reply can reference the specific thing the person described, in their words, which is the difference between a receipt and a conversation.
Median time to first human response
Track the median rather than the mean, because a single three-day outlier will quietly flatter the average. And measure the time until a person replies, since the automated acknowledgement tells you nothing about whether anyone engaged.
Track it weekly and put it somewhere the sales team can see. If you can't report that number, you can't claim a return on this build, and in ninety days you'll be arguing about whether it worked from memory.