The agent that found us our biggest customer

We're going to tell a specific story. The details are real. We've anonymized the customer.

About 4 months ago, our largest customer to date came through a channel we didn't expect: one of our own AI agents flagged them as a high-value inbound lead. The agent's classification triggered a human follow-up within hours. That follow-up turned into a sales conversation, which turned into a six-figure annual contract.

Here's how the agent worked, why it caught what it did, and why this matters beyond a single story.

The setup

Our website has a contact form. People filling it out range widely — students with questions, journalists writing pieces, developers curious about implementation, prospects evaluating purchase, and occasionally enterprise buyers ready to engage.

The pre-agent process for this form was: forms came into an inbox, someone (usually me) triaged them once a day, replied to the obvious sales prospects, ignored the rest.

The problem: by the time we triaged, hours had passed since the form submission. For enterprise buyers comparing 3-4 vendors, the first vendor to respond often won the conversation. Hours of delay cost us deals we never realized we were in.

The agent

We built an inbound lead qualifier as a custom Avery NXR agent. Pretty standard pattern. The agent:

→ Watches the contact form inbox → Reads each submission → Researches the company (via their email domain, LinkedIn data, website content) → Scores the lead against criteria (company size, industry fit, role of the submitter, language indicating active evaluation vs. exploration) → Sends a Slack notification to me with the score + reasoning + research summary → Drafts a personalized response email for review

The agent runs within minutes of form submission. By the time I check Slack, I have a structured summary instead of a raw form.

The specific submission

The submission that became our largest customer looked, on the surface, unremarkable. A form from someone with a generic-sounding name, a corporate email domain, asking general questions about our platform.

Pre-agent, this might have triaged as "casual inquiry, respond when I get to it tomorrow."

The agent's research changed the picture:

→ Email domain belonged to a 5,000-employee company in a regulated industry → The submitter's LinkedIn (auto-researched) showed VP-level title in operations → The questions in the form referenced specific concepts (data residency, audit transparency) that suggested they'd already done research → The company's recent news mentioned a $1.2B+ contract they were preparing infrastructure for

The agent's scoring: 9.4/10 — top of the priority queue. The Slack notification said "respond within 2 hours, this looks like a serious enterprise eval."

What I did with that

I dropped what I was doing. Read the agent's research summary. Wrote a personalized response (using the agent's draft as a starting point but rewriting significantly). Hit send within 90 minutes of their form submission.

The buyer's response (paraphrased): "We've sent the same inquiry to 3 vendors. You're the only one who responded today. Also you addressed the specific issues we asked about, which suggests you actually read the form."

That single thread became a sales conversation. The sales conversation became a multi-stakeholder enterprise eval. The enterprise eval became a contract.

What the agent enabled

This deal would not have happened without the agent. Or at least, it would not have happened the same way.

Without the agent: → Form would have sat in inbox for 18-24 hours → I would have triaged it as "interesting, follow up sometime" → Response would have been generic, sent 1-2 days later → By the time I responded, a competitor would have already booked a call → We probably wouldn't have made it to evaluation

With the agent: → Form was researched within minutes → The HIGH-priority signal got my attention in 2 hours → Response was specific to what they asked → We were the only vendor who showed up that day → That positioning carried through the entire eval

The agent didn't sell anything. It made sure I knew which inbound to prioritize. The selling was still human work. But the agent made sure I did the human work on the right thing at the right time.

The math

We've since added up what this agent has done over 6 months:

→ Total form submissions processed: ~840 → Submissions flagged as high-priority: ~45 → Of high-priority, took the call: ~28 → Became active sales conversations: ~17 → Closed business: 8 (including the big one)

Across those 8 deals, total contract value: ~$340K (with the largest being the six-figure one mentioned above).

Without the agent, conservatively half of those 8 would have been lost to slow response or wrong triage. Conservatively $170K in deals saved.

The agent's cost: a few hours to build + $29/month Avery NXR Pro. Maintenance: ~1 hour/month tuning.

ROI: not even close. It's the highest-leverage agent we've built.

Why this generalizes

We're sharing this story because the pattern repeats:

→ Inbound signals come in randomly → Most are low-value, but a few are high-value → Speed of response correlates strongly with conversion → Without help, high-value signals get missed in the noise → Agents can filter the noise so humans focus on signals

This isn't unique to sales inbounds. The pattern applies to:

→ Support tickets (high-urgency cases in the noise of low-urgency) → Hiring (strong candidates in the noise of mismatches) → Customer signals (churn risk in the noise of normal usage) → Investor inbounds (real interest in the noise of cold outreach)

In each case: a triage agent that handles the high-volume filtering work makes the difference between "missing the important signal" and "responding fast enough to win."

The bigger lesson

When you deploy an agent, the obvious benefit is "saves time on routine work." The non-obvious benefit is "raises the ceiling on what's possible."

The routine work was happening before the agent (manual triage). The agent absorbed it.

But the agent also enabled a behavior that wasn't possible before — responding to high-value inbound within 90 minutes of receipt, with personalized content. That capability didn't exist for us pre-agent. It does now.

The deals that capability unlocks are the hidden upside. They don't show up in time-saved calculations. They show up in revenue we wouldn't otherwise have.

What this means for buyers

If you have any kind of high-volume inbound signal (contact forms, support tickets, leads, applications), there's almost certainly an agent that would change your response capability.

The build effort is small (hours to days). The maintenance is small (1-2 hours/month). The upside can be career-defining.

Pick the inbound signal that, if you responded to it 10x faster, would change your business. Build an agent for it. See what happens.

For us, that signal was inbound contact forms. For your business, it might be different. The pattern is the same.

→ avery.software — Free Desktop tier. Build the agent that catches the inbound you'd otherwise miss.