How I Built a Website With an AI Agent That Captures Leads 24/7

Most company websites are brochures with a contact form. AM Shipping & Logistics needed theirs to work like a front desk: freight visitors arrive with concrete questions, often outside business hours, and a team of real people cannot staff a chat window at 2 a.m. The goal was simple to say and hard to build — a site that responds, qualifies, and captures while the team sleeps.

This is the build story of the live site (amslpk.com): what shipped, how the agent actually works, what I deliberately did not let it do, and what I would repeat on the next one.

The problem, stated plainly

A freight buyer’s questions are specific: do you move this cargo, what lane, pickup from where, can it start this week. Marketing pages answer none of that at the moment of asking. The previous pattern — read the site, find the form, wait for business hours — meant every hot question went cold overnight. The website was fine as a brochure. As a sales surface, it was asleep.

What we shipped

• A high-performance corporate website on Next.js — fast pages, clean architecture, and content the team can change without calling a developer.
• An embedded AI agent that greets visitors and answers inquiries in real time — grounded in the company’s actual services and routes, not generic freight copy.
• Automated lead capture: when a conversation turns commercial — a quote request, a lane, a volume — the agent collects contact details and hands them over instead of letting interest die in a chat log.
• CRM syncing, so captured leads land where the team already works instead of in yet another dashboard nobody opens.
• Freight booking operations streamlined around the clock: the agent works the front line at 24/7, and humans take over exactly where judgment or money commitments start.

The conversation design (where projects like this are won or lost)

The guiding principle: the agent is a front desk, not a freelance salesperson. Welcome, understand the shipment — type, lane, timeline — answer from approved material, and only then ask for contact details, and only when the conversation is actually commercial. Before the first line was written we defined what it must never do: invent rates, invent transit times, promise capacity. On those, its honest answer is “the team will confirm” — a sentence that protects the brand better than confident fiction.

The win was not the model. It was deciding what the agent is allowed to be wrong about — and keeping the answer to: almost nothing that touches money.

What the client says

“Ali built our company website and integrated an AI agent that welcomes visitors, handles customer inquiries in real-time, and captures leads 24/7. The results have been amazing — our operations are faster and far more efficient. Truly a standout developer.” — Intezar Rana, CEO, AM Shipping & Logistics

The honest numbers problem

I will not put a fake percentage in this case study — nobody can honestly claim a conversion lift without clean before-and-after data, and most small sites do not have it. What I can say is structural: leads that used to arrive as bare forms the next morning now start as conversations where the context already exists — cargo, lane, timeline, urgency. The team begins mid-sentence instead of playing Twenty Questions. First response went from hours to instant, because that is a property of the system, not a statistic.

What I would repeat on the next build

1. Scope the agent’s one job — front desk — before choosing models or frameworks. The role is the spec.
2. Feed it the real material first — services, routes, policies — and polish prompts after it stops making things up.
3. Ship with review in the loop: read transcripts daily for the first weeks, fix prompts weekly. Launch is the start of the work.
4. Start capture-first: the agent gathers and qualifies; humans commit. Auto-execution comes later, per action, once trust is earned with data.
5. Keep the fallback boring: if the agent is unsure, promise a human callback. Never improvise a deadline.

Want the same front desk on your site?

If your website gets visitors but your team gets the leads late, this pattern fits: a corporate site plus a front-desk agent with strict guardrails. Let’s talk and we will map your five highest-value questions first — before any model gets picked.