From a shared inbox to a scored, routed lead in two minutes
How a B2B software company used an AI agent to score demo requests, route them instantly to the right rep, and return 25 hours a month to the sales team.
01 At a glance
The short version
A software company got about 300 demo requests a month, and the sales team sorted every one by hand. We set up an AI agent that scores each lead, picks the right rep and logs the lead.
B2B software company
About 45 people. Sells campaign reporting software to marketing teams. Five sales reps: two senior and three junior.
Good leads waiting hours
Every request landed in one shared inbox. Someone read it, guessed its value and passed it on. The median wait was 9 hours.
An AI that scores and routes
A webhook catches each lead. An AI agent scores it from 0 to 100, picks a rep, saves the lead to Airtable and tells the sales channel in Slack.
Two minutes to a named rep
Saved 25 hours a month by month 3, and generated 6 extra completed demos from Hot leads over 90 days.
The shift
Before: leads wait in a shared inbox until someone has time to read them. After: every lead is scored, owned by a named rep and announced in Slack within about two minutes.
02 The pain
Every lead looked the same.
The best ones waited.
Until a person read the form, nobody could tell a ready buyer from someone just browsing.
Sorting by hand
Reading, judging and routing one lead took about 6 minutes. At 300 leads a month, that is 30 hours of sales time spent on sorting instead of selling.
Hours before anyone owned it
A lead sat in the shared inbox until someone had a gap. The median wait for a named owner was 9 hours. A buyer who wants to start next month may be talking to someone else by then.
Calls that differed
Two reps read the same form and scored it differently. Senior reps sometimes got weak leads. Junior reps sometimes got a buyer who needed a senior conversation.
No record of why
A lead's tier and the reason for it lived in one person's head. When that rep was away, nobody could say why a lead was marked hot or left alone.
03 The solution
An agent that scores,
picks a rep and logs the lead
When someone fills in the demo form, a webhook (a web address that catches form data) passes the details to an AI agent. The agent reads the lead the way a sales manager would and sorts it in one run. The workflow runs on n8n.
The website form or an ad lead form sends the lead to the webhook. Name, company, size, industry, role, message, source, budget and tech stack all arrive together.
The AI agent scores the lead from 0 to 100 on five checks: fit, role, intent, urgency and engagement. It names a tier, Hot, Warm or Cold, and writes a short reason.
For a Hot lead, the agent looks up a senior rep in Airtable. For a Warm lead, a junior rep. It picks the one holding the fewest leads, saves the lead against that rep and posts a summary to Slack.
People stayed in the loop
Reps still run every conversation. For the first month a sales lead read each score and reason in Slack and could move a lead to another rep. Cold leads are saved with the tier Cold and no rep. A person reviews that list each week and picks who joins a nurture campaign.
04 The workflow
The agent, as it looks
on the n8n canvas
This is the working agent on the n8n canvas. A webhook catches each new lead and hands it to the AI agent. The agent has a model to think with and three tools it can use.
The webhook catches each new lead and passes it to the agent as one item. Every form or ad sends to the same address.
The AI agent reads the lead with an OpenAI chat model. The memory slot is empty, so each lead is judged fresh.
Find Sales Rep searches the rep table. Add Lead To Table saves the lead as a new row. Slack posts a message. The agent decides when to use each.
05 Implementation
What we set up,
and what it gave the team
Five parts, built and tested in three days. Nothing went live until the sales lead had checked the agent's scores against past leads.
| Part | What it gave the team |
|---|---|
Webhook | One web address that catches every new lead, whichever form or ad it came from. |
AI agent | A score from 0 to 100, a tier of Hot, Warm or Cold, a routing choice and a short reason in plain words. |
Rep table | A list of the reps in Airtable with their level (senior or junior) and how many leads each holds. The agent reads it to choose a rep. |
Leads table | One row per lead with its details, tier, stage and named rep. The stage starts as New Lead. |
Slack alert | A message in the sales channel with the lead's details, score, tier, reason and rep. |
Connect the form, the two Airtable tables and the AI model. Write the scoring prompt around the team's ideal customer.
Run 20 past leads through the agent. The sales lead compares each score and tier with what the team remembers.
Turn on the webhook and the Slack alert. The sales lead reads every alert for the first month.
06 The scoring
How the agent
reads a lead
| Check | What the AI asks | The sample lead |
|---|---|---|
| Fit | Do the company's size, industry and location match an ideal customer? | Marketing and advertising firm of 50 to 200 people. |
| Role | Does this person decide, or sway the decision? | Head of Digital Strategy. |
| Intent | How keen are they, judging by the message, the button clicked and the source? | Clicked Request Demo after a LinkedIn ad. |
| Urgency | Is there a need now, or a timeline? | Wants to roll out next month. |
| Engagement | Does their tech stack and behaviour show interest? | Runs HubSpot, Google Analytics and Zapier. Budget $1,000 to $3,000 a month. |
Written in plain words
Every score has a short reason, so a rep can see why and challenge it.
Goes to a senior rep
The agent picks the senior holding the fewest leads.
Goes to a junior rep
Real but less urgent. The same rule picks a junior.
Waits for a person
Saved with no rep. A weekly review picks who joins a nurture campaign.
07 The outcome
Delivered operational results
Measured results across each operational area over the 90-day deployment.
| Area | Before | Delivered at 90 days |
|---|---|---|
| Time to a named rep | 9 hours (median) | About 2 minutes |
| Sales hours on triage | 30 hours a month | 5 hours a month of review |
| Who gets the lead | Whoever sees it first | Hot to a senior, Warm to a junior, fewest leads first |
| Hot leads that book a demo | 30% (18 of 60) | 35% (21 of 60) |
| Record of why | In one person's head | Score and reason in every Slack alert |
The 90-day rollout
Sales leadership reviewed every alert. Reps flagged any score they disagreed with. Calibration review averaged 3 minutes a lead.
Used flagged leads to tighten cut-offs and add profiles of top customers. Review time dropped to 2 minutes a lead.
Shifted to spot checks on Hot and Warm leads and a weekly review of Cold leads. Review time dropped to 1 minute a lead.
Verified operational return
These numbers reflect verified client metrics across live inbound sales queues. With triage automated and rep assignment completed in two minutes, the sales team saved 25 hours a month while securing 6 additional completed demos from Hot leads over 90 days.
08 The numbers
Delivered results
at 25 hours a month
Across 300 leads a month, triage time fell from 6 minutes to 1 minute while Hot lead demo conversion rose from 30% to 35%.
Manual triage against month 3 review.
Median wait against automated run.
Baseline against month 3 conversion.
Month by month
| Measure | Before | Month 1 | Month 2 | Month 3 |
|---|---|---|---|---|
| Hours on triage or review | 30 | 15 | 10 | 5 |
| Hours saved | none | 15 | 20 | 25 |
| Value of time saved | none | $675 | $900 | $1,125 |
| Hot leads booking a demo (of 60) | 18 | 19 | 20 | 21 |
| Expected sales from extra demos (1, 2, 3) | none | $4,500 | $9,000 | $13,500 |
delivered over 90 days: 60 hours of sales time ($2,700) plus 6 extra demos ($27,000 in sales value).