How to Build an AI Agent That Runs Your Weekly Pipeline Review

An AI pipeline review agent pulls every rep's open opportunities, activity and quota data from your CRM on a schedule, compares them against targets for coverage, pacing and % to plan, and sends managers and reps a written summary of who's behind, why, and what to do about it. Built with your CRM, an automation tool like n8n and an LLM, it replaces the manual weekly pipeline review with a consistent daily one.

Why manual pipeline reviews break down

Most sales teams run pipeline reviews the same way: a manager exports opportunities, scans for stale deals, asks each rep for updates, and builds a picture of the quarter from memory and gut feel. It works with three reps. It falls apart at ten.

The symptoms are familiar:

• Problems surface too late. A rep who's 40% behind pace in week three looks fine until week ten.
• Reviews are inconsistent. Each manager checks different things, so two reps with the same problem get different coaching.
• The CRM fills with dead deals. Opportunities with no activity in 60 days stay open, inflate pipeline and make the forecast unreliable.
• Reps don't know what to prioritize. They work the deals they like, not the ones most likely to close or most at risk.
• Managers spend hours on data prep instead of coaching.

None of this is a people problem. It's a systems problem, and it's very automatable.

What the pipeline review agent does

The agent runs on a schedule (daily for reps, weekly for managers) and works through five checks.

1. Pipeline coverage

For each rep, the agent sums open pipeline for the current period and divides it by remaining quota. If your team needs 3x coverage and a rep is at 1.8x, they're flagged. It also breaks coverage down by stage, because 3x coverage that's all in discovery is not the same as 3x coverage in negotiation.

2. Pacing and % to plan

The agent compares closed-won revenue to date against where the rep should be at this point in the period. A rep at 30% of quota halfway through the quarter is behind pace even if their total pipeline looks healthy. It reports both % to plan and the gap in dollars.

3. Activity inputs and KPIs

Outcomes lag; inputs lead. The agent pulls activity data (calls, emails, meetings booked, opportunities created) and compares each rep to their own targets and to team averages. A drop in meetings booked this month predicts a pipeline problem next month.

4. Inbound vs outbound mix

The agent calculates the share of each rep's pipeline that came from inbound vs outbound sources, and the same for the whole team. This answers questions managers usually can't: Is the team over-reliant on marketing leads? Which reps are building their own pipeline? Is outbound pipeline converting at a lower rate?

5. Dead and stalled opportunities

Any opportunity with no logged activity in a set window (commonly 30 to 60 days, depending on your sales cycle) is flagged. The agent can either notify the owner with a deadline to update it, or automatically move it to closed-lost with a reason code after a grace period. Either way, your pipeline reflects reality.

How the agent turns data into actions

The checks are just math. The useful part is what the LLM does with the results.

For each rep, the agent receives the numbers above plus the details of their open deals (stage, amount, close date, last activity, next step) and writes a short, specific brief:

• Where they stand against plan, in one sentence
• The two or three deals that matter most this week and why
• A concrete next step for each one
• Any deals that should be closed out or have their close date moved

For managers, it writes a team summary: who's on track, who's at risk, the main reason for each risk (low coverage, low activity, deals stuck in one stage), and suggested coaching topics for one-on-ones.

This is the shift that matters. Instead of a dashboard someone has to interpret, every rep starts the day knowing what to do.

How to build it

Data sources

• CRM (HubSpot or Salesforce): opportunities, stages, amounts, close dates, owners, lead source, last activity date
• Quota data: stored in the CRM as a custom object or property, or in a Google Sheet
• Activity data: logged calls, emails and meetings from the CRM or your sales engagement tool

Fields to get right first

The agent is only as good as the data. Before building, make sure you have:

• A required lead source field with clean values, so inbound vs outbound can be calculated
• Reliable close dates and stages, with stage definitions everyone agrees on
• A last activity date that updates automatically
• Quota by rep, by period

The workflow

1. A scheduled trigger in n8n runs each weekday morning.
2. n8n pulls open opportunities, closed-won deals for the period, activity counts and quotas from the CRM.
3. A code step calculates coverage, pacing, % to plan, activity vs target and inbound/outbound ratio for each rep.
4. A second step identifies opportunities with no activity past your threshold.
5. The metrics and deal details are sent to an LLM with a prompt that defines your targets, your sales process and the format of the brief.
6. The rep brief is sent by Slack or email to each rep; the team summary goes to the manager.
7. Stale opportunities are tagged in the CRM, and an owner task is created, or they're auto-closed after the grace period.

Guardrails

• Keep the LLM on interpretation, not calculation. Do the math in code and give the model the results.
• Log every brief so managers can see what the agent recommended.
• Start with notifications only. Turn on auto-closing of dead deals once the team trusts the rules.

What it looks like day to day

A rep opens Slack at 8:30 and sees: they're at 62% of plan with five weeks left, pipeline coverage is 2.1x against a 3x target, and the three deals to focus on today, each with a suggested next step. Two opportunities are flagged as stale with a deadline to update them.

The manager gets a Monday summary: four reps on track, two behind on coverage, one behind on activity, and the deals most at risk of slipping this quarter.

Metrics to track after launch

• Forecast accuracy (forecast vs actual, by month)
• Pipeline coverage by rep, over time
• Percentage of open opportunities with activity in the last 30 days
• Time managers spend on pipeline prep each week
• Win rate and sales cycle length, before and after