It’s Monday morning. The deal at the top of the forecast hasn’t heard from the buyer in two weeks. The rep isn’t worried. “They’re on vacation. I’ve got a follow-up scheduled for when they’re back.” Everyone nods, and the meeting moves on.

Three weeks later the deal is gone. Nobody lied. The rep really did have a task scheduled. But two weeks of silence was the deal losing momentum, and the pipeline review was the one place where somebody could have caught it.

That’s what a sales pipeline review is for. It’s the weekly meeting where a sales leader and reps go through the open deals and decide three things: which deals will actually close, where we’re dropping the ball, and whether we’ll hit the number. A good one ends with actions on specific deals. A bad one ends with updated numbers and the same deals sliding.

We asked 153 sales leaders where they run that review today. 61% run it straight from the CRM. 15% already use Claude or ChatGPT. 13% use docs or spreadsheets, and 7% use a tool like Gong or Clari.

We’ve run ours all of those ways. Here’s what we learned, how to run a better one with Claude, and what Claude can’t do for you.

Why most pipeline reviews don’t work

We’ve never had a perfect pipeline review. I don’t think anyone has. They tend to break in one of two ways.

The manager runs it. The rep sits under a magnifying glass and has to remember every detail about every deal, which realistically nobody can. It turns into an interrogation. Reps dread it, and they prepare for it by making their deals sound good.

The rep runs it. Now you’re hearing each deal through the person who most wants it to close. We love our reps. They work hard and they do their best. They’re also unreliable narrators. Not because they’re lying, but because they’re excited. They hear what they want to hear, and every deal comes out sounding a little better than it is.

People tell us AI can’t be trusted for forecasting. We’d argue humans can’t either. The answer isn’t to pick one. It’s to take the rep’s read and match it against what was actually said on the calls and in email, minus the emotion. That’s an objective forecast, and it’s what a pipeline review should be built on.

How our own pipeline review changed

Before: the CRM. We use HubSpot. Every week we’d pull up the deals set to close this month, the ones progressing, and what was happening in the later stages. It worked, sort of. The CRM only knows what someone typed into it, and the person typing is the rep who wants the deal to close. We were reviewing the optimistic version of the pipeline and calling it data.

This year: Claude. We connected Claude to HubSpot, our email and our call notes, and started running the review from an artifact that showed which deals were moving forward. It was better. It felt better. For the first time, the review was reading what buyers actually said, not just what reps logged.

It also never felt complete. The worst moments were when a rep looked at the review and said, “I don’t believe that’s true about my deal, and I don’t understand where this is pulling the data from.” Sometimes the rep was right. Usually they weren’t. But once a team stops trusting the review, it doesn’t matter how good it looks.

And there was a bigger problem we didn’t see at first. The review got better. The deals didn’t. We were spotting the right problems and then losing them the minute the meeting ended.

Now: Arrows. Today we run the review in Arrows. Every deal is in one place, we can filter it any way we need, and every deal has a momentum read built from the calls, the emails and the CRM: moving, drifting or stalled. And when we decide something in the meeting, it turns into an action on that deal, not a note in someone’s head. More on that below, because it’s the part that changed our results.

What Claude can’t tell you

Claude is a genuinely good place to start a pipeline review. It’s also worth knowing exactly where it stops, so you don’t mistake a better report for a better pipeline.

  1. It only knows what it can see. Connect only the CRM and Claude will repeat the CRM’s mistakes: an email nobody answered looks like activity, and a call where the buyer went cold looks like a meeting held. Connect email and your call recorder too.
  2. It won’t settle the “where did this come from?” argument unless you ask. Ask Claude to cite the email, call or CRM field behind every flag. Otherwise the review becomes your word against the rep’s.
  3. It doesn’t remember last week. Each chat starts fresh unless you set it up otherwise. It doesn’t know what you agreed on Monday, so it can’t tell you whether it happened.
  4. Every rep gets a different answer. If each rep runs their own prompts, you get five versions of the pipeline with five definitions of “at risk.” A team needs one definition, applied the same way to every deal.
  5. It can’t make anything happen after the meeting. This is the big one, and it’s where most pipeline reviews quietly fail.

Read every deal by momentum

The single most useful lens we’ve found is momentum. Forget the stage for a minute. Every deal is doing one of three things.

  • Moving. Great. You may not need to talk about these at all. Or look for the ones you could pull into this month.
  • Drifting. The deal had momentum recently and is about to lose it. The buyer has gone quiet for longer than usual, their side’s next steps are slipping, or something on a call signaled hesitation. These are the deals that are just about to slide off, and you can still catch them. Ask: what do we need to do this week to get it back on track?
  • Stalled. Momentum is gone. These are much harder to restart. Ask: do we close it out, or is there a Hail Mary worth trying?

Spend most of the meeting on drifting deals. That’s where your time changes outcomes. Moving deals don’t need you, and stalled deals rarely come back.

When the rep says “that’s not true”

It will happen. Two examples we see constantly:

“They’ve been on vacation for two weeks.” The review says the deal is slipping. The rep says the buyer is out and there’s a follow-up task for when they’re back. That may all be true. But two weeks of silence usually does mean lost momentum. It isn’t the rep’s fault. It’s still real, and it’s still worth a plan.

“I’m talking to the decision maker.” The review says the economic buyer has never been on a call. The rep says it’s handled: their contact has a direct line to the leader. A month later you find out the leader was looped in but wasn’t ready to decide, and you wish you’d sold to them directly. (That gap is exactly what a sales methodology is meant to catch. We wrote about what a methodology is really for, and why 43% of teams don’t use one.)

The rule we follow: the flags have to be accurate, and there will be nuance around them. Flag it anyway, and let the rep explain the nuance in the meeting. In our experience, more often than not the objective read turns out to be what happened. Sometimes a rep’s instinct catches something the data can’t see. That’s exactly why you look at both.

Pipeline review questions to ask

Whether you ask them in the meeting or ask Claude before it, these are the questions worth asking every week. Each one is about evidence, not feelings.

  1. “Which open deals have had no two-way contact in 14 days?” A reply, a call, a meeting. Not an email that went unanswered.
  2. “Which deals have no dated next step, on our side or the buyer’s?” A booked meeting counts. “They’ll decide by the 12th” counts. “Following up” doesn’t.
  3. “Where did someone promise something by a date that has passed?” Name who promised what. A broken promise is the earliest warning sign there is.
  4. “Which late-stage deals have never had the decision maker on a call?”
  5. “Which close dates have already passed, or don’t match what the buyer said?”
  6. “What will be done on each drifting deal this week, and by whom?”

If you use Claude, add one more line to every one of those: “Show me the email, call or field behind each answer.”

The fastest way to get all of this in one place is the free Arrows Sales Skills connector for Claude. Type “Run my weekly pipeline review” and you get a one-page report: what’s moving forward, what’s at risk and why, what to do this week, a rollup by rep for leaders, and the pipeline by stage. Think of it as a pipeline review template that fills itself in.

A simple pipeline review agenda

Run it at least weekly, in two versions.

Monday team standup (15 minutes). What is everyone committing to this week? Go rep by rep, drifting deals only. Each one leaves with a named action and a day.

Weekly 1:1 with each rep (30 minutes). What are you committing to this month and beyond? Walk the rep’s pipeline by momentum, check last week’s actions, and agree on this week’s.

Keep both short. The review isn’t the work. It’s how you decide what the work is.

The one thing to change: what happens after the meeting

Here’s the part almost nobody gets right.

You have a great review. You agree on what to do for every drifting deal. Then the rep goes back to their desk, opens their email, grabs a coffee, eats lunch, and moves on to the next deal they have to work on. Not because they don’t care. Because that’s what a rep’s day looks like. By Tuesday, everything you discussed is gone.

So ask yourself honestly: after your last pipeline review, how do you know the actions actually happened?

This is the gap a report can’t close, from Claude or anywhere else. It’s the reason we built Arrows.

We pull up the pipeline review inside Arrows and go deal by deal. When we see one drifting, we don’t write a note. We open the deal and just talk. Arrows has voice-to-text built in, so the leader and the rep say out loud what we’re doing and what needs to happen on this deal:

  • “Loop me in on this one. I know their founder. CC me on the next email.”
  • “Get the decision maker on the next call. Ask for it directly.”
  • “Multi-thread: reach out to their VP on LinkedIn.”

Arrows turns that into actions on the deal’s gameplan, with the context already filled in: who to email, what was said on the last call, what the next step was. After the meeting, the rep doesn’t have to remember anything. They open Arrows, review the drafts and send them. And the next Monday, the leader can see whether it got done, and whether the deal moved from drifting back to moving.

That’s the loop: review, action, follow-through, and next week’s review shows whether it worked. Most teams only have the first step.

Try it this week

  • Run one review with Claude. Install the free Arrows Sales Skills and type “Run my weekly pipeline review.”
  • Sort your deals by momentum and spend the meeting on the drifting ones.
  • End with actions, not notes. Every drifting deal leaves the meeting with a named action, an owner and a day.
  • If you want those actions to actually happen, see how Arrows does it.