Most founders already have the data needed for a first real diagnosis of their sales team sitting in a CRM export or a spreadsheet nobody's looked at properly in months. You don't need a data analyst or a dashboard build to get a useful first read on it. A few minutes with Claude and a well-built prompt gets you most of the way there.

What you need first

Export your pipeline data from whatever CRM or spreadsheet you use. At minimum you want deal stage, deal value, the date each deal entered its current stage, and who owns each deal. If pulling this together cleanly is already difficult, that's useful information on its own, it usually means the same disorganisation is showing up in how the team actually works, not just in the reporting.

The prompt to start with

Upload the export directly into a chat with Claude, then paste something like this alongside it:

Prompt
I've attached an export of our sales pipeline. Please act as a sales performance analyst and tell me, in plain English, not as a data analyst: 1. Pipeline coverage: what's our total open pipeline value compared to our target for this period, and is that healthy? (A healthy business usually needs 3 to 4 times its target in open pipeline.) 2. Stage bottlenecks: which stage do deals sit in longest, and how does that compare to the other stages? 3. Rep performance spread: how much does performance vary between our best and worst performing rep, and does that look explainable by territory or account differences, or does it look like a coaching gap? 4. Win rate trend: is our win rate improving, flat, or getting worse over the last few months? 5. Deal velocity: is the average time from first contact to close getting longer or shorter? For each one, tell me plainly what you found, whether it's a good sign or a warning sign, and end with the single thing you'd look into first if you were me.

What good output looks like

A useful response reads like a short, plain-English memo, not a table of raw numbers. If Claude comes back vague, or says the data isn't consistent enough to say anything confident, that's a real finding too. Messy or incomplete CRM data is itself one of the most common things a proper pipeline audit turns up.

If you want to go further with code

If you're comfortable with a bit of code, or want something you can rerun every week, you can ask Claude to write and run the analysis directly rather than just describing it. Something like this is a reasonable starting point:

Example script
import pandas as pd df = pd.read_csv("pipeline_export.csv") # Win rate won = df[df["stage"] == "Closed Won"] lost = df[df["stage"] == "Closed Lost"] win_rate = len(won) / (len(won) + len(lost)) * 100 print(f"Win rate: {win_rate:.1f}%") # Pipeline coverage against a quarterly target target = 250000 # replace with your actual target open_pipeline = df[~df["stage"].isin(["Closed Won", "Closed Lost"])]["deal_value"].sum() print(f"Pipeline coverage: {open_pipeline / target:.1f}x") # Average time spent in each stage df["days_in_stage"] = ( pd.to_datetime(df["stage_updated"]) - pd.to_datetime(df["stage_entered"]) ).dt.days print(df.groupby("stage")["days_in_stage"].mean().sort_values(ascending=False))

Your actual column names will differ depending on your CRM, and that's fine, you can just ask Claude to adjust the script to match your export instead of editing it yourself. Either the prompt or the code gets you a genuinely useful first pass at your own data.

What this won't tell you

An AI analysis is a strong starting point, not the whole answer. It's good at surfacing the pattern, a stage that's stalling, a rep who's an outlier, a win rate drifting down, but it can't sit in on a call to tell you whether an underperforming rep needs coaching or needs managing out, and it can't tell you whether a new comp structure will actually land with the team you have. That judgment still comes from someone who has built and run these fixes before, which is exactly where a proper audit picks up from here.