AI for the New Sales Rep: Built for the First 90 Days, Not the Tenth Quarter

Most sales AI is tuned on a rep's established baseline. A new rep doesn't have one yet. What actually helps starts counting from day one.

2026-08-03 · SARA — KEEL'S AI DEAL ASSISTANT · GETKEEL.IO

Priya's second week on the job, her manager pulls up her call recordings next to a rep who's closed deals here for three years.

Her talk-ratio looks bad next to his. It would look bad next to anyone's, four weeks in.

Most sales AI is tuned on a season of history that hasn't happened yet

Most sales tools are built around an established rep's baseline — years of calls, a settled rhythm, a talk-ratio that already found its groove.

A new rep doesn't have that yet. She has four weeks and a product she's still learning to describe without notes.

Measured against a tenured rep's numbers, almost anything she does reads as a gap. That's not a signal about her. It's a mismatch between what the tool assumes and what week four actually looks like.

The real problem: there's no baseline yet to build from

A rep with a year in the seat has an unspoken standard to check a new call against — her own last quarter.

A new rep has no "last quarter."

Every call she takes is the first entry in the record, not a deviation from one already there. Most tools weren't built for that starting condition — they assume the baseline already exists somewhere and just needs surfacing.

Scoring her against the team average isn't ramp. It's a verdict.

Call scoring and team scorecards exist to catch an established rep drifting from a norm. A new rep isn't drifting — she has nothing yet to drift from.

Score her against the team average in week three, and the scorecard reads like a verdict, not a coaching tool.

The gap it's measuring isn't a performance problem. It's a ramp problem, wearing a performance problem's clothes.

What ramp actually needs: a record that starts at zero, honestly

The useful version of this doesn't compare a new rep to anyone else. It just holds what she's learning, in the order she learns it.

The objection that caught her flat-footed in week two. The answer she fumbled and wants ready the next time it comes up. The call where a buyer's tone shifted and she didn't yet know what that meant — three calls later, she does, and the earlier note is what lets her connect the two.

None of that needs a tenured rep's numbers next to it to be useful.

This is the same gap ADHD reps and introverted reps already fall into

Different cause, same shape: a workflow built for one kind of rep quietly fails everyone outside that shape.

The ADHD case is about when the capture happens — before the detail evaporates. The introvert case is about where it happens — alone, not performed for a room. The new-rep case is about what it's measured against — nothing yet, and pretending otherwise is the mistake.

Built to start counting from day one, not quarter three

Sara doesn't score a new rep against anyone else's history. Talk through a call the moment it ends, and the read gets saved as its own first entry — not measured against a baseline that hasn't been built yet.

By the tenth call, she has ten entries to work from instead of a verdict handed down in week three. Founders Club is invite-reviewed. Apply at getkeel.io/founders.

The first 90 days were never really about hitting a tenured rep's numbers early. They're about building the record a tenured rep already has — one real call at a time, starting from the first one instead of being measured against someone else's tenth quarter.

By the team at Keel. We're building Sara, an AI deal assistant for the moments that don't get recorded.