Enhancing B2B SaaS Sales with AI-Powered Pre-Call Research

Picture an account executive, called an AE in this material, whose first meeting of the day is with a prospect she has researched carefully. The remaining meetings arrive one after another. By mid-morning she is no longer reading anything before she dials in. She knows this is a loss. Research before a call helps her understand who she is speaking with and what might matter to them, yet her schedule leaves no room to do it well.

The same pattern appeared in a real project. During early scoping, six discovery interviews were conducted, three with AEs at different levels of tenure. The AEs reported that they prepared for the first call of the day, skipped preparation for the rest, and felt uneasy about the gap. The useful question is not why the AEs were careless. The useful question is what constraint made skipping rational, and what a tool would need to do to remove that constraint. This article works through that reasoning and then follows the pilot built to test one answer.

The skipping was not caused by a lack of motivation. It was bounded by volume. Each AE was expected to take between eight and twelve calls a day. That range sets how many separate research sessions a day would be needed if every call received its own preparation.

The second number made the problem concrete. The existing research process took fifteen minutes for each call. The AEs had only about fifteen minutes between calls. Placed side by side, the two figures leave almost no slack: the research step alone consumes the entire gap the AE has available. Any preparation beyond that point has to come out of time the AE does not really own.

The diagram shows the timeline with one gap. Read it as a statement about who carries the preparation load. The AE's care about preparing is not the variable here. The time available between calls is the variable.

The scope agreed with the project sponsor was deliberately narrow. It was called the first slice and asked for one specific output: a pre-call briefing agent that produces a one-page brief twenty minutes before each scheduled call. The brief is delivered to the AE's email and to Slack. Three parts of this definition each answer a separate design question, so each gets its own treatment below.

The source does not describe how the brief is assembled internally, and this article does not guess at it. What the scope does establish is the pipeline's outer shape: a scheduled call triggers a brief, and the brief reaches the AE through two channels.

A briefing agent can be judged in more than one way, and the project chose two measures that answer different questions. The first is a self-reported usefulness score on a scale from 1 to 10, which the AE gives after using the brief. It captures the AE's own judgment of whether the brief helped. The second is briefing open rate, which tracks whether the brief was actually opened. The source calls open rate a leading indicator.

The two measures work best together. A brief that is opened often but rated poorly suggests the AEs are checking it and finding little use. A brief that is rated highly but rarely opened suggests the content may be good while delivery is failing to reach the AE at the right moment. Reading only one of the two numbers hides one of these possibilities.

The validation threshold combines the two measures. The pilot was planned as a two-week run with eight AEs. To pass, the average usefulness score had to reach 7 or higher, and the briefing open rate had to be above eighty percent. Both conditions must hold. Meeting one of them is not enough.