Pre-call Research Agent Project Brief
The proposed pre-call briefing agent generates a one-page briefing document twenty minutes before each scheduled call.
A Salesperson's Afternoon of Three Calls
Picture an account executive, the salesperson who owns a client relationship, with three calls on the calendar in one afternoon. Each call involves a different company. Each company has its own history: earlier conversations, recent press coverage, and people on the other side of the table whose backgrounds the salesperson may not remember. Walking into every call with that context in mind takes time the salesperson often does not have.
The source describes a proposed solution to this problem: a pre-call briefing agent. The word proposed matters. This is a system design to be built, not a finished tool with a track record. A portfolio brief describing it has to make the design clear enough that someone else could evaluate it, critique it, or build it. This article works through the design from three angles: when the agent runs, where its information comes from, and where its output goes. Once you can state those three things precisely, you have the core of the design.
The proposed pre-call briefing agent generates a one-page briefing document twenty minutes before each scheduled call. Read as a timed process, it has four stages: a trigger, a gathering step, a generation step, and a delivery step. The sections below follow those stages in order and then examine how a portfolio brief should describe and judge the design.
When the Briefing Is Produced
The trigger is time-based and tied to the calendar. The source specifies that the briefing is produced twenty minutes before each scheduled call. Nothing in the design asks the salesperson to request a briefing. The schedule itself sets the start of the work, so the salesperson does not have to remember to ask for one.
The timing is part of the design, not an afterthought. A briefing that arrives at the call's start would be far less useful than one that arrives twenty minutes earlier, because the earlier one leaves the salesperson a short window to read it. When you describe this agent in a brief, state the twenty-minute timing in the design itself rather than in a footnote.
Four Inputs to the Briefing
The briefing is assembled from four kinds of information: the company's CRM, recent news mentions, the last three calls in the relationship, and the public LinkedIn profiles of meeting attendees. The source names each one specifically, and a portfolio brief should be able to account for every input.
The four sources answer different questions. The CRM describes the relationship as the company has recorded it. The news mentions describe what has happened to the company recently, outside the sales process. The last three calls describe what was actually said in the relationship, which can differ from what the CRM records. The LinkedIn profiles describe the people who will attend this particular meeting. A briefing that draws on all four covers the account, its recent context, its history of conversations, and the audience for the next call. Dropping one leaves a blind spot the other sources cannot fill.
Where the Finished Briefing Goes
Once the one-page document exists, the design sends it to the account executive through two channels: the AE's email and a Slack channel that the AE monitors. The design sends the same briefing to each destination rather than producing a different document for each one.
The phrase Slack channel the AE monitors is a deliberate detail. The design does not say the briefing goes to a private message. It says the briefing goes to a channel the AE watches. That makes delivery depend on the AE's habits as well as on the system's output. A portfolio brief should name both channels and note that the Slack destination is one the AE already monitors, rather than a new place the AE must learn to check.
Misreadings to Avoid in a Portfolio Brief
A portfolio brief about this agent is easy to misread in predictable ways. The source's guidance points to the corrections below. In a portfolio brief, describe the agent as a proposed design, name every source and both channels, and mark any unspecified details as open decisions.
Common mistake Presenting the agent as a finished tool with results
Why it failsThe source describes a system design to be built. It has no track record in the source material.
Do this insteadWrite it as a proposed design, and describe what it is intended to do and how its parts fit together.
Common mistake Listing only some of the sources
What it looks likeNaming the CRM and news but leaving out the LinkedIn profiles.
Why it failsA portfolio brief should account for every input, and each source answers a different question.
Do this insteadName all four sources and state what each contributes.
Common mistake Describing delivery to one place only
Why it failsThe design sends the same briefing to the AE's email and to a Slack channel the AE monitors.
Do this insteadName both channels and note that the Slack channel is one the AE already watches.
Common mistake Placing the timing in a footnote
Why it failsThe twenty-minute lead time is what makes the briefing useful, because it leaves a window to read it.
Do this insteadState the trigger and its timing as part of the core design.
Common mistake Silently adding steps the source does not describe
What it looks likeAdding a step where the salesperson approves the briefing before it is sent.
Why it failsThat changes the design, and a reviewer cannot tell which version is being evaluated.
Do this insteadMark any added step as a change to the proposed design.
Predicting How the Brief Is Judged
What do you think happens?
A brief proposes a pre-call research agent and says its success will be shown by one number: the sales engineers themselves estimate how many hours per week the tool saves them. Is that enough to make the brief strong?
Reveal answer
No, a single self-reported number should be paired with a leading indicator
The source pairs a self-reported number with a leading indicator because self-reports can be shaded, even without intent, while observed behavior is harder to bend. In the source's example, the leading indicator is open rate.
Many people writing a brief for an AI tool assume the strongest brief is the one with the most ambitious promise. A brief that says the system will transform the sales process, learn everything about every prospect, and automate the whole account cycle sounds impressive. The source's critique points elsewhere. A strong brief is judged by how specifically it ties the system to a real job, how it fits the way people already work, how its quality is measured, who measures it, and how small the first build is. Breadth and sophistication of the AI are not the test. A brief that promises everything usually cannot say which single task it is responsible for, and that vagueness spreads into every other part of the brief.
The Five Features That Strengthen a Brief
The source identifies five features that make a brief hold up. Each one answers a different question an evaluator will ask. Taken together they turn a vague promise into a design someone can check.
Broad Pitch Versus Smallest Slice
| Feature | Broad AI pitch | Smallest end-to-end slice |
|---|---|---|
| Promise | Transforms the whole sales process and automates the account cycle | Produces one pre-call briefing from input to output |
| Job definition | Usually cannot name the single task it is responsible for | Names one job, such as pre-call research |
| Success metric | Hard to attach one metric when many jobs are mixed | A metric that measures exactly the one job |
| What it can demonstrate | Little that can be checked end to end | Whether the full path works for a real user |
Contrast between an ambitious pitch and a smallest end-to-end slice, based on the source's critique.
Critiquing a Sample Brief
Reading a weak and a strong brief
A student submits a brief: The agent will transform how account executives prepare, learn everything about every prospect, and automate the account cycle. Success will be shown by the sales engineers estimating the hours per week the tool saves. Evaluate the brief using the features from this article.
- 1Check the job
The brief names no single task. It claims to transform preparation, learn everything, and automate the cycle, so the job is not defined.
- 2Check the sources and channels
The brief names no data sources and no delivery channel, so a reviewer cannot account for the inputs or the destination.
- 3Check the rubric
There is no multi-criterion rubric. Quality is not broken into checkable dimensions.
- 4Check the metrics
The only metric is a single self-reported number with no leading indicator, so it is open to shading.
- 5Check the grading and scope
The grader is not named, and the scope spans the whole account cycle rather than a smallest slice.
The brief is weak. A stronger version would name pre-call research as the job, list the four sources and both channels, state the twenty-minute trigger, define a rubric with separate dimensions, pair the self-reported number with a leading indicator such as open rate, have account executives do the grading, and scope the first build to the single briefing.
Write a short critique of this sentence from a portfolio brief: The agent will read everything about each prospect and save the team time. Identify at least three specific features it lacks, and rewrite the success metric so that it pairs a self-reported number with a leading indicator.
Hints
- Ask which single job the sentence names, if any.
- Check whether the sentence names the four sources, the trigger timing, and both delivery channels.
- For the metric, a leading indicator is an earlier, observed signal that predicts the outcome that matters.
Key Takeaways
- The proposed pre-call briefing agent produces a one-page briefing twenty minutes before each scheduled call, triggered by the calendar rather than by a request from the salesperson.
- The briefing draws on four sources: the CRM, recent news mentions, the last three calls, and the LinkedIn profiles of meeting attendees, and each answers a different question.
- The finished briefing goes to the AE's email and to a Slack channel the AE already monitors, and a portfolio brief should describe the design as proposed and mark open decisions.
- A strong brief is judged by specificity: a named job-to-be-done, a multi-criterion rubric, paired self-reported and leading-indicator metrics, credible grading by the right stakeholders, and a smallest end-to-end slice.
- A broad, ambitious AI pitch can sound stronger but usually cannot name the single task it is responsible for, which weakens every other part of the brief.