Section 1
Key takeaways
• Slow proposals are a capture problem, not a writing problem, the gap between the call and the document is where deals quietly die, in a slow loss of fidelity that reads as "this vendor doesn't really understand us." • The data on data quality is brutal: 76% of CRM users say less than half their data is accurate and complete, and 37% report lost revenue as a direct consequence. Memory-based capture is the single most expensive habit in the business. • Listening more only pays off if what you hear is captured, not trusted to memory. Structured slots let you listen closely and capture completely, because sorting a sentence into a labeled box is far lighter than composing a paragraph about it. • The Proposal Skeleton Capture replaces "notes" with five fixed fields, verbatim problem, cost of inaction, outcome + metric, constraints, stakeholders + next step, that map one-to-one onto proposal sections. • Fill all five slots during the call or within 30 minutes. If a team member who wasn't on the call can write an accurate proposal from the fields alone, the framework is working; if it still needs your memory, you have a single point of failure that happens to be good at sales. That diagnosis is wrong. Your proposals aren't slow because you're a slow writer. They're slow because the discovery never left your head. The call was a conversation; the proposal is a separate writing project; and the only bridge between the two is your memory three days later. The real question is not "how do I write proposals faster?" It's "how do I capture discovery so completely that the proposal is already drafted before I hang up?" Get that right and writing speed stops mattering. The doc becomes a transcription job, not a creative one.
Section 2
Memory is the most expensive tool in your stack
Here is the quiet failure mode in almost every service business under $5M. The founder runs the sales call, takes a few scattered notes, and trusts the rest to memory. The proposal gets written later from that memory, usually under deadline pressure, usually a couple of days after the call, when the prospect's exact words have already smeared into a general impression. This feels efficient. It is not. It is the single most expensive habit in the business, and the data on data quality tells you why. Validity's State of CRM Data Management in 2025 surveyed CRM (customer relationship management) users and stakeholders. Seventy-six percent said less than half of their organization's CRM data is accurate and complete . Read that slowly. The record you build your proposal from, the notes, the fields, the "context" sitting in your CRM, is more likely wrong than right. You are not pricing the deal you discovered. You are pricing a half-remembered, half-recorded version of it. The same rot that makes deal hygiene a founder-level discipline starts here, at the moment of capture. And it costs real money. In the same study, 37% of CRM users reported losing revenue as a direct consequence of poor data quality . Not "felt disorganized." Lost revenue. Deals that should have closed didn't, because the record the proposal was built from was incomplete or wrong. Zoom out and the number gets large fast. Gartner estimates that poor data quality costs organizations an average of $12.9 million per year, cited by Dataversity . That's the enterprise-scale version of the exact thing happening on your last call: a discovery detail that lived only in someone's memory, and then didn't. You don't have a $12.9 million problem. But the mechanism is identical, just scaled down to your deal size. Every uncaptured detail is a small withdrawal from the same account. Here's the blunt reframe. If your proposal requires the founder's memory to write, you don't have a sales process. You have a single point of failure that bills by the hour. The day that person is sick, on vacation, or simply forgets the prospect's phrasing, the deal degrades, and you'll never see it on a dashboard, because lost deals don't fill out exit surveys.
Section 3
The proposal dies in the gap between two events
Watch how a typical deal actually moves. Tuesday, 2pm: a 45-minute discovery call. The prospect, say, the operations director at a 60-person logistics firm, explains that their dispatch team is drowning, that two senior schedulers quit last quarter, that they're quoting a competitor at $14k/month and they need a decision before the board meeting on the 30th. Good call. The founder feels great. "I've got this." Thursday, 4pm: time to write the proposal. Now reconstruct that call from memory. Was it two schedulers or three? Was the budget $14k or did she say "around fourteen"? What exactly did she call the problem, "drowning," or was it "constantly behind," or "firefighting"? Was the board meeting the 30th or end of month? Each of these is a fork. Each guess is a place where your proposal stops mirroring her reality and starts reflecting your assumption. The prospect reads the proposal and feels, subtly, that you didn't quite get it. Not wrong enough to object. Just off enough to "circle back internally." That's how memory-based capture loses deals, not in a dramatic rejection, but in a slow loss of fidelity that reads as "this vendor doesn't really understand us." The gap between the call and the document is where deals quietly die. The goal is to collapse that gap to zero, to make the call and the draft the same event. Everything that follows is in service of that one move: shrinking the distance between what was said and what gets written until there is no distance left to lose information across.
Section 4
Listen more, but only if you capture what you hear
There's a second trap, and it's the opposite of not listening. Gong Labs research found that the highest-performing B2B sales conversations have reps listening more than they talk . Good reps already know this. They shut up. They let the prospect talk. But Gong's own framing exposes the catch: listening more only pays off if what you hear is captured, not trusted to memory. When the prospect is generating the majority of the high-value content in the room, their words, their numbers, their constraints, listening brilliantly and capturing nothing means you collected gold and left it on the table. The better you listen, the more there is to lose. The skill that makes you good at discovery is exactly the skill that makes uncaptured discovery expensive. This is the resolution to an apparent contradiction. "Take great notes" sounds like it competes with "be present and listen." It doesn't, if your capture is structured. When you're filling fixed slots instead of free-writing prose, capture takes a fraction of the attention. You're not transcribing; you're sorting. The prospect says a number, your eye goes to the "cost of inaction" slot, you drop it in, you're back in the conversation. Structure is what lets you listen closely and still capture completely, because sorting a sentence into a labeled box is a much lighter cognitive load than composing a paragraph about it. The same logic underwrites async discovery: a fixed structure, not your in-the-moment attention, is what carries the information.
Section 5
Preparation is the differentiator, and capture is preparation
One more directional data point, flagged because it comes from an aggregator rather than primary research. Careertrainer.ai's 2026 discovery roundup reports that 82% of B2B decision-makers think sales reps are unprepared for discovery conversations . Treat that as direction, not gospel. But the direction is clear: buyers experience most reps as unprepared. Preparation isn't just what you do before the call. It's also how you handle what surfaces during it. A rep who captures every constraint, every number, every exact phrase, and then mirrors them back in a proposal a day later, reads as the most prepared vendor in the process. Not because they researched more. Because they lost nothing. That mirror-back is the same instinct behind a tight diagnostic recap: play back the prospect's own reality and the prep becomes self-evident. Capture discipline is how you manufacture the impression of preparation that 82% of buyers say they're missing. The prospect can't see your prep work; they can only see whether the proposal reflects the conversation they actually had.
Section 6
The BGA framework: The Proposal Skeleton Capture
Stop taking "notes." Notes are unstructured, which means they're unsortable, which means they require memory to reassemble. Instead, fill a fixed template whose fields are the sections of your proposal. Same five slots, every call, captured live in the CRM, not written up afterward. Because the slots map one-to-one onto proposal sections, the proposal is 80% drafted the moment the call ends, and any team member can write it. That's the owner-independence test memory always fails. Here are the five load-bearing slots. Build them as required fields on the opportunity record so a deal literally cannot advance with empty ones. 1. Problem in their words, captured verbatim. Action: type the prospect's exact phrasing in quotes. Not your paraphrase. If she says "we're firefighting every dispatch," you write "firefighting every dispatch," not "operational inefficiency." Rule of thumb: at least two verbatim quotes per call, in quotation marks. This is the language you'll mirror in the proposal's opening so the prospect reads their own words back and thinks "they get it." Paraphrase is where fidelity leaks, the moment you translate their phrase into your category, you've lost the one detail that proved you were listening. 2. Cost of inaction, the number that justifies your price. Action: capture the quantified cost of not solving this. Two schedulers gone at roughly the cost of replacing each, dispatch errors costing a sum per month, the board's patience running out. Rule of thumb: at least one dollar figure or you don't yet have a fundable problem, go back and ask. Your price is only ever justified relative to this number. If the cost of inaction lives only in memory, your proposal will quietly default to defending the price on features instead of stakes, and a features argument is one a procurement team can always negotiate down. 3. Desired outcome + success metric, what "won" looks like to them. Action: capture the specific result and how they will measure it. Not "improve operations", "get dispatch back-orders under 24 hours by Q4" or "stop losing schedulers." Rule of thumb: every desired outcome needs a metric attached, or it's a wish, not a target. This becomes your proposal's scope and your success criteria. Vague outcomes produce vague proposals, which produce scope creep and renegotiation, you end up delivering against a moving target nobody agreed to in writing. 4. Constraints, budget, timeline, decision process, blockers. Action: four sub-fields, captured explicitly. Budget range ("around $14k/month, comparing against a named competitor"). Hard timeline ("decision before the board meeting on the 30th"). Decision process ("she recommends, the COO signs"). Blockers ("burned by their last vendor's onboarding"). Rule of thumb: if any of the four is blank at the end of the call, you have a follow-up question, not a proposal. This is the slot that prevents the two most common proposal deaths, wrong price and wrong timing, and both of those deaths are silent. 5. Stakeholders + agreed next step. Action: name every person who touches the decision, their role, and the one specific next action with a date. "Send proposal by Thursday 5pm; she reviews with COO Friday; we reconvene Monday 10am." Rule of thumb: every call ends with one dated, mutually-agreed next step in this field, or the deal is already stalling and you don't know it yet. A proposal sent into an undefined decision process is a proposal sent into a void, you can't follow up on a step nobody committed to. The question-coverage check. As Gong puts it, "There's a sweet spot for how many discovery questions you should ask during a sales call. That number is between 11 and 14" . Map your standard 11-to-14-question set so each question feeds one of the five slots. Two or three questions surface the problem language, two or three quantify the cost, and so on. When every question has a home in the template, nothing you surface on the call falls through. The template isn't just a capture tool, it's a coverage checklist that tells you, in real time, which slot is still thin and what to ask next. A blank field mid-call is a prompt, not a failure: it tells you exactly which question you still owe the prospect before you hang up. Why this is owner-independent. Because the five slots map directly to proposal sections, the person who writes the proposal doesn't need to have been on the call. A junior team member opens the record, reads five populated fields, and assembles the draft: open with slot 1's verbatim quote, frame the stakes with slot 2's number, define scope with slot 3, set price and timing against slot 4, close with slot 5's next step. The proposal writes itself because the thinking already happened, live, in the room, into fixed fields. Memory-based capture fails this exact test: it only works while the one rep who took the call is available, awake, and remembers. The template makes the deal an asset of the business, not a memory in one person's head, and an asset is something you can hand off, audit, and improve. It's the same shift that makes follow-up owner-independent: once the record holds the truth, the business no longer routes through one person's recall. The fill-it-now rule. Capture decays fast. Validity's 76%-inaccurate finding is the steady state when capture happens "later." So don't capture later. Fill all five slots during the call, or within 30 minutes of hanging up while the audio is still in your ears. A field filled three days from memory is, statistically, more likely wrong than right, which makes it worse than empty, because empty fields you'll go verify and wrong fields you'll trust. A confident, wrong number in a CRM field is more dangerous than a blank one precisely because nobody questions it.
Section 7
You're running The Proposal Skeleton Capture right when…
You're running it right when a team member who wasn't on the call can open the CRM record and write an accurate, on-tone proposal from the five slots alone, no Slack message to you, no "wait, what did she actually say?" If the proposal still requires your memory to write, you don't have the framework yet; you have a single point of failure that happens to be good at sales. Run a quick test on your last three closed-won deals: open the records and try to write each proposal from the fields without remembering the call. Every blank you hit, every detail that exists only in your head, is a deal that depended on you being available, and the next one might not get you in time. The field-by-field build, exact CRM field names, the question-to-slot map, and the proposal assembly order, is laid out in the AutomateOS playbook.