A&A INSIGHTS
How much to propose at first contact: judge by the second order
Size the first proposal by whether the buyer can keep operating inside it, not by order value. From Anthropic’s buying-agent account and YC’s Growth Office Hours.
日本語で読む
THE STARTING POINT
Size the first proposal not by the largest amount you could win but by whether the buyer can keep operating inside that scope, and write down the conditions for not putting the larger proposal on the table before the conversation starts. Judge the design by the share of buyers who placed a second order — though that a smaller first proposal makes a second order more likely is our own hypothesis, not something the sources show.
Size the first scope by whether they can keep operating it, not by what you could win
A&A perspective
How much to propose to someone who has come to you for a conversation is decided neither by their budget nor by your own spare capacity, but by whether they can keep operating inside the scope you propose. Three steps. First, write down in prose, before the conversation, the conditions under which you will not put the larger proposal forward. Second, in the conversation, confirm only whether those conditions are met — and if they are, do not put it forward. Third, judge the quality of your proposal design by the share of buyers who came back with a second request, not by the average first order value. That is this article’s answer.
A&A perspective
The reason for that order is that engagements taken large at first contact stall in the buyer’s operation rather than in your proposal. If the scope exceeds what they can operate, the work is delivered and then not used, and work that is not used produces no second request. Our working assumption is that for a one- or two-person firm, taking a second engagement from an existing buyer costs less effort than sourcing one more new conversation. Neither source states this. So cutting the first scope is not giving up an opportunity; it is the design that sets total revenue per acquisition. What follows looks at a seller that published its decision not to steer buyers upwards, and at the argument for checking repeat behaviour before widening acquisition, and then turns both into conditions you can write down.
A seller that called not steering buyers upwards a feature
From the sources
In an article dated September 30, 2026, Anthropic describes rebuilding its inbound handling around a buying agent. The author is Carl Johnson, a sales development leader at the company. On the state before the rebuild, the article says "We had tens of thousands of inbound requests every month" and that "Reps spent their days answering questions the docs already covered". On the current arrangement it states that customers choose at the start whether to talk to an agent or a rep: "We intentionally left this experience as opt-in". Conversations are described as ending in one of three ways: a purchase, a hand-off to a rep, or a quick answer.
From the sources
Among the lessons, the article addresses steering buyers to higher plans directly: "The agent often points small teams to our Team plan instead of Enterprise, because that's the right answer for them in some cases." It continues, "We decided that's a feature.", and gives the reason: "A buyer who lands on the right plan instead of being upsold is more likely to stay and have a better experience." A separate lesson describes recording why each hand-off happened: "Each time the agent passes a buyer to a rep, it explains why." and "Early on, most of those reasons were things customers couldn't yet do on their own with our self-service products." It then places the improvement work alongside the outcome: "Those reasons shaped how we improve the experience, and the share of conversations that needed a person to help close the deal has fallen by about half." That proportion is the company’s own reported figure, with no comparison conditions published.
A&A perspective
Two shapes of judgement are borrowable here. The first is deciding in advance that recommending the smaller correct plan is an acceptable outcome rather than a miss. The company writes that its agent often points small teams to the lower plan because that is the right answer for them in some cases, and that it decided to treat this as a feature. Judge it afresh in each conversation and the scope widens the moment the buyer’s budget becomes visible. The second is the mechanism of recording why each hand-off happened. A solo firm can reproduce that by leaving one line per conversation recording why it did not put the larger proposal forward. Note, though, that the article does not isolate the revenue effect of either decision. What it states stops at the company’s own judgement that a buyer who lands on the right plan is more likely to stay and has a better experience.
A&A perspective
The difference in scale belongs here too. The article is describing tens of thousands of inbound requests a month and thousands of conversations a day. That is three or more orders of magnitude away from a solo firm fielding a few conversations a week. A change expressed as a proportion — "fallen by about half" — therefore does not become a forecast for a small business. What is borrowable is the format of writing the judgement down in advance, not the figures. The article is also a first-person account published by Anthropic to introduce its own sales organisation and its own agent platform; it is not an independent audit.
| What you confirmed during the conversation | Put the larger proposal forward? | First scope offered instead |
|---|---|---|
| Nobody has been named to operate the result yet | No | Limit the scope to what works until an operator is named, and say up front that the next proposal follows once they name one |
| The target process is not written down and differs by person | No | Take only the work of reducing the current process to a single written version. Building is not included |
| Exactly one person inside the buyer can touch what gets built | Conditionally | Confirm how many hours that person actually has, and cut the scope to what runs inside those hours |
| Budget expires on a date, and phasing means nothing starts this fiscal year | Yes | State the deadline constraint as the reason in the proposal, and ask them to name the operator up front regardless |
| It is a migration where replacing only part of the system breaks consistency | Yes | Split the migration along the seams of the work rather than the technology, and write the point you can roll back to into the proposal |
| They previously abandoned a similar implementation | No | Make the first engagement an investigation into why the previous attempt stopped. Defer the build proposal |
Count the rounds, not just the outcome
From the sources
The same Anthropic article reports what changed on the rep side. Of the customers the agent escalates to the team it states that "they turn into opportunities more than twice as often as leads from the old form, and close about five days faster", and that a rep starts the conversation already knowing what the customer needs and what they have been told. As an individual example it records an inside sales rep who "used to exchange about 10 emails to close a deal; now it takes about six", and says of the same rep that "He’s been able to 2.5x his output on closed won deals since launching the Buying Agent". All of these are the company’s own reported figures, with no comparison conditions published.
A&A perspective
What a service firm can take from this is to make the number of rounds an observed quantity. When a first proposal does not fit the buyer’s operation, the misfit shows up first not as a refusal but as extra traffic: questions circling back on the assumptions in the quote, requests to carve a piece out of the scope, extra explanation for whoever signs. Counting rounds from conversation to agreed proposal therefore surfaces a defect in your condition sheet earlier than waiting for the second request. The relationship between fewer rounds and a better-fitting scope is our reading rather than something the source states; in the company’s own account, the reason given for fewer rounds is that buyers arrived already well informed, not that the proposals were made smaller.
The ordering: check whether they come back before you widen acquisition
From the sources
In the transcript of Growth Office Hours, published by Y Combinator on December 1, 2017, YC partner Anu Hariharan says that the first thing to check before setting up a growth team is retention: "before you set up a growth team, the number one step you need to check is whether you have strong retention". She notes that too many companies form a growth team and then ask why they are not growing, and puts it as "Well, you have leaky bucket of water."
From the sources
In the same episode, YC partner Gustaf Alströmer sets the bar for what counts as a repeat: "Just the act of coming back isn’t meaningful unless you do something that gives you value from the product."
A&A perspective
This episode is a 2017 discussion grounded in SaaS and marketplaces, and repeat ordering in a service business cannot be measured with the same metric as product retention. The ordering, however, transfers. Before increasing the number of conversations, check whether the buyers you already took come back. And count a return only when a request of the same kind arrives from their side. An expansion you proposed and sold is evidence of your own selling, not evidence that the buyer could keep operating. For judging the size of a first proposal, that distinction is decisive. Note that neither source shows that a smaller first proposal causes a second order. That link is our hypothesis.
Write the conditions for withholding the larger proposal before the conversation
A&A perspective
Write the conditions in terms of facts about the buyer’s operation, not their budget or their industry. There are not many usable items. Has an operator been named? How many hours a week can that person actually give this? Is the target process written down? Have they abandoned a similar implementation before? Those four settle the capability side only. Two of the rows below — the budget deadline and the migration — come from the buyer’s constraints rather than their capability, and are not derivable from the four. Where a condition applies, write the alternative scope on the same line. Deciding only "do not propose" leaves you without an alternative in the room, and the large proposal goes out anyway.
Hypothetical example
The table below is a hypothetical condition sheet for an imagined small AI service firm. It is not a real deal and not a standard we operate. Two of the six rows say yes. One is where the budget expires on a date and phasing means nothing starts this fiscal year; the other is a migration where replacing only part of the system breaks consistency. In both, taking a wider first scope is better for the buyer. The purpose of a condition sheet is not to make scopes smaller; it is to make you able to explain to yourself when a larger one is right.
Judge by the share that produced a second order, not by the first amount
A&A perspective
How to count a second order — as a rate or as a list of names, which requests count, where to put the observation window — is handled in our published piece "Before scaling acquisition: how to count a customer as returned". The full definition is there and is not restated; above, only the part that bears on scope is borrowed. What this article adds is one thing: a column on that list for the first scope, so the resulting number feeds back into proposal design.
A&A perspective
So the metric for judging proposal design is not average first order value but the share of first-time buyers who came back with a second request. Three fields per line on that list: the first scope, whether a second request arrived, and if it did not, where it stopped. Settle the counting alone and, if the proposal scope is still judged on the spot each time, you will not know what you changed that produced the second order. Write where it stopped in terms of the operation rather than "budget": nobody was named to operate it, the process was never reduced to one version, the person who could touch it left. As a rough marker, once that list reaches about ten entries, which row of the condition sheet needs changing becomes visible. There is no measured basis for ten.
One line on why you withheld it, and the condition sheet maintains itself
A&A perspective
This is a different list from the one in the previous section: it covers every conversation, including the ones that never became an engagement. The part of the quoted article where each hand-off has to explain itself, and those reasons feed improvement, is almost directly reproducible at one-person scale. Leave one line per conversation recording why you did not put the larger proposal forward. Three fields: what they asked about, which condition applied, and what you offered instead. Run it for four weeks and the conversations stalling on the same condition become visible. If "no operator named" keeps being the cause, one option that appears is to make naming the operator the first paid engagement in itself.
Hypothetical example
This is a hypothetical operating example, not our record, and the counts are illustrative rather than measured thresholds. Starting it needs no tooling, only somewhere to write three lines per conversation. The important discipline is not to write the reason as "their budget did not fit". Budget is an outcome, not a condition. If you cannot write the reason in words that correspond to a row of the condition sheet, that conversation is of a kind the sheet does not cover, so mark it separately as a candidate for a new row.
What this article does not decide, and the next step
A&A perspective
The limits of the sources, stated plainly. The Anthropic article is a first-person account published in a promotional context, introducing the company’s own sales organisation and its own agent platform; it is not an independent audit. The about-half, about-five-days and 2.5x figures are all self-reported, with no comparison conditions published. The company handles thousands of conversations a day, three or more orders of magnitude away from a solo firm with a few conversations a week. The YC episode is a 2017 discussion premised on setting up growth teams at SaaS and marketplace companies, and repeat ordering in service work is not measurable with product retention metrics. And neither source shows that a smaller first proposal causes a second order. The link between first scope and continuation is our hypothesis, and we present no engagement or retention record of our own.
A&A perspective
The cases where this does not apply bear repeating. Where the buyer’s budget expires on a date and phasing means nothing starts this fiscal year, and where a migration breaks consistency if only part of the system is replaced, a wider first scope serves the buyer better. Unless those two are written into the condition sheet explicitly, the sheet degrades into a device for proposing small. The next step is to take your last five conversations and write out, side by side, whether you put the larger proposal forward and whether a second request arrived. For the whole flow, "AI-native GTM: a practical guide for solo founders and small teams" is the map; for productising once repetition is visible, see "Productizing repeat service work: judge by pass rate on hard cases". If you cannot settle which rows belong on your own condition sheet, that is a question for the advisory stage.
Size the first proposal by whether the buyer can keep operating inside it rather than by the largest amount you could win, and write the conditions for withholding the larger proposal before the conversation starts. Anthropic’s sales team wrote that when its agent points a small team to the lower plan, because that is the right answer for them in some cases, it treats that as a feature rather than a defect; it also recorded the reason for every hand-off so it could feed improvement. The YC discussion supplies the ordering — check repeat behaviour before widening acquisition — and the bar for what counts as a repeat. Applied to service work, the metric is the share of buyers who came back from their own side, not average first order value. Neither source shows that a smaller first proposal causes a second order; that link is our hypothesis. And where a budget expires on a date, or where partial migration breaks consistency, taking a wider first scope is what serves the buyer.
Sources & editorial note
Primary pages read for this article. Publication dates below belong to the sources; access dates record our research.
- How Anthropic’s sales team rebuilt inbound with Claude Managed Agents
Anthropic · September 30, 2026 (date shown on page)
Accessed 2026-10-07 - Growth Office Hours with Anu Hariharan and Gustaf Alstromer
Y Combinator · December 1, 2017 (date shown on page)
Accessed 2026-10-07
AI-assisted editorial production
A&A uses AI for research, writing, translation and editorial checks. Source facts, our analysis and hypothetical examples are labeled separately.
Editorial check: 2026-10-07