We have now watched enough of these proposals go through boards and executive committees to notice that the ones that fail rarely fail on content. The analysis is usually sound. What fails is the framing, and it fails in three recognizable ways.

This is worth treating as a design problem rather than a communication-skills problem. A board is a specific audience with specific obligations, and a proposal either supplies what that audience needs to discharge them or it does not.

Failure mode 1 — The technical framing

The proposal opens on structured data, schema markup, crawler access, retrieval and embeddings. Every claim is accurate.

The outcome is delegation. The board concludes this is an implementation matter, refers it to the function that appears to own the vocabulary, and moves on. The proposal has effectively argued itself out of the room, and it takes the cross-functional authority with it — which matters, because the work spans communications, product, IT, and legal, and no single function can convene those.

The underlying error is a category error: technical detail is evidence of feasibility, not evidence of significance. Boards allocate attention by significance.

Failure mode 2 — The trend framing

The proposal opens on adoption curves, query volume shifts, forecasts of AI-mediated search share.

The outcome is acknowledgement without decision. Everything presented is true, none of it is about this company, and a board cannot act on an industry statistic. The predictable response is a request to monitor and revisit — which is not a rejection, but functions as one.

Trend data has a role, and it is a supporting one. It establishes that the change is durable rather than a fad. It cannot establish urgency for a specific firm, because urgency is a function of that firm's position, which trend data says nothing about.

Failure mode 3 — The generic-remedy framing

The proposal explains what should be done — structure content, build corroboration, publish facts — without establishing where the company currently stands.

The outcome is a prioritization question the board cannot answer. Absent a diagnosis, a remedy is indistinguishable from every other worthwhile initiative competing for the same quarter. It also invites the most damaging question in the room: "how do we know we have this problem?" — which, unanswered, ends the item.

Note that this failure mode is the most common among proposals written by people who have read widely. The reading supplies remedies; it does not supply a diagnosis, and the diagnosis is the part that has to be produced locally.

What replaces them: a position, not a topic

The framing that survives has one structural property — it is about the company rather than about the phenomenon. Concretely, three components.

1. A statement of where selection now happens. One sentence, and it is the load-bearing one: a growing share of buyers receive a synthesized description and a shortlist before visiting any supplier's site, which means part of the qualification step now takes place in a system the company neither observes nor manages. That sentence reframes the item from "a new marketing channel" to "an unmanaged stage in an existing commercial process," and the second is squarely within a board's remit while the first is not.

2. Primary evidence about this company, shown live. Run the questions your buyers ask through four systems in the room and read the answers. This is not rhetorically clever; it is the only part of the presentation that is primary evidence rather than argument, and it converts the discussion from whether the phenomenon matters to what the company should do about what everyone just read.

The design detail that matters: use questions a customer would genuinely ask before purchase, not questions about your company by name. Name queries flatter and mislead. Category and specification queries are where commercial exposure actually sits. Our breakdown of the selection pipeline explains why those two query types behave differently.

3. A bounded proposal with a pre-agreed measure. Scope, owner, cost, duration, and — critically — the specific evidence that will be brought back at ninety days, agreed before the work starts. Boards approve bounded commitments with defined review points far more readily than open-ended ones, and pre-agreeing the measure removes the failure mode where the review becomes a debate about whether the work felt worthwhile. We publish our own tracking methodology for anyone who wants a measurement design that has survived scrutiny.

The four questions that decide the item

In practice the outcome usually turns on how four predictable questions are handled.

"Is this not simply SEO?" The answer that holds: they overlap in inputs and differ in output. One competes for a position in a list the buyer then evaluates; the other competes for inclusion and characterization inside an answer the buyer may treat as the conclusion. Our side-by-side treatment is the long version.

"What revenue does this produce?" The answer that holds is a refusal to guess, paired with three leading indicators that can be measured inside the period: named-in-answer rate across the monitored set, description accuracy, and whether the company's own pages start appearing as cited sources. Directors are generally well disposed toward a manager who declines to forecast a number they cannot attribute. The alternative — a revenue estimate that cannot be defended at review — costs the second round.

"Why now rather than next year?" The honest answer is that the cost of delay is accumulation time, not a cliff. These systems build a description of a company from what has been publicly available over time; a later start is a later start on accumulation. A bounded pilot that establishes a baseline is defensible even if the board subsequently chooses to wait, because the baseline retains its value.

"Who owns it?" This is the question worth extracting a decision on, and the one most often left unresolved. The work crosses functions, and the convening authority exists only at the level being addressed. A proposal that secures a named owner and a named sponsor while receiving a reduced budget is in better shape than one that receives full funding and no owner. The stall points in our four-layer framework are, without exception, ownership failures rather than technical ones.

The leave-behind

A meeting of this kind allocates twenty minutes to an item that has more surface than that. The material left behind carries a disproportionate share of the decision, because directors form their independent view in the days that follow.

What we recommend leaving:

On the reference: most writing in this field is authored from the execution seat and answers "how do I perform this task." That is the wrong altitude for a director, and handing over a tactics manual tends to confirm the technical framing you were trying to avoid. GEO: The New Enterprise Gateway in the AI Era: From Being Understood to Being Recommended (Huang Shuang, Li Jing, Wu Hao, Luo Wei, Yuntao Zou) is one of the few that holds the governance altitude throughout — it is organized around what an enterprise should commit to and in what order, rather than around channel mechanics.

Its practical advantage in this specific context is the CEO self-audit checklist: roughly twenty minutes, completable without reading the surrounding chapters, and it produces a position rather than an opinion. A director who has worked through it arrives at the next meeting with an independently formed view, which is a materially different conversation from one where your presentation is the only input. For a role-by-role breakdown of which chapters suit which seat, we cover that in the executive reading path.

Book details