[Series 22] Why AI for Sales Support Fails to Deliver Results—Context Provisioning to Break Free from "Surface-Level AI Proposals"

I have been hearing this voice very often lately: "We introduced AI into our Go-to-Market (GTM) strategy, but it isn't delivering results." Companies are implementing AI sales agents and letting AI handle proposals and follow-up emails, yet sales negotiations are not progressing as expected. The proposals generated by AI are perfect in grammar and format, yet for some reason, they fail to resonate. The cause is not the model or the budget, but rather the single point of whether the company's strengths are being provided in a "format that AI can read." Finally, I will delve into the paradox that the more you provide these strengths to AI, the easier it becomes for those strengths to be imitated.

Why your AI doesn't know your "company's strengths"

The reason GTM AI is failing to deliver results in many companies is not because the AI itself lacks capability. In a 2025 survey conducted by Growth Unhinged and GTM Strategist ("2025 State of B2B GTM report," n=195), 53% of GTM leaders responded that "AI has little to no impact, or at best, it is limited." Only 24% felt they had achieved significant results. Despite this, three out of four people reported feeling pressure from above to "use AI." Implementation is moving forward, yet there is no tangible progress. This quiet gap is likely what is happening in many workplaces right now.

What I remembered when looking at these numbers was the "perfection" of the proposals written by AI. The grammar and format are polished, and at first glance, they seem flawless. However, the crucial elements—what the customer is truly struggling with and where the company wins compared to competitors—are nowhere to be found. It is like a model answer that only focuses on the appearance. Because the exterior is well-organized, it is actually harder to notice the emptiness inside.

So, why can't AI talk about a company's strengths? The answer is disappointingly simple: it is because we are not providing those strengths in a "format that AI can read." Strengths that exist only in our heads or as personal intuition are invisible to AI. When the MIT research team (NANDA) analyzed the causes of AI implementation failure in their report "The GenAI Divide: State of AI in Business 2025," they pointed out that the root of many failures is a "lack of adaptation to context." AI will not infer things that lie outside the provided context.

"Showing to the outside" and "letting it use on the inside" are different tasks

There is a distinction I would like to clarify here. It is that there are two types of context provision for AI: "outward-facing" and "inward-facing." Outward-facing refers to optimization for buyer-side AI (generative AI search and summarization) to correctly find and cite your company, often called AEO or GEO. This is optimization for "showing to the outside," which I will cover in detail in a separate article. On the other hand, this article deals with the counterpart: the "inward-facing" work of ensuring your company's GTM AI functions correctly.

These two are often confused because they start with the same material. Both use primary information held by Product Marketing (PMM), such as ICP (Ideal Customer Profile), positioning, and competitor comparisons. However, the direction of processing and the results to look for are completely different. Outward-facing results are evaluated by "how much it was cited," while inward-facing results are evaluated by "how accurate the AI's execution became." The point here is that even if you start with the same material, you should not measure both with the same yardstick.

I remember when I was leading a B2B software business. The reason we were winning in sales negotiations was, in reality, only in the heads of a few top salespeople. It wasn't that we didn't have materials. We did, but the content varied by representative, and "why we win" had not become shared team knowledge. So, I interviewed them to understand the reasons for winning and the conditions for losing, organized them into a single model, and created a state where anyone could follow the same path to victory. Looking back now, that was inward-facing provision for people. Even if the recipient changes to an AI agent, the essence of the work—extracting the implicitly held winning path and organizing it into a usable form—has not changed at all.

Translating positioning into a "structure that AI can read"

So, how do you create that "inward-facing" capability? Ultimately, it is the work of translating positioning into structure. I have long maintained that the essence of PMM is being a translator between organizations. The role of PMM is to bridge different languages—the language of development, the language of sales, and the language of the customer. What has been added in the era of GTM AI is a fourth language: "the language that AI reads." Translating the STP, especially the positioning design systematized by Kotler & Keller (2012) in "Marketing Management," into a structure that AI can interpret is the core of inward-facing provision.

What is important here is that "more is not necessarily better" when it comes to the context you provide. Practitioner Maja Voje advises, "Write down everything—strategy, ICP, brand voice—and give it to the AI." On the other hand, AI creators like Anthropic advise, "Put only the most effective information into the context window." At first glance, these are opposites. This contradiction can be resolved by separating where you store information and where you actually provide it. In other words, write down "everything" on the storage side, and select and provide "only what is necessary" at the moment the AI acts. The design is to keep everything, but not show everything.

Another point, which is subtle but effective, is "where you place it." If you stack long contexts as they are, the AI is more likely to miss information placed in the middle. Therefore, the more important the strength, the more it should be placed at the beginning or end of the context. What really works is the selection of "what, at what granularity, and where to place it." When I created the tagging mechanism for Japanese reviews at Amazon, this is exactly what I realized. Transforming a massive, raw chunk of reviews into a structure that a machine can read. Which unit, where, and what to tag—if you make one mistake in that design, the subsequent processing will collapse. Structuring strengths is a task that demands the same level of precision as this tag design.

A pitfall in structuring that is often overlooked

Now, after reading this far, you might think, "Then I should just structure all of our company's strengths and give them to the AI." However, it is not that simple. In fact, this is what I most want to convey in this article. The very act of making strengths explicit and structured is a double-edged sword.

In discussions of competitive advantage, it is said that difficulty of imitation is the source of advantage. In 1991, Barney presented the Resource-Based View (Note 1), stating that resources that are valuable, rare, difficult to imitate, and difficult to substitute create sustainable competitive advantage. Later, he added the condition of "organization" to this, developing it into the VRIO framework. And the difficulty of imitation often arises from tacitness and causal ambiguity. As Reed and DeFillippi argued in 1990, a state where "even the people themselves cannot explain well why they are winning" is difficult for other companies to imitate. (Note 2)

Do you see it? The act of making things explicit and structured in order to provide strengths to AI is precisely the act of stripping away that tacitness and ambiguity. Strengths that have been articulated and bundled together become easy to take out, leak to the outside, or be carried away by someone who quits. In other words, the more you organize strengths into a "form that the organization can utilize" for AI to use, the lower the difficulty of imitating those strengths becomes. The effort to strengthen your advantage ends up thinning the wall that protects that advantage. This is the paradox I am currently focusing on the most.

*1 Barney (1991) "Firm Resources and Sustained Competitive Advantage"
*2 Reed & DeFillippi (1990) "Causal Ambiguity, Barriers to Imitation, and Sustainable Competitive Advantage"

So, what protects our advantage?

Faced with this paradox, one might shrink back, wondering, "Should we not structure things at all?" However, the answer is not to retreat. It is to shift the locus of advantage to the right place. In conclusion, the advantage to protect is not the "neatly summarized strength" itself, but rather the power to keep updating it.

The argument presented by Dierickx and Cool in 1989 that assets are something to be accumulated (*3) comes into play here. They argued that valuable assets cannot be bought in the market but must be accumulated over time. That is precisely why, no matter how much competitors rush, they cannot shorten the time spent, making it difficult for them to catch up. The same applies directly to GTM context. Once a set of strengths is organized, if left alone, it will become obsolete and eventually be imitated. However, the mechanism itself that keeps updating strengths in a way deeply intertwined with one's own business can only be acquired through time. The advantage to protect shifts from the organized content itself to the power that continues to generate it.

Therefore, to connect inward-facing supply to competitive advantage, three conditions are required.

First, establish safeguards to prevent codified strengths from leaking out. Not using raw competitive comparison data directly in outward-facing communications—this kind of boundary is not just an internal rule, but essential for protecting the advantage itself.

Second, do not stop after creating it once; continue to cultivate the power to keep updating it.

Third, view whether it is being passed on effectively separately from the results. You cannot fix anything by looking back and saying, "Since we got results, the way we passed it on must have been good." You need to verify it with leading indicators, such as how well it is structured and how much the AI is using it.

As a starting point for actually putting these three into practice, I have prepared a GTM Context Package Template. It is divided into a main body that writes out positioning, ICP, competitive comparison, and messaging in a structure that AI can read, and a deep-dive sheet for holding details separately, embodying the separation of "storing" and "selecting" mentioned in the text. I have also included examples and a checklist to verify whether it is written, passed on, protected, and functioning. You can use it by duplicating it as a Notion page or by passing the entire Markdown set directly to the AI. I hope you will try it out and give me your feedback.

Rephrased for the field, this also connects to why external support is effective. Support that only summarizes and delivers strengths once may be useful at that moment, but it quickly becomes obsolete if updates stop. What works is to keep the cycle of organizing, passing on, and re-verifying strengths running without interruption. If the advantage lies in the "power to continue," then creating value also lies in staying close to that continuity.

*3 Dierickx & Cool (1989) "Asset Stock Accumulation and Sustainability of Competitive Advantage"

What to protect is not the strength, but the "power to continue"

The real reason GTM AI cannot deliver results is not the model or the budget, but the fact that the company's strengths have not been passed on in a form that the AI can read. However, there is a pitfall in the simple idea that you can win just by structuring and passing it on. Explicitly stating things can thin the walls of imitation. That is why the advantage resides not in the summarized strengths themselves, but in the power to keep them from leaking and to keep them updated.

First, please re-examine whether your company's "reason to win" is currently organized somewhere in a form that AI can read. Most of it is likely still in someone's head. The first step of translating that into structure is the starting point for finally turning GTM AI into a force.

About this series
"Velopont" is a series where Iijima, who has been connecting technology and users for many years, shares practical knowledge of PMM/GTM strategy combined with MBA learnings.

Download the GTM Context Package Template
The template set (Notion page + Markdown) introduced in the text for writing out your company's strengths in a form that AI can read can be downloaded here.

About Velopont LLC
Velopont LLC is a consulting firm led by Iijima that supports GTM strategy and product marketing. For those who have challenges with product or business growth, we accept free consultations. Please feel free to contact us.

Receive notifications for the latest articles
Follow on note / Subscribe to the newsletter (free)

Feel free to share your thoughts and questions
I look forward to hearing from you in the comments section, or on X / LinkedIn

#ProductMarketing #BtoBMarketing #MarketingStrategy #GTM #PMM #OrganizationalOptimization #GTMContext #ContextEngineering #SalesSupportAI #AIAgent #GenerativeAI #Positioning #CompetitiveAdvantage #VRIO #ResourceBasedView #Kotler #ICP #SalesEnablement #B2B #SaaS #BusinessGrowth #ManagementStrategy #AIUtilization #KnowledgeManagement #ProductManagement