---
title: "AI Consumer Researchers Are Emerging — Will Traditional Market Research Be Replaced?"
author: "Unique Research"
sourcePublication: "Unique Research Substack"
originalPublishedAt: "2026-07-15T11:51:03+00:00"
canonical: "https://ffcap.cn/en/research/src-20260715-04html"
source: "https://uniqueresearch.substack.com/p/src-20260715-04html"
language: "en"
---

# AI Consumer Researchers Are Emerging — Will Traditional Market Research Be Replaced?

[![](https://substackcdn.com/image/fetch/$s_!tznP!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F544994a7-4275-40d7-acda-cbf87d9a9370_2048x1152.png)](https://substackcdn.com/image/fetch/$s_!tznP!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F544994a7-4275-40d7-acda-cbf87d9a9370_2048x1152.png)

_Original · Unique Research · 2026-07-15_

_Editor's note: The interview and its judgments belong to the original Chinese author and the interviewee, Xiao Jian (肖建), partner at Honyuan Technology (宏原科技). This English rendition translates the full article in source order, including the opening essay, six analysis sections, and all twenty Q&A items. All company and person names are preserved as source attributions. Industry figures and client examples are speaker claims, not independently verified findings._

AI Industry Observation

When Every Brand Uses AI to Produce Content, Consumers Start Believing Nothing

"When the cost of manufacturing information approaches zero, the cost of trust approaches infinity."

AI is lowering content production costs at an unprecedented rate.

What used to take several people a week — copy, posters, short videos — now AI can produce dozens of versions in a day. Brands can chase trends faster, update social media more frequently, and batch-generate content for different platforms and demographics.

This is certainly progress.

But when every brand can infinitely produce content, another problem emerges: what do consumers actually believe anymore?

A seeding note might be AI-generated, a seemingly authentic product review might come from paid promotion, and comment sections contain both real users and water armies, influencers, and brand-manufactured voices. There is more and more information, but trustworthy information has not increased in step.

From the brand side, the problem is also changing.

In the past, brands worried about not having enough content to reach consumers; now, what brands really need to worry about may is saying a lot, but not a single sentence being believed by consumers.

Xiao Jian (肖建), partner at Honyuan Technology (宏原科技), has an understanding of AI marketing that differs from the mainstream. He believes the most worthwhile direction in AI marketing is not continuing to help brands "say more," but first helping brands "listen clearly."

Listen to what consumers actually believe, what they worry about, and what they want, then turn these real voices into verifiable growth actions.

In his view:

"When the cost of manufacturing information approaches zero, the cost of trust approaches infinity. This may be the most important sentence for understanding the next phase of AI marketing."

Brands Don't Lack Content — They Lack Why Consumers Should Believe Them

Currently, most products on the market called "AI marketing" focus on a few areas: generating copy, generating images, editing video, automated ad placement, and optimizing ad creatives.

These capabilities are certainly valuable, but they still solve the same problem: how to let brands produce information faster and in greater volume.

Xiao Jian uses a vivid analogy:

Understanding AI marketing as generating copy and images is like saying a phone is for making phone calls. You're not wrong, but you're missing the truly important part.

Because today's core problem for brands is no longer "not saying enough," but "no one believes what's being said."

When brands can batch-produce content with AI, consumers can likewise use AI to detect patterns, compare information, query ingredients, and verify reviews. The information advantage formed purely by content quantity becomes increasingly hard to sustain.

If you see this as a content arms race, brands likely won't truly win.

Honyuan Technology chose another path: let AI first "listen," then "judge," and finally "recommend."

First, from comment sections, social media, e-commerce reviews, and customer service records, listen to what consumers are truly anxious about, believe, and desire; then judge whether these voices are real signals or noise, which doubts are affecting purchases, and which appeal points have a chance to drive action; finally tell the brand what evidence to add this week, what messaging to adjust, and which competitor to watch.

This is not simply helping brands generate content, but answering a more fundamental question:

What should brands say for consumers to possibly believe it.

So generative AI is solving expression efficiency, while Marketing AI must solve judgment efficiency and decision quality.

The former makes marketers faster; the latter determines whether marketers are right.

What Consumers Say in Surveys May Be Completely Different From What They Think When They Pay

For brands to understand consumers, they have traditionally relied mainly on surveys, interviews, focus groups, and market research reports.

These methods haven't stopped working, but they are increasingly unable to fully capture real consumer decisions.

The reason is not just that markets change fast.

More importantly, when a consumer sits down to fill out a survey, what they give is often an organized, explained, and even beautified answer.

They may tell researchers they value product quality, brand reputation, and value for money. But when they're actually ready to place an order, what flashes through their mind may be a different sentence:

Would giving this to my mother-in-law look cheap?

What consumers say in formal research, discuss in focus groups, and what they actually think when they pay, or complain about in friend groups, are often not the same thing. Especially before entering a purchase scenario, consumers have usually already browsed Xiaohongshu, Douyin, and e-commerce reviews. Real doubts, trust, and desires are hidden more in these spontaneous discussions than in enterprise-designed survey options.

"Would giving this to my mother-in-law look cheap" is a typical example.

Literally, it involves "gift," "mother-in-law," and "cheap." Emotionally, it expresses worry.

But for a brand, this sentence contains far more information than that.

It means the product is being placed in the "elderly gifting" scenario; the consumer has a desire to show filial piety but worries the gift isn't dignified enough; the brand hasn't provided enough evidence to help her resolve the doubt; if a brand can prove its product is dignified and reputable, there's an opportunity to push her to complete the purchase.

General large models can understand the literal meaning of a sentence and identify emotion, but the truly difficult part is:

In the context of a complete consumer decision, what does this sentence actually mean?

Xiao Jian believes the biggest technical challenge in AI consumer insight is not scraping data, nor just semantic and sentiment judgment, but understanding real business scenarios.

Hearing what users say is only the first step. More critical is judging which purchase scenario this sentence belongs to, what desires and doubts it reveals, what evidence is missing, and what action the brand should take next.

This is also the clearest distinction between general language capability and vertical marketing capability.

General models understand language; vertical AI must further understand scenarios, trust, and business consequences.

Turning Consumer "Chatter" Into the Brand's "Operating Ledger"

Consumers' real expressions are usually very fragmented.

It might be a negative review, a bullet comment, a photo post, or a question that repeatedly appears in customer service conversations.

Alone, each piece of information looks like noise.

Traditional social media analysis might tell brands that consumers mention "price," "packaging," or "ingredients." But keyword statistics like this are still far from operating decisions.

"Nice packaging" and "would this look cheap for my mother-in-law" both involve packaging, but they are completely different kinds of signals.

The former is a general evaluation; the latter is a dignity doubt that may directly block purchase.

Honyuan divides this process into three steps.

The first step is to convert each consumer expression into a traceable semantic card.

Not just looking at what words the user mentioned, but judging who is saying it in what scenario, whether it expresses desire or doubt, whether the missing piece is evidence, or whether there is a product charm that can be amplified.

The second step is connecting fragments into paths.

A user might first save content on Xiaohongshu, then a few days later ask on an e-commerce platform whether the product works, then enter a livestream to place an order, and finally share an unboxing experience in a private group.

If these behaviors are scattered across different platforms, they're easily seen as four isolated actions. But from a user decision perspective, they may form a complete path: first attracted by the scenario, then developing doubts, completing purchase after finding evidence, and finally sharing based on experience.

The third step is aggregating many individual paths into market signals.

When a certain type of doubt keeps increasing, when a certain appeal point is repeatedly validated by competitors, when trust in a certain channel is declining, what brands see is no longer scattered reviews, but changes happening in the market.

These changes ultimately need to enter product, content, channel, and sales actions, not stay in a pretty research report.

The interview mentioned an example from a Dong'e Ejiao (东阿阿胶) project.

Traditional pharmaceutical marketing habitually emphasizes product efficacy like "strengthening the spleen and stomach" and "chronic disease treatment," but consumers discuss more about "takeout stomach villains" and "daily wellness routine" — everyday-life expressions.

The product itself hasn't changed, but there is a gap between product language and consumers' real lives.

After identifying this gap, what the brand needs to do is not necessarily modify the product, but respond to the scenarios and doubts consumers live in, in a way they can truly understand and relate to.

This is also the difference between Marketing AI and ordinary sentiment analysis.

It's not compressing ten thousand reviews into five summaries, but translating consumers' "chatter" into "operating language" brands can use.

Content Marketing May Be Where Marketing AI Creates Value First

Marketing AI can enter many areas.

It can help enterprises judge product demand, formulate launch strategies, choose channels, and analyze KOL-brand fit.

But Xiao Jian believes the scenario where Marketing AI most easily demonstrates value is still content marketing.

The reason is not that content is most important, but that content is closest to consumer action and easiest to verify.

Suppose AI finds that consumers repeatedly ask before buying a product:

Is this ingredient safe for pregnant women?

The enterprise doesn't need to wait for product re-development, or immediately adjust the entire channel system. It can add the relevant evidence on the detail page, in social media content, and in customer service scripts the same day, then observe changes through save rates, cart rates, and conversion rates.

This feedback loop can be as short as one week.

In the past, the starting point for brand content was usually "what I want to say." Even when using AI to generate copy, it just completes "what I want to say" faster.

Marketing AI changes the starting point of content.

It shifts brands from "what I want to express" to "why don't consumers believe yet."

Content then becomes not just a communication vehicle, but a trust system designed around doubts and evidence.

Consumers worry about safety — the brand provides safety evidence; consumers worry about efficacy — the brand shows verifiable results; consumers don't know what scenarios the product fits — the brand needs to explain usage scenarios clearly.

Xiao Jian summarizes this change as: content shifting from creative-driven to doubt-driven and charm-driven.

Content no longer pursues just being pretty, lively, or emotionally valuable, but gradually builds trust along consumers' decision barriers.

AI Fishes Out Signals — Humans Decide Whether to Believe and Use

AI finding consumer signals doesn't mean enterprises can directly make decisions from them.

From a data insight to actual business action, there's still a deep "trust gap" in between.

For example, AI finds that negative discussions about a brand are increasing on a platform.

But is this genuine consumer dissatisfaction, or competitor-planted water army? Is it a widespread trust crisis, or an extreme case amplified by platform algorithms?

Enterprises need to go back to the original content and cross-validate with other platforms, sales data, and customer service feedback.

Even if the signal is real, enterprises still need to judge what it means.

Users frequently saying "too expensive" might mean the price is genuinely too high, or the brand hasn't clearly communicated value, or a competitor just started a promotion that changed consumers' psychological reference point.

The same word, in different business contexts, may point to completely different actions.

Further along, there's an even harder question: does the data-supported opportunity align with the brand's long-term direction?

AI might find that young women have demand for high-end tonics in "self-reward" scenarios, suggesting the brand emphasize light luxury expression.

But if this brand has long wanted to build a "professional, rigorous, medical-grade" image, light luxury content might bring short-term traffic while diluting long-term brand equity.

This tradeoff can't be made by a model alone.

The organization still needs to decide how many resources to invest, how long to validate, which demographics and channels to choose, and where the stop-loss line is if the experiment fails.

Finally, even if the recommendation logic is correct, consider whether the organization has the ability to execute.

Can the content team produce the corresponding content? Does the channel pricing system cooperate? Can sales and customer service scripts be adjusted synchronously? Do departments share common goals?

So Marketing AI doesn't automatically make decisions for enterprises; it makes the judgment process that used to be hidden in experience more transparent, verifiable, and reviewable.

AI is responsible for fishing signals out of noise; humans are responsible for turning signals into decisions.

AI approaches reality; humans bear judgment and responsibility.

Many Bosses Need Not AI, But "AI That Proves Them Right"

Talking about the difficulties of enterprises adopting Marketing AI, the first things that come to mind are data, models, and budget.

But Xiao Jian's judgment is more direct:

The biggest challenge is not technology, but organizational cognitive misalignment and resistance to change.

The first layer of resistance comes from decision-makers.

The CEO may repeatedly emphasize the importance of AI, but when AI's judgment conflicts with their own experience, they still trust the executives who have followed them for years.

When AI and the boss agree, the system is considered very accurate; when they disagree, it's usually the AI that is first questioned.

Behind this is not an algorithm problem, but a power problem.

"Whether an enterprise is truly willing to use AI depends on whether the CEO can accept being corrected by data. Many bosses verbally need AI, but what they truly need in their hearts is an AI that proves them right."

The second layer of resistance comes from execution teams.

When AI starts recording the basis, process, and results of every action, much previously ambiguous work will be put under the spotlight.

Brand managers may worry that AI exposes logic gaps in past decisions; content teams worry about being replaced; operations staff may not be used to every action requiring documented justification.

On the surface, they'll raise issues about system inaccuracy, inconvenient operation, and incomplete data; but the deeper reason may be:

People aren't used to being constrained by data.

The third misalignment is treating AI like an outsourcing company.

Many enterprises hope that after connecting AI, consumer insights, content strategy, and ad optimization can all be done automatically, and the team just waits for results.

But a true AI decision system doesn't mean enterprises can stop thinking.

On the contrary, it requires enterprises to explain more clearly why they made a decision, how to verify it, and who is responsible for the results.

AI doesn't make organizations more relaxed; it forces them to be more honest and rigorous.

This also explains why many enterprises have data middle platforms, marketing middle platforms, and numerous dashboards, yet still rely on experience and intuition for decisions.

Past middle platforms solved "where is the data," but not "what does the data mean."

After AI enters, three changes may occur.

First, from people finding data to data finding people. AI continuously monitors anomalous signals and proactively tells business teams what problems are emerging.

Second, from looking at metrics to looking at relationships. Enterprises don't just see conversion rates dropping; they see their relationship with negative reviews, competitor discounts, channel pricing, and customer service feedback.

Third, from reviewing the past to guiding the next step. Data is no longer just a rearview mirror, but starts becoming a navigation system.

In Xiao Jian's words:

Middle platforms connect the data; AI makes it alive. The middle platform is the warehouse; AI is the worker.

Without workers, no matter how much is piled in the warehouse, it's hard to turn into business.

The Stronger the Large Model, The More Vertical AI Must Answer "Who Is Responsible"

As large model capabilities keep improving, whether application-layer companies will be replaced by model vendors is a question all vertical AI companies face.

Xiao Jian believes the long-term barrier for Marketing AI companies is not the model itself.

Large models can understand language, summarize reviews, judge sentiment, and generate a complete marketing recommendation.

But it still struggles to independently answer three questions.

First, is this information true?

Among a hundred positive reviews, which come from real consumers, which are commercial content, and which are water armies or AI-generated content? If the data itself is polluted, all subsequent decisions lose their foundation.

Second, is this information important?

"Packaging is decent" and "would this look cheap for my mother-in-law" are both user feedback for a large model, but for brand growth, the latter is obviously closer to a purchase barrier.

This kind of commercial weight judgment depends on industry corpora, knowledge graphs, and long-term project experience.

Third, after this recommendation is used, who is responsible?

General models can give answers, but won't take responsibility for brand growth results. Real business decisions must clarify the boundary between AI and humans: AI is responsible for observation and recommendations; humans are responsible for judgment, decisions, and consequences.

Therefore, what's truly hard to replicate in vertical AI is not a particular algorithm, but the combined capability built over time:

Identification of real user voices, industry knowledge graphs, business scenario understanding, client business context, and the ability to embed into enterprises' daily decision-making processes.

When a brand starts using the same language to discuss consumers, reviews market signals at a fixed rhythm, and continuously deposits historical judgments into the system, AI is no longer an isolated tool but gradually becomes part of how the enterprise grows.

Model vendors make AI smarter; vertical AI companies need to make it trustworthy, usable, and accountable within specific industries.

These two are more like a division of labor than simple replacement.

Future Marketing May Become a Discipline of "Trust Engineering"

Looking at the marketing industry over the next three years, Xiao Jian offers three judgments.

The first change is that trust will be engineered.

In the past, trust was usually considered a long-term accumulated brand asset, hard to measure accurately and hard to decompose.

But after AI intervenes, enterprises can continuously observe whether consumer doubts are decreasing, whether brand evidence is being accepted, whether content is driving search and purchase, and whether product experience is delivering on brand promises.

Trust is no longer just "I feel consumers trust us more," but gradually becomes a system that can be observed, verified, and improved.

The second change is that marketing organizations will move from functional division to system coordination.

Brand, e-commerce, content, media, and sales departments can no longer only look at their own KPIs.

Because content lacking evidence may affect e-commerce conversion; channel price chaos may damage brand trust; product experience may invalidate earlier ad spend.

When consumer voices, product delivery, channel performance, and operating results are connected, department walls become harder to maintain.

The third change is that the CMO role may be redefined.

Over the past decade, media buying has increasingly been controlled by platform algorithms; creative production has been dispersed among influencers, vendors, and AI; growth responsibility has been split across e-commerce, private domain, and sales departments. CMOs in many enterprises have gradually become managers of communication and budget.

But if understanding consumers, building trust, and driving growth start to become a system, the CMO may again become the chief architect of that system.

He not only needs to understand communications, but also data, technology, organization, and human decision-making.

Future marketing teams may have AI consumer researchers, evidence gap detectors, content strategy advisors, growth review auditors, and risk early-warning agents.

But this doesn't mean a few robots will replace the entire marketing department.

The more likely form is that every marketer has a set of AI roles alongside them: AI continuously listens, organizes, and alerts; humans are responsible for judging trends, making tradeoffs, coordinating the organization, and creating content that truly moves people.

AI won't end marketing.

What it truly eliminates may be the marketing approach that relies on information asymmetry, content bombing, and gut-feeling experience.

As content production costs keep falling, the truly expensive brand capability will no longer be saying more, but hearing accurately; not manufacturing more information, but identifying real signals; not telling a story by feel, but continuously proving through action that it deserves to be believed.

So what Marketing AI ultimately needs to do is not just help enterprises improve efficiency.

It needs to, in real scenarios, hear real users expressing real tasks in their own language, identify the action signals emerging within them, and then through content, product, and channel actions, verify whether these judgments truly bring growth.

When every brand can use AI to produce content, the end point of competition won't be who produces more.

It will be who is closest to real consumers, who sees their doubts earliest, and who can continuously produce credible evidence.

"From this perspective, the next phase of AI marketing may not be a content production revolution. It is a trust reconstruction."

Selected Interview Q&A

Q1: Without using an official introduction, how would you explain what Honyuan Technology does?

Xiao Jian: We help brands figure out what consumers truly believe, fear, and want, then turn these real voices into verifiable business growth, and deposit them into a continuously running business system.

Simply put, we don't just help brands produce more content; we help brands understand consumers more accurately and make growth decisions based on that understanding.

Q2: Why are traditional surveys, interviews, and focus groups increasingly unable to meet enterprise needs?

Xiao Jian: The biggest problem with traditional methods is that they all ask consumers "what do you think."

But what consumers say in surveys, discuss in focus groups, and what they actually think when paying or complain about in friend groups are often not the same thing.

Today's consumers, before entering a purchase scenario, have already browsed Xiaohongshu, Douyin, and e-commerce review sections. Their real doubts, trust, and desires are hidden more in these spontaneous discussions than in enterprise-designed survey options.

So it's not that traditional research is wrong, but that it's hard to capture the unadorned consumer decision process in the real world.

Q3: Entering from Marketing AI, what problem in the marketing chain does Honyuan most want to solve?

Xiao Jian: There's a long-neglected fault line in the marketing chain.

On one side is massive consumer voice and business data; on the other is what the brand should do next. In the past, the two were mainly connected by human experience and intuition. But today there's so much data that people can't look at it all, let alone see it accurately.

What we want to solve is the conversion gap from data to trust, and from trust to decisions.

AI doesn't just write copy for brands; it identifies which of thousands of real user voices are genuine doubts, which are credible evidence, and which are charm points that can drive action, then turns them into weekly verifiable and reviewable growth actions.

In one sentence: use AI to turn "understanding consumers" from a craft into a runnable system.

Q4: Many people think AI marketing is just generating copy, images, and ad creatives — what do you think?

Xiao Jian: This understanding isn't wrong, but it's like saying "a phone is for making phone calls" — correct, but incomplete.

AI generating copy, images, and video solves the brand's "say more" problem. But today's biggest dilemma for brands is not saying enough, but saying things no one believes.

When the cost of manufacturing information approaches zero, the cost of trust approaches infinity. Brands batch-produce content with AI; consumers also use AI to detect patterns. In this content arms race, brands can't truly win.

We choose to let AI first "listen" to what consumers are truly anxious about, believe, and desire; then "judge" which voices are real and which doubts are unaddressed; finally "recommend" what evidence brands should add this week and what expressions to adjust.

Generating content is only the surface of AI marketing; what truly matters is helping brands re-understand people and rebuild trust.

Q5: Compared with traditional market research, what's the biggest change in AI consumer insight?

Xiao Jian: Traditional research is "I ask, you answer"; AI consumer insight is "I'm always listening to you."

When designing a survey, the enterprise has already preset which dimensions are important, and consumers can only answer within that framework.

But in the real world, consumers use their own language, in their own life scenarios, expressing doubts and desires to people they trust.

AI's value is being able to continuously and at scale understand these unadorned real expressions.

Q6: Finding truly valuable information from massive consumer voices — what's the biggest technical challenge?

Xiao Jian: The biggest challenge is not data acquisition, nor just semantic understanding and sentiment judgment, but understanding real business scenarios.

For example, a user says: "Would this look cheap if I give it to my mother-in-law?"

The model can identify "gift," "mother-in-law," "cheap," and judge that she's worried. But for the brand, what does this sentence mean?

It means the product has entered the elderly gifting scenario; the consumer wants to show filial piety but worries it's not dignified enough; the brand lacks evidence that would reassure her; if reputation and dignity can resolve the doubt, it may drive purchase.

The truly difficult part isn't hearing what the user said, but placing this sentence in the complete picture of brand growth, judging which link it belongs to, what it means, and what to do next.

Q7: How does AI turn a complaint or review into business insight a brand can use?

Xiao Jian: Roughly three steps.

The first is turning fragments into semantic cards. Not simply judging whether the user mentioned price or packaging, but identifying who is saying it in what scenario, what desire or doubt it expresses, and which evidence gap it exposes.

The second is connecting fragments into paths. A consumer might first save a Xiaohongshu note, then go to an e-commerce platform to ask about efficacy, then enter a livestream to order, and finally share in a private group. These are not four isolated actions, but a decision path from interest, through doubt, to building trust and completing purchase.

The third is aggregating many paths into growth signals. Which doubts are rising? Which charm points have been validated by competitors? Which channel's trust is declining?

AI's value is not summarizing information into a pretty report, but turning consumers' "chatter" into the brand's "operating ledger."

Q8: Will every enterprise have an "AI consumer researcher" in the future?

Xiao Jian: I think yes, and it will become part of enterprise infrastructure, like financial systems and CRM.

It can listen 24/7, never missing an emotional late-night negative review, a sudden topic in a mid-tier influencer's comments, or a small-scenario discussion with low volume but signaling new demand.

More importantly, it doesn't do simple keyword monitoring, but listens with a business framework. It knows "would this look cheap for my mother-in-law" isn't ordinary price sensitivity, but a dignity doubt in a gifting scenario.

Its value is not just reviewing the past, but guiding the next step: what new doubts emerged this week, what evidence should the brand add, and how effective was the action.

It won't replace researchers, but frees people from information filtering so they can judge what's a trend, which opportunities are worth pursuing, and what content truly moves people.

Q9: In which scenario does Marketing AI most easily demonstrate value?

Xiao Jian: If I could only choose one, I'd choose content marketing.

Product R&D, new product launches, channel selection, and KOL matching are all important, but content is the shortest path connecting "understanding consumers" and "driving consumer action."

For example, consumers repeatedly ask: "Is this ingredient safe for pregnant women?"

The brand doesn't need to wait for product re-development; it can add evidence on detail pages, social media content, and customer service scripts that same day, then observe results through saves, cart adds, and conversion. This loop might complete in a week.

In the past, brands started content from "what I want to say"; even with AI-generated copy, it's just expressing yourself faster.

Marketing AI turns content toward "why don't consumers believe yet." Content thus transforms from a communication vehicle into a trust engine.

Q10: How do AI-generated marketing insights actually become business decisions?

Xiao Jian: From insight to decision, there are at least five layers of human judgment.

First, is this insight credible? Is it a real consumer signal, or distortion caused by water armies, extreme cases, or platform algorithms?

Second, what does this signal mean? Users saying "too expensive" could be a pricing problem, unclear value communication, or a competitor promotion.

Third, does it align with the brand's long-term direction? A short-term traffic opportunity may dilute long-term brand equity.

Fourth, how big of a cut are we willing to use to validate? What's the validation period, resource investment, success criteria, and stop-loss line?

Fifth, is the organization ready? Can content, channels, sales, and customer service execute together?

AI fishes signals out of noise; humans turn signals into decisions. AI approaches reality; humans bear judgment and responsibility.

Q11: How will AI change enterprise data middle platforms and marketing middle platforms?

Xiao Jian: In the past, many enterprises spent heavily building data middle platforms, solving "where is the data" but not "what does the data mean."

After AI enters, three changes come.

First, from "people finding data" to "data finding people." AI doesn't wait for business users to ask questions; it proactively discovers anomalous signals and pushes them.

Second, from "looking at metrics" to "looking at relationships." Enterprises don't just see conversion rates dropping; they see relationships with negative reviews, competitor discounts, channel pricing, and customer service feedback.

Third, from "reviewing the past" to "guiding the next step." Data is no longer just a rearview mirror but starts becoming a navigation system, directly triggering the next round of observation, judgment, action, and review.

Middle platforms connect data; AI makes data alive. The middle platform is the warehouse; AI is the worker. Without workers, no matter how much inventory is in the warehouse, it can't become business.

Q12: What's the biggest challenge for enterprises introducing Marketing AI?

Xiao Jian: The biggest challenge is not technology, data, or budget, but organizational cognitive misalignment and resistance to change.

The decision layer's problem is "Lord Ye's love of dragons." The boss says AI is important, but when AI's judgment conflicts with their experience, they usually suspect AI first. Many CEOs need not AI that truly challenges them, but AI that proves them right.

The execution layer's problem is fear. AI records the basis, process, and result of every action; problems previously covered by experience and vague expression get exposed, and many people aren't used to being constrained by data.

The third problem is treating AI as an outsourcer. Enterprises want AI to finish all the work while they just wait for results. But AI is more like augmented intelligence — it organizes evidence and makes recommendations; humans still need to judge, decide, and take responsibility.

What enterprises truly need to prepare is to accept being corrected by real market voices.

Q13: As large models get stronger, what is the long-term barrier for Marketing AI companies?

Xiao Jian: Large models solve general understanding and generation capability, but they struggle to independently solve three problems.

First, it doesn't know what's real. Among a hundred positive reviews, which are real consumers, which are commercial content, water armies, or AI-generated?

Second, it doesn't know what's important. It can summarize a thousand reviews, but may not judge which signal truly affects brand growth.

Third, it can't take responsibility. Large models can give recommendations, but after the brand actually adopts them, the business consequences still need humans and the enterprise to bear.

So Marketing AI's barrier is not the model itself, but industry corpora and knowledge graphs, anti-distortion and evidence governance capabilities, and long-deposited client business context and decision paths.

The stronger the large model, the happier we are. Model vendors make AI smarter; we make brands more trustworthy.

Q14: For Honyuan Technology, what is the truly hard-to-replicate capability?

Xiao Jian: Not one model, nor a single technology, but the time compounding formed by multiple capabilities layered together.

First is anti-distortion capability. Identifying water armies, distinguishing real UGC from commercial content, judging whether a signal is noise or trend — these require repeated accumulation in real projects.

Second is industry knowledge graphs. "Safety," "trust," and "efficacy" mean completely different things in different industries and must be gradually deposited through real user language, operating data, and project experience.

Third is client context and decision paths. When clients start using the same language to discuss consumers, review at a fixed rhythm, and deposit historical judgments in the system, they're no longer just using a tool but changing how they make growth decisions.

This barrier is not simple addition, but multiplication of industry corpora, anti-distortion, knowledge graphs, organizational processes, and client decision memory.

Q15: What's the biggest difference between vertical AI and general large models?

Xiao Jian: General large models understand language; vertical AI understands scenarios, trust, and decision responsibility.

General AI can read "this product is terrible," but may not know whether it comes from a real user, competitor water army, or an extreme case amplified by the platform.

It can summarize user feedback, but may not know that "would this look cheap for my mother-in-law" is closer to a purchase barrier than "packaging is decent."

It can also give logically correct marketing recommendations, but that recommendation may not align with the brand's long-term direction.

Most importantly, general models don't take responsibility for business results.

General AI gives answers; vertical AI needs to place the answer within complete business logic, telling the enterprise whether the answer is credible, important, aligned with brand direction, and who is responsible after adoption.

Q16: Over the next three years, what's the biggest change in the marketing industry?

Xiao Jian: I'd use three keywords: trust engineering, organizational systematization, and the redefinition of the marketer's identity.

First, trust goes from being a byproduct of marketing to being the core deliverable. Brands need to continuously discover consumer doubts, provide credible evidence, and verify whether trust truly improves through behavior and operating results.

Second, marketing organizations move from functional silos to system coordination. Brand, content, e-commerce, media, and channels can no longer operate independently, because one channel's price chaos can damage the entire brand's trust equity.

Third, the CMO role is redefined. Over the past decade CMOs have increasingly looked like communication procurement managers; in the future they may become chief architects of consumer insight, trust building, and growth systems.

Marketing gradually shifts from an art driven by creativity and budget to an engineering driven by trust and systems.

Q17: What AI roles will appear in a brand's future AI marketing team?

Xiao Jian: First, an AI consumer researcher, responsible for continuously listening to consumer voices and pushing market changes to decision-makers.

Second, a competitor and substitute threat tracker. It watches not just traditional competitors, but all solutions that might replace the brand's position in users' lives.

Third, an evidence gap detector, specifically looking for questions consumers repeatedly ask but the brand hasn't effectively addressed.

Fourth, a content strategy advisor. It won't just write copy, but judge what content to create this week based on consumer doubts and evidence gaps.

Additionally, there will be GTM rhythm advisors, growth review auditors, and risk early-warning sentinels.

But people won't disappear. People calibrate AI, judge signals, make strategic tradeoffs, create content that truly moves people, and handle interest coordination within the organization.

The future marketing team is not a few robots replacing a group of people, but a group of people with AI partners.

Q18: Will AI widen the efficiency gap between brands?

Xiao Jian: Yes, and the gap will be brutal.

But the real gap is not between using AI and not using AI, but between using AI to understand consumers and only using AI to generate content.

The first layer is content production efficiency. AI lets a team go from ten pieces of content a week to a hundred. But once everyone can do it, this advantage quickly flattens.

The second layer is decision efficiency. Brands not using AI do research and make decisions quarterly; brands truly using AI well can discover problems, test actions, and observe feedback weekly.

The third layer is cognitive efficiency. Excellent brands treat AI as an external brain for continuous cognitive calibration, not a one-time tool. The longer it goes, the closer their understanding of the market approaches reality; brands still relying on a few executives' experience may increasingly disconnect from consumers.

Q19: In current industrial application-layer AI startups, where is the biggest opportunity?

Xiao Jian: The biggest opportunity is using AI to solve problems that were previously too complex, too tacit, and too people-dependent to be systematized.

The first opportunity is turning tacit knowledge into runnable systems. Every industry has a few "old masters" who can judge water armies, identify risks, and spot opportunities by experience, but this capability is hard to replicate. The value of AI startups is externalizing these tacit judgments into frameworks and systems.

The second opportunity is handing weekly repetitive information collection and preliminary judgment to AI, so humans can do the strategic judgment and creation that only humans can do.

The third, and biggest, opportunity is rebuilding business logic around trust.

When information manufacturing cost approaches zero, trust becomes the scarcest commercial resource. Whoever can use AI to help enterprises build, manage, and verify trust captures a more fundamental need of this era.

Q20: How should traditional enterprises start embracing Marketing AI now?

Xiao Jian: The first step is not buying a tool, but starting from the boss, being ready to accept being challenged by real market voices.

AI may tell you facts you don't want to hear: the brand in consumers' eyes isn't what you imagine; the real competitor isn't the one you've been watching; the product selling point you think is most important may be meaningless to consumers.

When these voices are laid out, does the enterprise face them, or is the first reaction "the data is wrong"? This choice determines whether AI truly enters the business, or becomes another expensive ornament.

Before purchasing an AI system, enterprises can do one simple thing: pull together e-commerce reviews and social media discussions from the past month, and sit down with the team to read them carefully.

See what consumers are worried about, comparing, and desiring.

This action doesn't need AI, but it helps the organization build a reverence for real voices. With that habit, using AI is adding wings; without it, no amount of AI is just Lord Ye loving dragons.

---

Original publication: https://uniqueresearch.substack.com/p/src-20260715-04html
On-site reading page: https://ffcap.cn/en/research/src-20260715-04html