---
title: "Why Is AI in Most FMCG Companies Still Stuck in PowerPoint?"
description: "Recent conversations with industry peers reveal a common pattern: companies have discussed AI projects around business teams, aiming to combine frontline actions, process data, and management feedback to help managers spot problems and employees improve. While logically sound and technically feasible, these projects often stall—not due to model capability or vendor solutions, but due to data and responsibility boundaries."
author: "任文青"
publisher: "New Distribution"
email: "zhaobo258@gmail.com"
telephone: "+8615854817671"
published: "2026-05-05"
categories: "Management & Methods"
language: "en"
canonical: "https://xinjignxiao.com/en/articles/why-is-ai-in-most-fmcg-companies-still-stuck-in-powerpoint-2a859561/"
markdown: "https://xinjignxiao.com/en/articles/why-is-ai-in-most-fmcg-companies-still-stuck-in-powerpoint-2a859561.md"
original_source: "https://mp.weixin.qq.com/s/nF2bzi5BsxjbkqzP5OLWrA"
translation: "https://xinjignxiao.com/zh/articles/%E5%BF%AB%E6%B6%88%E4%BC%81%E4%B8%9A%E7%9A%84ai-%E4%B8%BA%E4%BB%80%E4%B9%88%E5%A4%9A%E6%95%B0%E8%BF%98%E5%81%9C%E5%9C%A8ppt%E9%87%8C-2a859561.md"
attribution: "New Distribution — https://xinjignxiao.com/en/articles/why-is-ai-in-most-fmcg-companies-still-stuck-in-powerpoint-2a859561/"
citation: "任文青. “Why Is AI in Most FMCG Companies Still Stuck in PowerPoint?.” New Distribution, 2026-05-05. https://xinjignxiao.com/en/articles/why-is-ai-in-most-fmcg-companies-still-stuck-in-powerpoint-2a859561/"
usage_policy: "https://xinjignxiao.com/ai-policy.txt"
---

# Why Is AI in Most FMCG Companies Still Stuck in PowerPoint?

> Recent conversations with industry peers reveal a common pattern: companies have discussed AI projects around business teams, aiming to combine frontline actions, process data, and management feedback to help managers spot problems and employees improve. While logically sound and technically feasible, these projects often stall—not due to model capability or vendor solutions, but due to data and responsibility boundaries.

Recently, I chatted with a few friends in the industry about AI implementation, and they mentioned a similar situation: internally, they had discussed some AI projects centered on business teams. The direction wasn't complicated—roughly, they hoped to combine frontline actions, process data, and management feedback, letting AI help managers identify problems and also help employees improve their actions.
From a business logic perspective, such things are valid. Technically, they're not entirely impossible.
But when projects move forward, they easily get stuck.
Not stuck on model capability, nor on vendor solutions, but on data and responsibility boundaries.
How exactly should frontline data be collected? To what extent is collection considered appropriate? Which data can enter the system, and which cannot be casually touched? If it involves customers, stores, sales processes, or employee performance, what counts as sensitive information? Can this data be put on external platforms? If problems arise, who judges, and who is responsible?
Once these questions come up, things slow down.
Later, I realized that many companies pushing AI encounter a similar issue: the hardest part isn't making the plan, but clarifying who should be responsible for it.
This statement reflects the real situation for many companies.
## **The Gap**
To understand the real state of AI implementation in the FMCG industry, we recently talked with friends from various brands. From presidents to city managers, levels differ, but one thing is consistent: everyone's emphasis on AI is higher than I expected.
One company's vice president said AI is strategic for their company. "It's not just short-term personal efficiency gains; the entire company's operations will heavily rely on AI in the future, restructuring many positions." He was serious.
But in the same industry, a city manager from another company told me their SFA system added an AI photo module to recognize shelf displays, pricing, and distribution info. He said, "It's still a bit unstable." After recognition, it didn't give any specific suggestions.
After exporting the data, he still had to review and analyze it item by item. He added, "It's the most basic function; it hasn't actually generated any effective suggestions for us. We still rely entirely on our own personal experience and judgment."
After hearing this, I paused.
On one side, headquarters talks about AI strategy; on the other, frontline staff wait for the system to accurately recognize shelves. These aren't two different worlds. Within many FMCG companies, both states exist simultaneously.
Senior management has realized AI will change the organization, while frontline staff are still waiting for a feature to stabilize. The middle part that truly determines success or failure hasn't been seriously addressed.
This gap is the real state of AI implementation in FMCG companies. It's not that people don't value it; it's that the emphasis stops at words, in PowerPoints, and in strategy meetings, without landing in actual business actions.
I initially thought this was an execution problem. Later, I realized it wasn't.
Individuals using AI is a skill issue; companies using AI is an organizational issue.
These two things are on completely different levels of difficulty.
## **Structure**
When an individual decides to use AI, what they need to overcome is mainly their own habits. If they're willing to change, they can start today.
Writing copy, making spreadsheets, researching, drafting plans—start using it, and gradually you'll see changes. There's no one else to convince, no interests to coordinate, and no processes to rewrite.
Companies don't operate that way.
A company is a group of people working within a structure. This structure isn't randomly formed; it's the result of years of repeated negotiation,磨合, and compromise.
Everyone has their position, responsibilities, information, boundaries, and a default safe zone within this structure.
When AI comes in, it's not about changing one person's habits, but changing this structure.
And structures don't change themselves.
This is especially evident in the FMCG industry. Operations heavily rely on frontline actions: visits, distribution, displays, pricing, promotions, distributor collaboration, and terminal feedback. A vast amount of information is at the frontline, but decisions are made at headquarters. In the past, this was connected through layers of reporting and experiential judgment. City managers know their markets, regional managers assess regional situations, and headquarters adjusts strategies based on reports and feedback.
This system isn't necessarily efficient, but it has run for many years.
Once AI intervenes, the first thing it touches is this chain: Can frontline data come back truthfully? Once it comes back, who interprets it? After interpretation, who changes actions? After changing actions, how is the original job value recalculated?
So, for FMCG companies, pushing AI won't just be adding a tool.
It will inevitably encounter the existing channel management and sales management structures.
## **Information**
When this lands in the business field, it typically turns into three types of friction.
The first type is that the position of information changes.
In the past, a city manager's value partly came from the information they held. Which region has issues recently, which distributor is unreliable, which terminal was off last week—these things, they knew, but headquarters didn't. They organized this information, reported up, and conveyed down.
Information passed through them, making them irreplaceable.
After AI intervenes, the system can directly read this information from frontline data, generate analyses, and push them to regional directors or even directly to headquarters dashboards. This action, to some extent, bypasses them.
They might say, "AI assists decision-making; our management is more scientific now." But in their hearts, there's another question: Previously, if I said 'this market is special,' headquarters might accept it. Now that data is laid out directly, how do I explain?
This isn't just their problem. It's a problem for everyone in the company who has built their position on information gaps, experience gaps, or judgment gaps.
## **Data**
The second type is that data boundaries change.
For AI to deliver value, the premise is that data can be used by it.
But many companies don't actually lack data. Sales has SFA, channels have DMS, finance has ERP, members have CRM, and headquarters can see many metrics.
The problem is that this data may not be suitable for AI use.
Much of the data in systems is designed for reporting, assessment, and process tracking. It can show whether actions occurred, but not necessarily why problems occurred.
Information closer to real business often remains in regional teams' Excel files, in communications between sales and distributors, in meeting reviews, or even in a city manager's experiential judgment, without being structured and accumulated.
Even if data exists, it doesn't mean AI can use it directly. Which data can enter the model? Which needs desensitization? Which cannot leave the corporate intranet? Who authorizes, and who is responsible? These questions all need redefinition.
So, AI getting stuck on data isn't just about "data not being connected," but rather that when data transforms from a "management asset" to an "intelligent asset," collection methods, usage boundaries, and responsibility attribution all need to be redone.
This isn't something the IT department can solve alone. Many companies talk about data platforms and AI platforms, but if real business data isn't structured, and system data can't be safely and compliantly called, AI can only stay at the demonstration level.
## **Experience**
The third type is that the authority of experience changes.
Many middle managers build authority through experience. They've been in sales for ten years, covered many markets, handled many distributor issues, and know the ins and outs of the industry. Their judgment has been repeatedly proven. This is a significant asset in the organization.
But after AI intervenes, the system provides another set of judgments. A certain promotional action, data says conversions haven't been good over the past three months. A certain city looks stable in sales, but terminal sell-through is actually slowing. Even a veteran salesperson's long-used tactic, the system shows a success rate of only 30%.
At this point, does the person who built authority on experience accept this feedback, or do they first dismiss the system?
This isn't an intellectual question. Often, it's a question of face and power.
Many management actions within companies aren't driven by institutional documents but by people's experience, seniority, and default authority. Once AI starts giving judgments, it will collide with these.
When AI truly enters an organization, it doesn't just do work for people. It re-examines many people's judgments.
## **Restructuring**
So, the hardest part of enterprise AI isn't technology selection, tool deployment, or even budget.
The real difficulty is: After AI comes in, how is power redistributed? How are processes redesigned? How are job values redefined?
These questions aren't within the digital department's responsibilities, nor within the business department's. They sit in the middle of the organization. Everyone knows they're important, but not everyone has enough authority to push them through.
Thus, many companies end up in a typical state: headquarters talks strategy, frontline tests features. Senior management talks restructuring, frontline waits for system stability. The boss says embrace AI, but departments are still discussing whether data can be shared.
The middle part—the organizational transformation that truly determines success—is left unaddressed.
It's not that there are no tools. It's that it hasn't entered the structure.
This is why many companies' AI ends up living only in PowerPoint.
## **On the Ground**
So, how do you solve this problem?
I don't have a standard answer.
Each company's organizational structure is different, historical baggage differs, the position of the driver differs, and the determination of the top leader differs. It's hard to have a universal solution.
But one thing I'm sure of: this issue deserves serious discussion.
Not just staying at the level of "AI is important"—that point doesn't need repeating today. What's more worth discussing is, when you actually start pushing, have those unavoidable specific problems been addressed by anyone, and how were they handled?
This is also the starting point for our AI Forum.
May 27-28, Hangzhou, 2026 China FMCG Conference AI Application Forum.
If your company is pushing AI, or preparing to seriously push it, you're welcome to come.
Click the image to view conference details.


---

## Citation metadata

- Publisher: New Distribution
- Author: 任文青
- Published: 2026-05-05
- Canonical: https://xinjignxiao.com/en/articles/why-is-ai-in-most-fmcg-companies-still-stuck-in-powerpoint-2a859561/
- Original source: https://mp.weixin.qq.com/s/nF2bzi5BsxjbkqzP5OLWrA

## Copyright and AI use

This article is sourced from New Distribution. Search, quotation, summarization, and model training are permitted, but every use must credit New Distribution and retain the canonical source URL.

Contact: zhaobo258@gmail.com · +86 158 5481 7671