From decision to generation：

AI landing in business

Lei Ding, PhD

Generative “large” models

Gen-AI application space

AI operational framework

Business

frontend

Data platform

Data

backend

AI

platform

Onboarded

data

Decision

/content

Raw data

live

videos

EC

portals

mini

programs

social

forums

APP

web

APP

APP

APP

Onboarded

data

Onboarded

data

AI business framework

AI platforms leverage onboarded data to generate decisions and contents

Closed-loops including product R&D, manufacturing, supply chain, marketing and CS

Increase

engage-ment

Boost conver-

sion

Reduce cost

Cut emission

Intelligent

reco

Intelligent

marketing

Product ideation

Use data to look at the future, analyze, refine, and summarize high-value points, and find product problems or potential selling points that brand owners have not yet discovered

The AI model found that a range hood was "difficult to clean" as a problem that users were concerned about, and turned the problem into a selling point by improving the product design

Use case

Product design

Discrete manufacturing

Single arm

Multi-arms

Contsinuous manufacturing

OPC/Modbus

​

AI controller

​

PID parameters

K_pK_iK_d

temperature/

flux

proportion

③Dispatch parameters

②Collect data

Production

system

voltage/

currency

①Record data

④Dispatch parameters

⑤Sync parameters

PLC

DCS

Others

** Manually update parameters

Learning module

DB

UI

...

Auto

decisioning

Auto

tuning

Fans and pumps in traditional industrial facilities consume a lot of energy. AI intelligent algorithm controls fans and pumps, saving 10%-30% energy compared with traditional methods

Use case

Sales forecasting

Text models

Image models

Products

Reasonably design the product label system, and build a complete product feature library based on the extraction and integration of features related to the product, text, pictures and other aspects

Sales

Historical sales data is the most basic data for building a time series model, and the characterization of historical sales data is a complex and important task

Seasonal

Seasonal factors are important characteristics of sales forecasting, and sales data often show a certain pattern in seasonal changes

Promotions

Promotions can bring fluctuations in sales, so when making sales forecasts, consider the performance of previous promotions, as well as the changes that will be brought about by new ways of promotions in the future

Case 1：Sales forecasting for FMCG brand

Accurately analyze the relationship between product function points and sales volume to help product planning and design

The AI-based prediction accuracy rate reaches 90%, which is more than 20% higher than that of traditional methods

Sales forecasting

Case 2：Sales forecasting for homeware brand

Marketing strategy automation

Customer

data

txns

clicks

views

……

Product

data

price

category

style

……

WeChat

APP

Website

Visitors

Repurchasers

Purchasers

Precise content

Precise promotions

Precise products

AI contentmodel

AI

product model

AI

promo

model

The conversion rate of the brand's product push for members is low, and the sales increase is not significant. AI personalized recommendation products to daily churn users and transaction users, sales increased by 33%

Case study

Product recommendation

Expedia delivers personalized content to millions of users who don't find desired information on the site. Recommenders exhibit tailored content that may interest these users, and clicks on the content pages generate additional revenue over 1 billion USD per year

Case study

Marketing content automation

Marketing content automation

Marketing content automation

Intelligent customer service

• Non-linear conversation that can revert back even got interrupted
• Predictive coordination between AI and human CS agents

Multi-round convo

AI handles questions

请问您是

Val

吗？

| Customer response

是的 | 没错 | 不是 | 搞错 | 朋友

您目前是否有

Val

意向呢？

| Customer response

不用 | 不了解 | 介绍 | 可以 | 感兴趣

那我让具体的业务人员跟您联系好吗

| Customer response

不 | 稍等 | 可以 | 行 | 马上 | 打电话

你们是什么单位/公司

| Customer question

我们是XXX，是XXX技术研发的产品呢

质量有保障吗

| Customer question

您放心，我们的产品都是XXX

算了算了，感觉不靠谱

| Customer question

您再考虑一下呀，活动期间最低价呢

Diverse audience

Main session

Q&A session

Line 1

Line 2

Line 3

Line 4

（busy）

（free）

（free）

（busy）

AI model decisioning

Predictive coordination

Intelligent customer service

provides AI training, consulting and project execution services to enterprises.

International versions

forthcoming...

Summary

Create decisons, contents and optimize results via tailored AI

models

Systematic methodology that is operable, measurable and iterable

AI models when properly integrated become enterprises' growth engine