I had a good discussion interview with Arthur du Passage and Edouard de Mezerac of Artefact, a data consulting and digital marketing company. And I thought they had a really good vantage point for how AI / ML is being used on the ground in China. I’ve summarized my 3 take-aways from our discussion.
But first…consider joining Asia Tech Strategy, my podcast and subscription newsletter on the strategies of China / Asia tech companies.
Some background on Artefact.
Artefact is a publicly-listed French company with +900 staff around the world. Their business is basically to bridge the gap between nice AI PowerPoint presentations and actual business improvements. How can AI and machine learning improve your business right now? That’s a really good consulting-type engagement. And as a company founded by both ex-McKinsey & Co consultants and data and software technologists, they seem to have the right skill set for this.
Artefact has offices in Shanghai, Hong Kong, Malaysia and Singapore – and the client list shown on their webpage is pretty much a who’s who of multinationals and brands:
- Banking and insurance, including JP Morgan and Allianz.
- Travel and hospitality, including Accord and ClubMed.
- Automotive, including Nissan, Audi, and Hertz.
- Media and entertainment, including Disney, Ubisoft, and Mattel.
- Cosmetic and luxury, including L’Oréal, Chanel, and Dior.
- Industry, telecom and high tech, including Samsung, Canon, and Thysskrupp.
Edouard is Managing Partner Asia, based in the Shanghai office. Arthur is a Data & Consulting Director based in Shanghai. They have over 100 staff in China. Ok, on to my main take-aways. Note: these are my words, not theirs.
Take-Away 1: Everything starts with data visibility: having access to the data and being able to leverage it into better understanding, decisions and efficiencies.
As companies go digital, there are opportunities to do more with less, which is really the point of technology. Yes, you can replace factory workers with robots. Yes, you can replace people with software. But this is usually not about ending a department. It is usually about giving staff tools that make them dramatically more productive and effective. The low hanging fruit of AI and software is to make existing operations and organizations more efficient and effective. But where do you start?
Well, it’s not really even about technology or software. It’s about data and data analytics. Getting more complete and rapid data visibility is a good starting point.
Most organizations already have tons of data (i.e., information) on their operations. They have inventory reports, performance assessments, supply chain info and so on. It’s just that the information is located in emails, documents, reports and Excels sheets scattered across the company and its people. For example:
- Staff in the purchasing office have tons of emails with suppliers about shipments and invoices.
- Warehouse managers scan and track current inventories in their own Excel spread sheets and notes.
- Branch managers have tons of documents and reports floating around.
The net result is there isn’t that much reporting to middle and senior management. Let alone company-wide sharing of information. And when reporting does happen, it is usually incomplete and slow. Staff periodically take inventories. Summaries are sent. And it is all done in reports that take time and effort to create.
But AI / ML is good at scouring companies’ systems, emails, excel sheets, documents and reports (whether digital assets or physical assets that have been scanned in) and pulling all that information into a centralized database that can be analyzed and searched. Suddenly, management has dramatically improved data visibility. It is far more complete and it happens continually and in real-time. There is a big difference between sort of knowing the inventory in your warehouses and stores and knowing everything, every hour. Complete and real-time data visibility can significantly improve operations. And it can improve day-by-day decisions. Complete and real-time data visibility should make the whole organization smarter and faster.
Take-Away 2: AI / ML gets you lots of improvements in existing operations.
Arthur and Edouard mentioned quite a few different use cases, which, of course, varies industry-by-industry. But I think a lot of this can be put under the category of AI/ML resulting in improvements in current operations. Making operations more efficient and more effective. Such as:
- Fraud detection. I was at a dinner with Alibaba President Michael Evans a few months back and he kept bringing up fraud detection. It’s an ongoing challenge for Alibaba’s ecommerce platforms and also for Ant Financial. The IT systems that detect unusual behavior have not kept up with the scale and speed of expansion of ecommerce and financial services. AI/ML is the go-to solution for these types of operational challenges.
- Customer Service. I did a couple of podcasts about JD’s AI initiatives. And I was surprised that customer service was a big focus on this. Call centers just don’t scale well and bigger AI is seen as the solution to this operational problem.
- Quality control. It turns out AI is quite good at computer vision. It is one of those areas that is moving particularly fast and that is pretty easy to commercialize. Cameras with computer vision software can check and count inventory in warehouses and on store shelves. It can detect visible defects in produced and shipped products. It can replace a lot of security guards. Computers can just see particularly well and that is good for quality control, whether it’s checking products off an assembly line or rental cars being returned.
I put all the above examples in the “spell-check” category. Spell check on a computer is just awesome software. It’s free on every website and computer application. It checks all text for mistakes and you then you only have to review the flagged items. That looks to me like the future of fraud detection, quality control and a lot of these AI focused operational tools.
We can put all of these under the category of cost and effectiveness improvements within existing operations.
Take-Away 3: AI is a key tool for engaging with customers.
In an age of abundance with endless options for products, services and entertainment, companies are in an arms race to capture and retain the time, attention, engagement and spending of customers who now have a sea of choices – but limited time and attention.
And as certain platform business models (Alibaba, Pinduoduo, JD, Tencent, Google, Facebook) dominate consumer attention, every merchant and brand is confronted with the challenge of both how to succeed on these platforms and how to win with consumers directly. So being on the technology frontier of AI/ ML is really important in this arms race. McKinsey had a nice paper about this a few years ago, arguing that consumer loyalty is basically dying. That you need to continually fight and re-fight for consumer attention, consideration and purchasing.
Edouard and Arthur spoke a lot about customer-centric use cases for AI / ML. And they had a good phrase, which was to use AI/ML to “better understand, anticipate, engage and serve” customers. I like that break-down of activities. This could apply to B2B as well, but I’m going to focus mostly on B2C (i.e., consumers).
On “understanding” consumers.
So much of marketing is based on crude demographic information, such as age brackets and location. Millennials vs. Gen Z and first tier vs. third cities. The rise of social networks (i.e., who you know) offered some improvement on this. But not much.
The real key to understanding and marketing to consumers is behavioral information. You want a 360-degree view of the behavior of your current and potential consumers. You want to know what they looked at before they bought. What they didn’t buy. What they also looked at. You want all the data about their behavior long before and after they purchased. And, ideally, you want to focus on what Alibaba Vice Chairman Joe Tsai calls those with “high consumer intent”.
And it turns out AI with the right data is very good at this. But, unfortunately, this is also the data that platforms like Alibaba, JD and Pinduoduo are not going to give brands and merchants. They will give customer purchase information. But not the 360-degree behavior information before and after purchase. For that, brands/merchants will have to use the platform’s dashboards and analytics. Alibaba calls this uni-marketing.
So the 360-degree behavioral information is what merchants and brands are going to have to get directly from consumers where they can. And they will have to painfully and progressively rebuild (through what we call today CDPs, Consumer Data Platforms). At best, this is going to be an imperfect jigsaw puzzle as consumers are spending most of their time on ecommerce platforms. Although WeChat mini-programs may be changing this situation.
With behavioral data, you can start to make predictions, which is really the whole point of AI. Note: whenever I see “AI”, I mentally replace it with the term “cheap and fast prediction”. And as these are pretty standard at this point, it’s really not about the algorithms for most companies. It’s about getting behavioral data.
I asked Arthur and Edouard what the most popular use cases were on the customer-side – and they said:
- Propensity modeling to increase conversion or repurchase. Once transaction data is tied back (even partially) to consumer and behavioral data, you can train a model to learn who is likely to do what. Better targeting becomes easy.
- Improving spending on digital marketing. This is about improving visibility on marketing spend. And making it more efficient (i.e., less wasted money). Reconciling marketing spend with the consumer journey is critical. And this means capturing data from the important touch points. And getting a 360-degree view of consumer behavior. Advanced ML techniques can help understand the weight of the different touch points between brands and the clients.
- Micro segmentation. Once you have a 360-degree view of consumers, you can do precision marketing. This should change the impact of media spending.
- Demand forecasting. Obviously, a big deal for ecommerce and retail. Using consumer information to drive improvements in forecasting and inventory.
Those are my three main take-aways from our discussion. And I think they provide a nice simple framework for AI use cases in China today.
- Take-Away 1: Everything starts with data visibility: having access to the data and being able to leverage it into better understanding, decisions and efficiencies.
- Take-Away 2: AI / ML gets you lots of improvements in existing operations.
- Take-Away 3: AI is a key tool for engaging with customers.
One last comment.
Arthur and Edouard mentioned that most industries are now being impacted by AI/ML to some degree. Most all CEOs are thinking about it. And B2C brands and retail especially feel the pressure right now. But the biggest problems CEOs have is a combination of strategy, staffing and execution. How do you get a company moving in AI when it probably does not have the staff and expertise to execute today? That’s a good role for consultants.
Hence the idea of getting a running start. Hire consultants and advisors. Start training internal staff. And go after 2-3 easy first AI projects at the same time. Build expertise and staff around smaller projects. I’ve suggested several here.
And, there is an urgency to act. Yes, speed and execution are important. But what matters even more is when you get into the race. It is not too late. Companies need to start moving ASAP.
I write and speak about digital China and Asia’s latest tech trends.
I also run Asia Tech Strategy, a podcast and subscription newsletter on the strategies of China / Asia tech companies.
My subscription newsletter offers:
Deeper insights into the strategies of the tech giants of China / Asia. I help investors see around the corner – both with tech giants and rising companies.
See the big picture. Get a better understanding of Asia’s digital ecosystem. What are the important tech themes? What will the future look like? Where to hunt for opportunities?
A unique view from on the ground of digital China / Asia.
Note: This content (articles, podcasts, website info) is not investment advice. The information and opinions from me and any guests may be incorrect. The numbers and information may be wrong. The views expressed may no longer be relevant or accurate. Investing is risky. Do your own research.