3 Lessons in AI/ Machine Learning from Artefact (Tech Strategy – Daily Article)

Facebooktwitterredditlinkedin

I had a good 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.

An Introduction to Artefact

Artefact’s 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 is a publicly-listed French company with +900 staff around the world. Edouard is Managing Partner Asia, based in the Shanghai office. Arthur is a Data & Consulting Director based in Shanghai. The company has offices in Shanghai, Hong Kong, Malaysia and Singapore – and the client list 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.

Ok, on to my three lessons. Note: these are my words, not theirs.

Lesson 1: Everything Starts with Data Visibility. That’s How You Get Smarter, Faster and Better.

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 digital 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?

It’s about data, data visibility and data analytics.

Most organizations already have tons of data (i.e., information) about 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 have tons of Excel spreadsheets and notes about current inventories.
  • Branch managers have tons of documents and reports floating around.

The net result is most data visibility is just occasionally reporting to middle and senior management. There is little 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 or physical) 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 massive difference between sort of knowing the inventory in your warehouses and stores and knowing everything. Business has always been like a pitch black factory and management just has a flashlight. They can point it at things and see them for a while. That is how reports, analysts and traditional management operate. But with data visibility, it is like the lights get turned on. Management can suddenly see everything. 

Data ecosystem company Snowflake calls this creating “a single source of truth”. It’s not just that all the data is suddenly visible. It’s that everyone can see it. And everyone can immediately start running data analytics on it. I wrote about this here:

Once management becomes data-driven, it immediately becomes smarter, faster and better. Decisions are no longer made based on intuition and some analysis. Everything is based on data. And you get immediate feedback on whether your decision is working and not.

In my 6 levels of digital competition, I didn’t even put smart, fast and data-driven management as one of the levels. I made it a standard part of operating performance for all digital businesses. It is assumed that all digital businesses are smart, fast and excellent. You can see this listed on the below left (right below Operating Performance).

 

Lesson 2: AI / ML Gets You Lots of Improvements in Existing Operations.

Arthur and Edouard mentioned quite a few different AI 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 making existing operations more efficient and effective. Some examples are:

  • Fraud detection. I was at a dinner with Alibaba President Michael Evans a few years back and he kept bringing up fraud detection. It’s just 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 the ecommerce and financial services. AI/ML is the go-to solution for fraud issues in operations.
  • Customer service. I did a couple of podcasts about JD’s AI initiatives. And I was surprised that customer service was a big focus for them. Call centers just don’t scale well and better AI is hopefully the solution to rapidly increasing customer service requests.
  • Quality control. AI is quite good at computer vision. It is one of those areas that is moving particularly fast and 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 really well and that is useful in quality control, whether it’s checking products off an assembly line or rental cars being returned.

I put all the above examples in the “like spell check” category. Spell check is just awesome software. It checks all text for mistakes and 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. They will check tons of things continually. And flag the problematic items for review. 

We can put all of these under the category of cost and effectiveness improvements within existing operations.

Lesson 3: AI is the Key Tool for Improved Customer Engagement.

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. These customer now have a sea of choices – but only limited time and attention.

Most every merchant and brand is confronted with the challenge of both how to reach and win with consumers directly. It’s an arms race and AI/ ML is a really important tool in this fight. 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. Forever.

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 “Better Understanding” Consumers.

So much of marketing is based on crude demographic information, such as age brackets and location. Millennials vs. Gen Z. First tier vs. third cities. The rise of social networks and detailed targeting based on social and interest graphs improved this a lot. But it makes companies dependent on Facebook.

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. You want to know about their behavior long before they purchased. And, ideally, you want to focus on what Alibaba Vice Chairman Joe Tsai calls customers with “high consumer intent”. You also want to know what happened after they purchased. 

And it turns out AI with the right data is very good at creating 360-degree views of consumer behavior. But, unfortunately, this is also the data that platforms like Alibaba, JD and Amazon are not going to give brands and merchants. They will give customer purchase information. But not the 360-degree behavior information. For that, brands/merchants will be dependent on the platform’s dashboards and analytics services. Alibaba calls this uni-marketing.

So merchants and brands are going to have to get 360-degree behavioral information themselves. They will get a lot of this data directly from consumers. They will partner with other companies for some data. And they will purchase other data. They will have to build an imperfect jigsaw puzzle of data about their customers.

And with this behavioral data, the AI can start to make predictions. Note: whenever I see the term “AI”, I mentally replace it with the phrase “cheap and fast prediction”. 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 lessons from our discussion. I think it fits in nicely with my own frameworks.

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 really 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 in AI. Hire consultants and advisors. Start training internal staff. And go after 2-3 easy first AI projects at the same time. Companies can 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. Most companies need to start moving ASAP.

———-

Related articles:

From the Concept Library, concepts for this article are:

  • AI – Cheap Prediction
  • Data visibility and analytics
  • Smile Marathon: Machine Learning

From the Company Library, companies for this article are:

  • n/a

I write, speak and consult about digital strategy and transformation.

My book Moats and Marathons details how to measure competitive advantage in digital businesses.

I also host Tech Strategy, a podcast and subscription newsletter on the strategies of the best digital companies in the US, China and Asia.

With my subscription newsletter, you will:

Get a deeper understanding of the strategies and business models of the best digital companies.

Get specific frameworks for measuring competitive advantage in digital businesses and for traditional businesses doing digital transformation.

Get an edge in predicting what is going to happen next and who is going to win.

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.

Photo by Possessed Photography on Unsplash

———

twitterlinkedin
Facebooktwitterredditlinkedin

Leave a Reply