The AI Magic Triangle
This article is a continuation of Warwick Matthews’ discussion with Dr Anastassia Lauterbach, about trends in AI for business, and what it means for Japan.
アナスタシア・ラウターバッハ博士は、ポツダムの人工知能、データ、データ倫理の教授であり、起業家、テクノロジー戦略家でもある。国際的な社外取締役であり、英国、ドイツ、ロシア、米国の上場・非上場企業の監督・諮問委員を務める.
ラウターバッハ博士は言語学の博士号を持ち、ダイムラー、Tモバイル、ドイチェ・テレコム、クアルコム、マッキンゼーといったフォーチュン500社に20年以上勤務している.
技術者ではない人々、特に若者にAIリテラシーを提供することを目的とした会社、AI Edutainmentの創設者。 スイスのルツェルンを拠点に活動.
In our previous article, we learned of Dr Lauterbach’s concept of the “magic triangle” for AI adoption in an organisation:
In this blog, we will dig a little deeper with Dr Lauterbach into how an organisation makes an AI dimension fit into its overall business strategy and operations.
Can you tell us more about the “Magic Triangle” of AI adoption?
The first part is reviewing your business’s strategy and identifying paths for how AI will fit into that. Rather than finding a home for AI in your business (a.k.a. “solution looking for a problem”), we need to step back and ask the fundamental question: What does our business do to be competitive?
Strategy starts with making a list of your company’s USP (“unique selling proposition,” i.e., differentiators) to answer the “What” in a given market and competitive environment. How do you justify why your customers come to you and pay your invoices? What do you do to attract the best partners and vendors? Where are you best at, and what are your bottlenecks?
The second point is the “How”: How do you implement the strategy based on where you are now? Do you need to establish something new if you have a legacy technology platform or a crumbling supply chain? Do you need to insource infrastructure and talent to speed up your time to market? We talked here about the “Trail of Tears” for digital transformation. Countless transformations around the globe fail, as companies lack the skills and patience to execute these successfully.
Touching our technology stack is always risky, so sometimes it will be smarter to build something parallel to the old system and then transition, rather than try to repair everything.
You might decide your business will benefit significantly from the ability to work with unstructured data, for example, to process customer feedback or complaints, but this then creates a requirement for your IT infrastructure. It is more than just some models, analytics, and an LLM license.
So there is a need for significant investment: every single business has a shadow in Data, and nowadays this “digital shadow” drives the value of almost every business on this planet, whether it be agriculture, healthcare, or financial services. We must take the investments in this side of our business seriously, and it is not just about paying the big consulting firms huge amounts of money to deliver “AI in a Box”.
A business in this space invests 25% of its revenue in Cloud and associated technologies. If they are brilliant, they will spend 15% on data hygiene, ensuring the best data is used in the best ways for the business. This is obviously easier said than done, and it can be seen as a very tedious task just to sit down, go file by file, applying some algorithms to aggregate certain files into clusters, trying to understand what is happening there, and just cleansing, cleansing, cleansing.
At CDL, we enjoy building the data cleansing and the MDM, and the intelligence to fuel these other things.
This is fantastic and can be the backbone of any successful AI implementation.
So, without cleansing the data, you cannot successfully release any model, including building a customised LLM or a forecasting framework.
And then we come back to “Who,” i.e., talent, the third corner of the triangle. As we said last time, it’s not just about your internal employees; it’s also about how you pick vendors to take advice from or partner with.
Beyond that, we also need to differentiate between technical brilliance and communication. I am an outward-oriented introvert, so I can speak in front of 10 people or 1000 people, but I’d always rather sit and read a book in the corner. Some of the technical people I have worked with are like me multiplied by 1000. They want to work with their data and algorithms; they will be drawing on the whiteboard, and you will just see their back while they are working.
AI experts often prefer to be doing AI rather than explaining it to the Board.
Suppose you are trying to communicate with the Board of Directors, from which you are asking for millions of dollars in Cloud and AI funding. In that case, you need to be able to tell stories that they can understand. I coach CTOs and CIOs as part of what I do in my AI Edutainment business because it is so vital that technical people can talk about these things in a language that “Muggles” (the word for non-Wizards in the Harry Potter stories) can rely on and can understand without any problems.
As you know, I always say, “If someone who is pitching you AI has more Sales and Marketing personnel than data scientists, be very cautious about working with this company.” However, the ability to communicate these complex concepts slickly is also an important skill. So we need both.
This is why I say it works best for a company to create advisory committees (rather than expand the Board) to bring ideas and R&D results back to the Board and have people around who can then explain the potential, explain the risks, and help the Board make an informed decision about AI and technology investments.
This also applies to cybersecurity and infrastructure; you want people advising you with real-life experience building and running these things.
Do you have some overall advice for bringing together the triangle’s three points?
Talk to your people, go through the systems, identify those crown jewels in data, and then talk to outside specialists about how to give more power to those Crown Jewels to level up your products, level up your value proposition, and perhaps open up new channels and partnerships. If you are serious about this exercise, you will detox your digital assets and can answer the question: “How many records do you need to succeed?” Not only will it allow for a more systematic AI implementation, but it will also allow for better cybersecurity risk mitigation!
Look at other angles, e.g., democratisation of data, perhaps creating a data trust for the benefit of the industry as a whole (without giving up competitiveness or market advantages).
Then, for the boards considering this journey — and this has nothing to do with technology per se — work with real coaches who can help with courage and the necessity of driving change. These coaches are formed executives with a proven ‘can-do’ mentality! It is normal and rational for people to hesitate and dislike change, especially in regulated industries like FIs. So, forming a group of those you can trust is imperative to immediately apply lessons learned elsewhere and develop psychological resilience to drive change.
Finally, as we touched on last time, it’s crucial to understand that AIs have some fundamental limitations that are not going away in the foreseeable future. AI and automation generally can take a considerable chunk of costs out of your business. Still, keeping all these things in mind, we must never forget the need for “human in the loop”. Look at the aircraft that are flying today, mostly automatically. The human pilots understand where to take over and what is going on. Of course, there are still communication mistakes between the pilot, the air traffic control, and the machine. But in general, it’s safe, and we are not planning to change that anytime soon. Or, we could look at humans in the loop while doing telepresence medicine, and we would say there should be humans in that context.
We cannot leave this to regulators or just say we must be “human in the loop everywhere” because regulators don’t know what they don’t know. These things must be reviewed on a domain basis, so perhaps different rules are needed for the automotive, energy, healthcare, or banking industries.
In our next article, we will discuss with Dr. Lauterbach the regulation of AI and cybercrime and how governments can or should react to this rapidly evolving area.