Filip Tichý | 14.2.2024 | News
At the end of 2022 the world was shocked, when Chat GPT notoriously took the world by storm, as an app that with the use of AI (Artificial intelligence) is capable of making presentations, successful admission tests or articles. The astonishment caused by Chat GPT, finally fully shed the light on what has been happening in the technological sector for years. The technological advancement in artificial intelligence has moved this topic on the undisputed first place in strategic thinking of large companies. In the last 10 years, the largest (not only) technological companies invested billions into AI development and acquisitions connected to artificial intelligence. We will discuss how large digital companies have already implemented artificial intelligence into their work, which caused their unprecedented vertical and horizontal growth. Just as Chat GPT and other generative AI applications are daily causing changes to business, especially in areas, which use text generation of any form (education, media, consulting, entertainment), the other AI apps will continue to do the same to everything else.
Artificial intelligence influences and changes every sector and area of the operations of a company, similarly as the introduction of electricity or internet. This is why AI (like electricity, telecommunications, internet) is considered to be a “General Purpose Technology“ (abbreviation in Chat GPT does not refer to General Purpose Technology, but it is a rather funny coincidence). In the world and neither in Slovakia so far we do not see a massive automation, the replacement of people by robots, etc. Quite the contrary, there is a lack of people and the main drive of robotisation is the scarcity of a human workforce. How long will it stay this way? What impact will the automation have on different sectors? Will it be an incremental change to the operation of everyday processes? Or will my specific sector become unrecognizable? The scale to which the specific sectors will be transformed by artificial intelligence will vary. However, what is certain is the fact that each sector and each profession will definitely be impacted (just as each profession was impacted by the internet). Therefore, the change is inescapable, yet the question remains, when will it come and how big the change will be. It is universally expected that artificial intelligence will bring an unprecedented increase in effectivity, decrease in cost price and that it will open the door for new presently unknown areas. When it comes to competition between companies, the company that will be the first to take full advantage of the opportunities and benefits artificial intelligence can provide, will gain a huge head start and in extreme cases, the transformation process can lead to cleansing of the market, meaning that some companies may cease to exist during this process.
Authors of the book “Competing in the Age of AI“ (Iansiti, M., & Lakhani, K. R. [2020]. Competing in the age of AI: Strategy and Leadership When Algorithms and Networks Run the World. Boston, MA, Harvard Business Review Press) present an interesting concept of a so-called “AI-centric organisation“. This company has its entire production and operating activities handled by the artificial intelligence. People work here solely to ideate, identify new activities and models, code them and subsequently perform monitoring, maintenance and evaluations. This means that client engagement, accepting of orders, pricing, purchases, production and supply of products or services, invoicing, complaints, etc. (all production activities) are performed automatically. Service activities like marketing campaign targeting, accounting or legal and compliance activities are performed automatically as well. In other words, human input is eliminated from the core of operative activity of the company and it is moved to the edge. The beauty of this model is, that when artificial intelligence can perform a certain activity, scaling is practically infinite, regardless of whether it is required to process a thousand or a million orders. The only thing needed after scaling up to million orders is a bit of additional computing and data capacity. A huge part of variable expenses (in the case of intangible products or services, all variable expenses) will change to fixed and they do not increase proportionally to the increase in turnover. This model is therefore extremely profitable when higher turnover is achieved.
A real example of a company like this is Netflix, where human employees work on creation and coding of new products, control, and evaluate existing apps. A very simplified example: they decide to fix (lower) the rate at which the users end their subscription. They code a new AI model, which uses machine learning and learns to identify new individualised trends in user behaviour before they end their Netflix subscription. When users begin to show signs, which had in the past lead to the cancelation of a subscription, Netflix will offer them a discount, or perform a different preventive action to change the user‘s mind. Everything is fully automatic, in real time and fully individualised for each user. At first, the new model will go through several verification phases: sandbox testing, A/B tests (model will be available only to a portion of the population and the results will be compared between the population with the AI model and the one without it). Following this testing, the model will be fully released and the human team will begin to monitor and evaluate this new application. Afterwards, the operating activity is fully automatic and millions of users are offered discounts without any human interference, approvals, etc. Other cases of an AI company are for example Booking.com or Airbnb. Searching, booking, payments, complaint handling, all of that is fully automated. On top of that, these companies have significantly transformed the hotel industry by inputting the “software layer“ or the “operating system“, which eliminates the need for human dependent travel agencies, travel agents and hotel networks. It increases effectivity, profitability, but also its value for the user.
The core elements of the power of artificial intelligence are data and computing power. The more we possess of either one or both, the more powerful artificial intelligence we can create. The structure of an AI-centric organisation is, naturally, very different from regular companies. There are no separate departments like sales, purchasing or marketing; there is just one digital unit. What is even more important is the fact that these data are all in one place. A regular company has its own departments, and for each of system, and can be centralised in a SAP type ERP system. Data in ERP system (recorded orders, invoices, delivery notes, complaints, etc.) are nowhere near the full extent of all the data the company holds. Up to the moment of accepting the order itself, a lot of (very important) data is generated. First approach of the customer, customer response (speed, manner, type, level of interest), communication about the services and prices offered, how the customer accepts the offer (speed, manner, type, what was decided, what was considered, what was rejected, relationship with the prices or discounts offered). All these data are store somewhere. Whether it is in e-mails, Excel sheets, in logs, in an employee’s memory (yes, that’s data too, very important and definitely hard to access), analytical data of our website or online forms. Other departments like sales, purchasing or marketing also have their own relevant data regarding to their activities, stored in various forms. Therefore, when we talk about an AI-centric organisation having all its data in one place, we include the data not only from the ERP system, but also all the other data from z Excels, mails, logs and other sources, which are integrated into one “data lake”, in a readable and easily processable form, as well as in a proper modality (text, picture, sound, sheet, sequence, etc.). All of these data are stored in one big “data lake“, most often on the cloud (for example, the cloud solution S3 from AWS, which was built solely to store data for utilisation by an artificial intelligence). Then the individual AI models are being coded over this one set of data, whether it is for setting the price, manufacturing or service providing, marketing offers. AI model learns (by machine learning) from the company’s full history (or even better, on the complete history of all companies of the given sector). And when the AI model is carrying out its main task, it takes into consideration all relevant and possible trends, correlations and linked actions and reactions across all the activities of the company. Since the main raw material for functioning of the artificial intelligence is a large amount of data for its machine learning, AI-centric organisations identify not only the internally generated data but also external sources of data they can use. These data are either freely accessible (datasets of mobile phones, ECB, finstat, etc.) or anonymised data purchased on the market (credit card companies, companies monitoring consumption, companies that aggregate and sell user data, for example from Google or Facebook, analytical companies, satellite data, etc.). Structure setup in this way, allows the AI-centric organisation to work as one digital unit, where different AI models, which work above the large integrated data pack, perform the operative activities of the company. Structure setup in this way, allows the AI-centric organisation to work as one digital unit, where different AI models, which work on top of a centralized, fully integrated data warehouse, perform the operative activities of the company.
Sounds futuristic? This AI-centric organisation only uses a “weak AI“, which from a technological point of view exists for quite some time and many companies have already built their present success on this model. Development of artificial intelligence aims toward a much stronger artificial intelligence, “strong AI“, whose potential still exists only in theory, but holds power to change even an AI-centric organisation beyond recognition. So, how can an AI-centric organisation be an inspiration for a manufacturing company, logistics centre or a hospital? Not every company will be able to become a 100% AI company. However, it is interesting to analyse, whether my company can try this concept at a 5%, 10% or 25% rate. Implementing the concept of an AI-centric organisation only at a rate of 5%, or simply into just one department, can be the decisive factor for the future success or failure of our company.
Considering that artificial intelligence will in one way or another affect all sectors and activities, leaders of all companies should think about how their sector will be affected and when and which steps should be taken. In terms of long-term strategic planning, it is important to analyse the possibilities of utilisation of artificial intelligence already today, and think about how my sector will be changed in the future. Many companies are organising internal conferences on this topic as well as creative manager workshops. It is important to realise that the topic of artificial intelligence is not a topic isolated to IT. Its usage must also be discussed from the user’s point of view in terms of processing improvements, effectivity and performance. We do not have to understand AI from the technical side, but we can hold a workshop on the topic of: “Imagine that for one day you have a magical Harry Potter colleague, who can do anything. Literally anything. How would you use him? Make him handle orders that are seemingly endless or full of mistakes? Make him perform never-ending calculations of selling prices until it is acceptable (which is an activity utterly futile with a human work force)? Or have him run the whole production part of my business? How would I use the freed-up capacities? Which new products could I sell to my customers with this new magical advantage? What horizontal or vertical growth could arise from this superpower? Structuralised workshop on such topic can help us formulate a corporate AI strategy, but also explore a series of untapped scenarios for our business and give the company a new innovative impulse.
The fact is, that if a company decides to implement artificial intelligence, it could not be launched immediately. Following the steps taken by an AI-centric organisation, our first step should be the identification of sufficient amount of internal and external data, and compile them into a one functioning package. This can be achieved in multiple steps, where each step represents a complex and difficult project, that could take up to several years. Also, it is possible to take a different strategic approach: OK, I will not be the “first adopter“ of artificial intelligence in my sector, because my company is, for example, small. However, I still need to consider, that in only 5 to 10 years, artificial intelligence will become mainstream. And then all kinds of AIs will be already operational, tested, tweaked and widely available. When this time comes, every company will be forced to implement artificial intelligence (or transform their business model into a niche “artisan“ or “fully-human“ retro model). So, which steps should be taken already now in preparation for a future AI adoption (in whatever form)?
Most companies are already fully digitalised, but there is still a fair bunch of companies, where some of their areas are still “hardcopy”. Some accounting documents and signed delivery notes are still in binders, important contracts are in safes etc. It is very likely that in 5 years time, we will have easily accessible, cheap and effective accounting AI apps. For a small company to be able to implement it in these 5 years (and benefit from lowered costs and increased effectivity), it must begin its full digitalisation process now. Digitalisation also offer many other advantages (in security, internal control), but that is for another discussion.
Which data sources, that are directly or indirectly relevant for us, do we have in the company besides our accounting and ERP system data? E-mails? Different Excel sheets? Client folders? Are there some important internal data, which are not recorded, because they are stored in minds of employees themselves (e.g. important client negotiations)? How can we simply record this data? Is it legal to record (and automatically process) such meetings? All internal data sources must be identified and mapped. Identified data must be “cleaned“, tagged and adjusted into a form, that can be easily processed by an artificial intelligence.
The more data we possess, the more powerful artificial intelligence becomes. The more data we possess, the more complex and more effective can artificial intelligence model become. Also, the more data we possess (historical, from the sector, from similar sectors), the more can an artificial intelligence learn, which leads to more accurate and better decisions. Large companies undertake expensive and difficult steps to improve their AI models, so that they can collect a wide range of external data. Smaller companies, e.g. in retail, can think about data acquisition from their competitors, from companies that collect and monetize data, from national regulators, or simply try the free “internet mining“ method. Data quality can differ from an AI’s point of view, internet mining collects low quality data (missing tags, verification of information credibility, context), whereas historical data of real transactions from our or similar company represents data of the highest quality, which is perfect for an artificial intelligence.
After identifying and recording all internal and external data, we need to integrate them into one big data package – the data lake. This phase can begin as soon as the second step (internal data identification) is concluded. It is a difficult project, which may require a significant change of the IT infrastructure. The simplest solution to the data integration would be storing the data on a cloud service (e.g. the aforementioned AWS S3). This does not necessarily mean a transfer of the whole IT infrastructure to the cloud, but also an automatic and instant interface or mirroring of the local data sources to the data lake on the cloud. For the AI models to be working efficiently, new data must be very quickly (preferably immediately) made available on the cloud. This would make the company from the data structure point of view “AI ready“ for implementation of effective AI apps.
Artificial intelligence will bring changes to all areas and sectors. If my business is, for example, finances, retail or healthcare, the potential of weak AI is enormous. The same could be said about companies that collect huge amounts of data from thousands of customers or users, like in logistics, infrastructure, energy, telecommunication, transport, online advertising, e-commerce etc. To consider the impact and implementation of artificial intelligence is a relevant strategic topic in any sector. Companies should begin to build their AI strategy: Do we want to be a 5% AI company? 10% AI company? Which area or department has the most potential for artificial intelligence? How will artificial intelligence change my sector and what will my competition look like once powered by artificial intelligence? Any company, that wants to stay relevant and successful even in 5 or 10 years, must begin to seek answers to these questions and begin to adapt and change their IT data structure. It is expected, that very soon new technologies and apps similar to Chat GPT, will appear each year. And although their impact will be incremental, they will swiftly change business and the world as we know it today.
Ľubomíra Murgašová | 10.9.2024 | News
Fotovoltika v daňových súvislostiachSolárna energia je v posledných rokoch veľmi zaujímavou alternatívou spomedzi…
