A “Multi-Agent System” approach to complement new business models – Case study: Cleveland Cavaliers

original article : https://www.linkedin.com/pulse/psd2-game-changing-directive-banks-salim-benhoussa?trk=portfolio_article-card_title


A business model describes the rationale of how an organization creates, delivers, and captures value, in economic, social, cultural or other contexts. It is a simple explanation of how do you intend to make money out of your company.

One of the most used framework is undoubtedly the Canvas Business model. The reason behind this approbation among practitioners and managers remain its simplicity and its ability to encompass key elements that “explain” how an organization creates, delivers, and captures value using 9 building blocks: Key Partners, Key Activities, Key Resources, Value propositions, Customer Relationships, Channels, Customer Segments, Cost Structure and Revenue Streams.

Although this template allows a better understanding of the “current” context within an organization, it falls short when it comes to explaining the performance of a corporation and defining strategies to act accordingly. This observation holds even more for digital companies given all the disruptiveness that characterize their environment.

How can we keep using such “primitive” frameworks to take strategic decisions in companies that claim to enhance processes previously executed by human beings using Artificial Intelligence?

The first part of this post explores the possibility of a more rigorous approach that would allow harnessing the latest concepts and technologies to assist managers in making strategic decisions.

The second part aims to use this approach to help the Cleveland Cavaliers – an NBA Basketball organization – trade for the right players in the short term, and make the right strategic business decisions in the long term.


Part 1: The King is dead, long live the KingS!

Nowadays, success stems from Information. With new legislation that impose standards on companies to share their data using APIs or web services, it is possible to extract large amount of information from multiple sources in a cost effective way and use it to make the right decisions. Clearly, current “business models” frameworks are stern, rigid and too generic to take into consideration all this available data and adjust on a regular basis to recent events.  


One of our main capabilities as human beings is our ability to learn from past events, experiences, artefacts and use this knowledge to face new challenges (which imply trying to make the right decisions).

Let’s take the example of a two-year child who is learning to distinguish different animals. During what we call the “training phase“, the child learns to associate different images of cats to the word cat, and different images corresponding to different breeds of dogs to the word dog.

The “prediction phase” occurs when the same child sees a new breed of cat that he has never seen before. During this process, he intuitively goes through all the stored images in his memory and tries to guess or predict the class/category of this unseen animal.


Scientists have been trying to mimic the processes within our brain to create “artificial neural networks” (ANNs) able to solve specific tasks better than humans.

An analogy could be drawn from the previous example to understand how ANNs could help businesses from a strategic standpoint. All what we have to do is replace images by a set of different inputs from financial performances, current customers, to investments on specific business units, etc. The output or the response (what our AI will predict) can be either a label such as positive/negative growth (classification algorithms) or a numeric continuous value of a predicted feature such as the expected revenue (regression algorithm).


Hmm… No Clippy, not really….


One of the limitations of this method is its inability to modelize complex interrelations between players within the same ecosystem. The figure below shows the generic “Ecosystem” of a Multisided Platform and the “value net” or role of the different players interacting with the company.


Customers, competitors, complementors and suppliers are key elements that should be taken into consideration while assessing or predicting the future outcomes of strategic decisions. Additionally, businesses have to create a healthy ecosystem in order to face all the disruptions and avoid losing important parts of the market. The last statement explain the huge investments operated by big IT companies to foster innovation and include startups within their ecosystems. Another example is Facebook and its willingness to encourage developers to use tools of its ecosystem by providing thorough documentation related to its APIs and web services.

Creating Vs Capturing value

The complexity and the diversity of ecosystems nowadays with several involved actors creates dependencies that are really hard to incorporate within the previous predictive model.  Here it is important to distinguish between two notions: creating value and capturing value.

Value creation occurs whenever an action is taken for which the benefits exceed the costs, or whenever an action for which the costs exceed the benefits is “prevented”. It can include different aspects of production: innovation, method of production, quality, quantity, relationships between the firm and its employees, etc.  

Capturing value occurs through changes in the distribution of value in the chain. The customers are considered as a part of the chain and thus increasing the price of a product means that the company captures more of the value created. The added value was not created but was before captured by the customers since they were paying less for the product or service. Value-capturing can occur by monetizing users (data-driven revenue model), pricing effectively and providing liquidity to stakeholders.


In general, creating value is essential in early stages to acquire customers and achieve critical mass. In a competitive context, the created value should be mainly captured by consumers and users.  Hence, optimizing one response such as the expected net profit which characterize “capturing value” might not be a good strategy. In fact, besides pushing some customers to switch to other competitors, it can have a disastrous impact on other players within the ecosystem. A possible solution to circumvent this limitation is what we call “feature engineering”. This means that we will have to create predictors (inputs) and responses (output) that englobe all the missing relevant concepts. With a large amount of historical data regarding the company or similar corporations, better results should be reached… at least in theory.


The Ockham’s razor principle was coined by William of Ockham – a logician, scholastic philosopher, and theologian – in the 14th century. This principle has been used by several scientists such as Isaac Newton and Leibniz. It is based on several statements; the most relevant ones are:

“The simplest explanation is usually the right one”

“when you have two competing theories that make exactly the same predictions, the simpler one is the better.”

“The explanation requiring the fewest assumptions is most likely to be correct.”

Accordingly, the best model to “predict” the outcome of a strategic decision should be characterized by the fewest assumptions among all other models. This means that this particular model should reproduce somehow all the inherent features and interrelations of the ecosystem. A representation using a “multi-agent system” with several Autonomous agents can help us achieve this objective.


Multi agent systems or MAS are systems in which “agents” interact within a certain environment to achieve a goal or objective and are generally used to solve complex problems in a distributed, decentralized and regulated fashion. Several conceptualizations exist, one of them relies on four dimensions: Agent, Environment, Interaction and Organization (Demazeau, 1995).

Agents refer to the autonomous entities in the system that are proactive, reactive, social, able to take part to an organised activity, in order to achieve its goals, by interacting with other agents and users in a shared space called environment.

The main difference between Multi-agent systems and the normal AI approach is the “social” interaction part as agents need to communicate and delegate tasks to other entities in order to achieve an objective. Several implementations of multi agent oriented programming platforms exist, one of them is called Jason. This interpreter is open source and extends the AgentSpeak(L). It is mainly based on Prolog – a logic programming language associated with artificial intelligence and computational linguistics – and enables a modelaziation of a problem using the MAS paradigm.

In addition to the Agent, other key notions such as the Goal, Plans, Beliefs and Events are of key importance. The (1) “Goal” refers to the ultimate objective of the agent; (2) “Beliefs” characterize the agent’s perception about itself, the environment and other agents; (3)  “Plans” describe how an agent should behave given a Goal, Belief or an event i.e. the different task that it has to perform; (4) “Events” are perceived events by the agent that can trigger the execution of a plan. More details about how to implement a MAS can be found on this link.

The figure below shows one possible way to modelize the interrelations between all the concepts in Jason.

A more complex model based on Cartago (Ricci et al., 2009), Jason (Bordini et al., 2007) and Moise (Hubner et al.,2009) can also be used if one needs to introduce the Environment and Organization dimensions.

I can imagine that. The next part of this article will tackle some examples of possible MAS implementations. Hopefully, it will help you understand how such systems work.

Part 2: Case study – Cleveland Cavaliers

One of my favourite Basketball players is LeBron James.  At the time I’m writing this article, the Cleveland Cavaliers (Cavs) – LeBron’s team – is struggling to win games due to the addition of new players. Two phenomena can explain this series of losses to average opponents: a lack of chemistry or a lack of complementarity. If the first option can be easily fixed with time and additional training sessions, the second one is much more problematic and usually require making trades to sell some players and add new pieces.

During this period of the season, Organizations such as the Cavs are able to trade for available players in the market to improve the team and adjust to the level of opponents. Trading for good players usually require giving several assets or role players in exchange.

This section explores the possibility of creating a MAS to find which available player would fit with the current team and add value to the roster. Dan Gilbert, the owner of the National Basketball Association’s Cleveland Cavaliers could use this system to define whether he should trade for DeAndre Jordan, Kemba Walker or Anthony Davis and if such an acquisition would enhance his chances of beating the Golden States Warriors in the Finals.

Modelizing NBA players by agents

During a game, each  NBA team consists of 5 active players on the court.  It is possible to assign to every professional NBA player an avatar or an agent with a Goal, beliefs, plans and events. One could suppose that the goal of each player is “win the game” but other more creative possibilities exist such as assigning “get a triple double” goal to a player like Russell Westbrook ,“Beat the stacked Golden State Warriors in the finals” to LeBron James or “get a Max Contract” to Isaiah Thomas.

A further analysis of NBA games shows some common characteristics between players that can be used to optimise the process of setting or defining the initial attributes of all agents corresponding to NBA players. In fact, each player has a specific role in his team.

The Center is usually the tallest player on the team, has more physical strength and overall athleticism and is in charge of rebounds in both sides of the court. Offensively, he uses pivot moves to hit a variety of short jumpers, hook shots, and dunks. On defense, the Center’s main responsibility is to keep opponents from shooting by blocking shots and passes in the key area.

The Power Forward is usually the next tallest players in the team, he is the closest to the center in physical attributes and playing style, but with more speed.

The Small Forward is the all-purpose player on offense: aggressive and strong; tall enough to mix it up inside but agile enough to handle the ball and shoot well. He must be able to score both from the perimeter and from inside.

The Shooting Guard is usually taller than a point guard but shorter than a small forward. He is generally the team’s best perimeter shooter.

The Point Guard is usually the shortest player on the team. He is supposed to be the team’s best passer and ball handler; not primarily a shooter.

Given the characteristics of each role, it is possible to take advantage of the “inheritance” concept in object-oriented programming languages to create 5 super class, each corresponding to one role and its implied goal, beliefs, plans and events.

For instance, the Goal of the center could be “win the game”. This goal leads to the execution of two plans:

  1. protecting the rim and getting rebounds during the defense phase

  2. Use pivot moves and score short jumpers during the offensive phase



As a result, each player can inherit all the previously stated attributes that correspond to its role or class. Furthermore, this modelization enables a flexible way to test several tactical schemes involving a player in different spots or roles. For instance, LeBron James – arguably the best player in the nba for the last 10 years – can play at all positions thanks to his incredible athleticism. This property allows his coach Tyronn Lue to use him in different ways depending on the opposition and available players (some players might miss some games due to injury). This MAS artifact could enable Tyronn Lue to test or run several simulations and figure out which team, rotation, and tactical scheme is the most appropriate against the opposite team.

Additionally, each agent could use messages to ask for the execution of a specific tactical scheme such as a “pick and roll” or a “screen”. Let’s see how a screen can be performed by agents in a system created to simulate the outcome of a game between the CAVs and the Golden State Warriors (GSW).




(1): The CAVs are running the offense. Each GSW player is guarding the player in the opposite team with the same role. Hence, the agent Isaiah Thomas (IT) – point guard of the CAVs – finds himself facing the agent Stephen Curry. There are no open players and due to his height (1m75 or 5 foot 9 inches), it is difficult for IT to shoot the ball over Stephen Curry (1m90 or 6 ft 3 in). The implementation of the agent IT includes asking the center Tristan Thompson to “set a screen” by sending a message to him.

Note that plans of all possible beliefs should be included in the description of the agent (code). This was omitted in this example for pedagogical purposes.




(2) Once the center Tristan Thompson (TT) receives this message, he starts moving towards Stephen Curry to set what we call a pick or screen. The center of the opposite team ZaZa Pachulia has as a belief “guarding the center” and as an associated plan “follow the center” to avoid giving him the opportunity to score easily if he gets the ball. At the same time, IT is running on the opposite direction of TT.




(3)  Stephen Curry has also “guarding the PG” as a belief and “follow the PG” as associated plan. However, he is now blocked by TT. As a result, he will execute a secondary plan “get around the player” before “following the PG”. This creates a lapse of time in which IT is open (no one is guarding him).




(4) Seeing that IT is open, the SF of the GSW Kevin Durant runs in the direction of IT to avoid giving him an open shot. In general, a normal point guard would have passed the ball to LeBron James, the CAVs’ SF as he becomes open now. But IT has “Get a Max Contract” as a goal and consequently, he will try to shoot the ball and score in order to impress General Managers (GMs) of NBA organizations and get the money that he deserves. Once he shoots the ball, the two PFs try to have a good position around the rim in order to get a rebound in case Isaiah misses his attempt.   

Here we can see that getting information from the environment (basketball court) is important in this example so that all agents can update their beliefs depending on the position of the ball, the handler of the ball, etc. By doing that, it is possible to trigger events in each situation. Those events would lead to adding new beliefs and plans that would enable agents to act accordingly.

In addition, the incredible amount of statistics about individual players in the nba such as the average points per game, the average assists, rebounds, blocks, steals, 3-points and free throw percentages etc, enables us to add more specificities regarding each player and to introduce probabilities for making shots and assists. Several free APIs such as “stats.nba.com” exist for this purpose and allow fetching and extracting this data. Additional data related to preferred shooting spots, preferred plays could complete the set of plans for each agent.  One must keep in mind that our objective is to provide the best approximation of real games in order to increase the accuracy of our simulations or predictions (winning team in each game). Once the process of defining the attributes of the agent (Goal, Beliefs, Plans, Events) is done for all NBA players, it is possible to simulate a game and find the right combination of players that would lead to maximize the chances of the Cavs to win games.

Regarding our initial objective, running a simulation involving the new lineup of the CAVs with the new potential acquisitions in one hand; and Golden State Warriors players as opponents in the other hand; can help assess the value of some possible trades.


What about the long term strategy?

In addition to players, NBA Organizations are able to trade draft “picks”. Those picks allow teams like the CAVs to acquire new young players from High School or College. Those assets are highly demanded by Organizations since they might give you the opportunity to acquire the next future superstar. Dan Gilbert has the Brooklyn Nets pick, a potentially top 10 overall draft pick. Adding this asset to some role players can facilitate the acquisition of NBA stars like DeAndre Jordan and Anthony Davis. However, Dan Gilbert is reluctant to do that because he couldn’t get the commitment of LeBron James for the next 3-5 years. As a result, the owner is considering this pick as an opportunity to acquire a new talented prospect to prepare the future of the franchise. The risk of this strategy is to frustrate LeBron and push him to leave the Cavs for another team to increase his chances of winning championships.

Let’s try to use all the previous business notions to build a MAS system that would help Dan Gilbert make the right decision. We will start by the ecosystem and the value net to describe the different actors.


Customers:We have mainly two types of customers:

(1) Businesses such as the NBA(organization), media platforms (TV, radio) and news outlets (newspapers, magazines …)

(2) End consumers or fans who buy tickets and products online and offline(stores).

Online & Offline commerce: These are the parties responsible of selling CAVs products to consumers both online and offline.

Players & staff: This includes all NBA players of the organization as well as the coach and other members of the Staff.

The owner: Dan Gilbert, an American businessman who is also the owner of the American Hockey League’s Cleveland Monsters and the Arena Football League’s Cleveland Gladiators.

Partners and Sponsors: Nike and Good Year are two Sponsors of the Cavs and LeBron James can be considered as a partner. Actually, this player has been the “face” of the NBA for the last 10 years and has a stronger brand than some NBA organizations. He is considered by some specialists as one of the rare talented prospects that exceeded expectations through the history of the NBA.

The Cleveland Cavaliers franchise value dropped from $476 million in 2009 to $355 million in 2011 after LeBron James left the Cavs to the Miami Heat according to Forbes (Rishe, 2014). When LeBron decided to come back to cleveland in 2014, LeRoy Brooks – a professor of finance at the Boler School of Business at John Carroll University – estimated that this move could generate $500 million for the local economy (Gregory, 2014). In 2017, a study made by Daniel Shoag – a Harvard professor – and Stan Veuger – a member of the American Enterprise Institute – highlighted that LeBron James had “a statistically and economically significant positive effect on both the number of restaurants and other eating and drinking establishments near the stadium where he is based, and on aggregate employment at those establishments. Specifically, his presence increases the number of such establishments within one mile of the stadium by about 13%, and employment by about 23.5%” (Shoag & Veuger, 2017).

All These observations confirm the “privileged status” of LeBron and why he could be considered as a partner for this organization.


Value net (Role of players):

Customers: Refers to Businesses and End consumers.

Complementors: Complementors are entities that complement the products and services of the CAVs. The Cleveland Cavaliers organization needs the NBA to reinforce its brand. On the other hand, the NBA needs success stories such as the one of LeBron, a superstar born in Akron (Ohio state) to a 16-year-old mother who raised the child on her own. The NBA needs emblematic figures like LeBron to increase its conversion rate among young generations especially given the popularity of football and baseball in the US.

Competitors: Here we can distinguish between two types of competitors:

  • Direct competitors: other NBA teams and in particular the Celtics in the Eastern conference and the Golden State Warriors in the Western conference.

  • Indirect competitors: this category mainly includes Ohio state teams in other Sports such as the Cincinnati Reds (Major League Baseball), the Cleveland Indians (Major League Baseball), the Cincinnati Bengals (National Football League) and the Columbus Blue Jackets (National Hockey League).

Suppliers: Represent all actors involved in selling products and services related to the CAVs from Manufacturers and Distributors to Wholesalers and Merchants. Investors could be seen as suppliers in terms of financial resources. Since it is hard to define all suppliers of an institution like the Cleveland Cavaliers, Suppliers won’t be taken into consideration in this section.


Created value VS Captured value

In 2014, the NBA announced a nine-year $24 billion media-rights deal with ESPN and Turner Sports. This deal corresponds to a raise from $930 million per year to a little bit more than $2.6 billion annually and took effect in the 2016–2017 season (Draper, 2014).

The increase in created value generated by the CAVs organization was not solely captured by this institution; players benefited from it as well. A recent study shows that the new 2016 TV contract had “a positive impact on players’ salary in the NBA” leading to an increase in the value of the salary cap (Kelly, 2017). This observation emphasizes the importance of having a healthy ecosystem for NBA organizations and maintaining a good relationship with players.

The 2011 lockout exemplifies the last statement.  In fact, due to contradictory positions between owners and players regarding salary cap and luxury taxes (Thompson, 2011), teams could not trade, sign or contact players, and players could not access NBA team facilities, trainers or staffs for almost 6 months and the regular season was reduced from 82 to 66 games (with all associated losses for NBA organizations in terms of tickets, TV broadcasts, etc.).

The Graph below shows that the effect of a new TV deal wasn’t as important for previous similar contracts.  

After this short analysis of the different stakeholders involved, it is clear that creating a good environment for LeBron James is fundamental for the CAVs given the small size of the market at Cleveland in comparison with Los Angeles (Lakers) or New York (Knicks). Additionally, having the “best player” of the last decade helps the franchise to “globalise” its brand internationally and compensate its popularity within the US market. Hence, Dan Gilbert HAS to make the necessary trades and use the valuable assets that he possesses including draft picks to build a competitive team and convince “the chosen one” to stay in the “land”.

The Chicago Bulls are still selling Michael Jordan items among different customer segments. His talent and ability to transcend during decisive moments to win important games had an astonishing impact on a whole generation. LeBron James is considered by many people as the second best player of basketball who ever lived and has the potential to become the local Hero of an entire state and generation if he stays at Cleveland and helps his team win other championships.  


Number of fans, Conversion rate & Reach

A businessman has to be prepared for worst-case scenarios including seeing LeBron James taking his talent to another competitive team to beat the Golden State Warriors. Here again, a MAS system could be built to assess the repercussions of this event in terms of number of fans.

For consumers, we will represent each customer segment by a defined number of similar agents (let’s say 1000) that have the same goal, beliefs, plans and events.  The goal of the agent could be “enjoy life” (as this agent represents a human being). Once a belief of type “become a CAVs fan ?” is created after receiving a message from the agent CAVs, the agent becomes a fan of the team with a probability that is equal to the conversion rate. If the agent becomes a fan, then a belief “Cavs fan” is created. This belief triggers plans of buying a certain amount of tickets and CAVs products such as jerseys or hats depending on the characteristic of its initial customer segment.  


A simplified demonstration

Step 1:


Step 2:


Step 3: (we suppose here that the generated random number p is smaller than the conversion rate)


Step 4:


We could repeat this process for businesses and create an agent for each type, that is TV, radio, newspapers, magazines. Note that there might be a correlation between the number of converted consumers and let’s say the number of articles about a player of the organization for magazines because media are in general a “mirror” of consumers and try to provide them interesting content in order to increase their reach and engagement.

Assuming that historical number of fans by franchise can be found on some articles, it is possible to apply this process for all NBA organizations, define the conversion rate for each one and hence predict their performance or approximate their generated revenue.

Note that this performance depends mainly on the conversion rate which depends on the caliber of superstars within the team as well as the sportive results (LeBron reached the finals the last 7 years and is a “guarantee” of good sportive results for organizations). If LeBron James leaves the cavs at the end of the season to another franchise, the conversion rate of the CAVs will drop considerably while LeBron’s new team will witness the opposite effect.


Also one must keep in mind that the number of new fans N equals to the reach multiplied by the conversion rate   (  N= reach * conversion rate ) . Consequently, it is possible for Dan Gilbert to reduce his losses by investing on marketing campaigns to increase his reach.  



It is important to understand that this method is not perfect but is supposed to give better predictions than normal approaches. In fact, traditional methods fail to encompass the impact of other actors and competitors on potential customers – a customer can switch to a competitor, and it is harder to convert fans previously converted by other competitors.

The accuracy of this method relies on two pillars:

(1) a good description of the context, all actors and their interrelations.  

(2) available data about all created agents and their behaviors.

Nowadays, people spend a major part of their lives requesting and consuming products/services on internet through several “touchpoints”. Those touchpoints can be social media platforms, apps, websites, etc. Interestingly, several platforms offer you the possibility to login using your email address or Facebook account. Also, it is recommended to add other personal data such as your phone number for security purposes. The end result is that it is possible to combine all data from different touchpoints to create an avatar for each person. This avatar encloses the behavior of its users and can be used as a base for constructing the consumer agent. With cloud computing and the evolution of hardware, storing huge amount of data is no longer a problem and hence this approach takes advantage of the new available technologies to enhance one of the oldest processes: making strategic decisions.

A similar methodology can be applied to the different actors within the ecosystem of the CAVs to define which distribution of the created value leads to the healthiest ecosystem and the best economic growth in the long term. Our analysis showed that Dan Gilbert has to make trades and convince LeBron James to stay in Cleveland, otherwise the value of the Franchise will drop considerably. However, LeBron James has expressed several times his willingness to acquire a franchise once he retires. Supposing that he wants to buy the Cleveland Cavaliers, leaving this organization could help him reduce the cost of such an acquisition in few years from now.  

If LeBron stays at cleveland, he will be remembered as the hero of the city (or even the state) in the game of Basketball like Michael Jordan. If he leaves, he will land in a competitive team to maximise his chances of winning several championships and will have the opportunity to buy the CAVs at a competitive price when he retires. Once again, “the chosen one” will be seen as the saviour of the “land”.  In both cases, LeBron James wins.


Demazeau, Y., & Demazeau, Y. (1995). From Interactions To Collective Behaviour In Agent-Based Systems. IN: PROCEEDINGS OF THE 1ST. EUROPEAN CONFERENCE ON COGNITIVE SCIENCE. SAINT-MALO, 117–132. Retrieved from http://citeseerx.ist.psu.edu/viewdoc/summary?doi=

Boissier, O., Bordini, R. H., Hübner, J. F., Ricci, A., & Santi, A. (2013). Multi-agent oriented programming with JaCaMo. Science of Computer Programming, 78(6), 747–761. https://doi.org/10.1016/j.scico.2011.10.004

Ricci, A., Piunti, M., & Viroli, M. (2009). Environment Programming in Multi-Agent Systems – An Artifact-Based Perspective. Retrieved from http://www.emse.fr/~boissier/enseignement/maop14/DOC/cartago/env-prog-cartago-tech-rep.pdf

Ricci, A., Piunti, M., Viroli, M., & Omicini, A. (2009). Environment Programming in CArtAgO. In Multi-Agent Programming (pp. 259–288). Boston, MA: Springer US. https://doi.org/10.1007/978-0-387-89299-3_8

El Fallah Seghrouchni, A., Dix, J., Dastani, M., & Bordini, R. H. (Eds.). (2009). Multi-Agent Programming. Boston, MA: Springer US. https://doi.org/10.1007/978-0-387-89299-3

Klusch, M., Bergamaschi, S., Edwards, P., & Petta, P. (Eds.). (2003). Intelligent Information Agents (Vol. 2586). Berlin, Heidelberg: Springer Berlin Heidelberg. https://doi.org/10.1007/3-540-36561-3

Hubner, J. F., Sichman, J. S., & Boissier, O. (2007). Developing organised multiagent systems using the MOISE; model: programming issues at the system and agent levels. International Journal of Agent-Oriented Software Engineering, 1(3/4), 370. https://doi.org/10.1504/IJAOSE.2007.016266

Bordini, R. H., Hübner, J. F., and Wooldrige, M. (2007). Programming Multi-Agent Systems in AgentSpeak using Jason. Retrieved from http://home.mit.bme.hu/~eredics/AgentGame/Jason/Jason_konyv.pdf

Rishe, P. (2014). A Tale Of Two Franchise Values: How LeBron’s Return To Cleveland Impacts Cavaliers, Heat. Retrieved January 29, 2018, from https://www.forbes.com/sites/prishe/2014/07/11/a-tale-of-two-franchise-values-how-lebrons-return-to-cleveland-impacts-cavaliers-heat/#797d4fb65fdf

GREGORY, S. (2014). Economist: LeBron Worth Over $500 Million to Cleveland | Time. Retrieved January 29, 2018, from http://time.com/2981583/lebron-james-cleveland-cavs-money/

Shoag, D., & Veuger, S. (2017). Taking My Talents to South Beach (and Back) Evidence on Local Externalities from a Superstar Athlete *. Retrieved from https://scholar.harvard.edu/files/shoag/files/south_beach_and_back_01.pdf

Kevin Draper. (2014). What The NBA’s Insane New TV Deal Means For The League And For You. Retrieved January 29, 2018, from https://deadspin.com/what-the-nbas-insane-new-tv-deal-means-for-the-league-a-1642926274

Kevin Draper. (2014). What The NBA’s Insane New TV Deal Means For The League And For You. Retrieved January 29, 2018, from https://deadspin.com/what-the-nbas-insane-new-tv-deal-means-for-the-league-a-1642926274

Kelly, T. (2017). Effects of TV Contracts on NBA Salaries. Retrieved from http://blogs.colgate.edu/economics/files/2017/08/Troy-Kelly-Final-Thesis.pdf


DEREK THOMPSON. (2011). The NBA Lockout: Here’s What You Need to Know. Retrieved January 29, 2018, from https://www.theatlantic.com/business/archive/2011/06/the-nba-lockout-heres-what-you-need-to-know/241251/



PSD2: a game-changing directive for banks?

original article : https://www.linkedin.com/pulse/psd2-game-changing-directive-banks-salim-benhoussa?trk=portfolio_article-card_title

In an attempt to anticipate the future implementation of the Payment Service Directive 2 (PSD2), fintechs and startups are trying to disrupt the banking industry by providing customers with innovative and smart financial services using the latest technologies and deep learning algorithms. With these new financial services, end users will be able to track their expenses, switch from one account to another, compare the fees of their accounts and aggregate data of multiple accounts in one single screen. To provide these services, fintechs need access to payment accounts from banks. However, as these startups pose a competitive threat to established corporations, banks may try to find ways to prevent such access.

This is exemplified by the AFAS case (Rechtbank Midden-Nederland, 2014). The central issue of the case arose when AFAS Software asked the customers of ING Bank to enter their credentials on the website of AFAS to log on directly to ING’s secure online banking interface and initiate the transactions on their behalf. The Dutch bank claimed that its general terms and conditions do not allow customers to share or disclose their data (credentials…) to a third party and therefore denied AFAS access to this information. The bank also insisted on the security risk behind such a request. On the other hand, AFAS argued that it tried to anticipate the future implementation of PSD2 with all its implications in term of competition, single market, data privacy. The decision of the Dutch court was in favor of ING bank considering credentials as sensitive data that shouldn’t be disclosed for security purposes. It added that since PSD2 is not yet implemented in Netherlands, the argument of AFAS cannot be taken into consideration.

This post aims to explore the possible implications of the implementation of PSD2 on future litigations between Banks and startups/fintechs.

PSD2 in a nutshell:

As part of the Digital Single Market Strategy, all EU members will have to implement Directive 2015/2366/EU known as PSD2 ‘(Payment service directive 2) by January 13, 2018 (Payment services in the EU, 2016). The purpose of PSD2 is to amend the previous directive (PSD) that has been in place since 2007 to fix some open points such as the lack of security and standardization of payment solutions, the several generic exemptions in the directive, the lack of interoperability and the difference in transaction fees across the European Union.

The main objectives of PSD2 are:

  • Secure the payment procedures
  • Reduce the costs for customers
  • Increase the efficiency of the processes through the standardization of infrastructures
  • Unlock customer data by providing access to account information for fintechs (third parties)
  • Increase the efficiency and integration of the European payment market
  • Improve customer protection

For consumers, PSD2 will have mainly good repercussions as it will increase convenience (websites, mobile apps) and transparency, reduce considerably the vulnerabilities (breaches) of the system and diminish the costs of transactions/services. In addition, consumers would be able to consider offers from other countries which will strengthen their bargaining power.

Banks will have to share their customer data with third parties. Consequently, they will have to bear the cost of developing the necessary APIs and compete with new actors such as fintech and startups. As a result, the directive might promote innovation in the banking industry, reduce the architecture complexity in the IT department and simplify the scalability through European countries with the harmonization of the legal frameworks in the EU. European banks showed their skepticism towards this directive by the voice of the European Banking Federation (EBF). This federation which represents 32 national banking associations stated in 2015 that PSD2 “provides yesterday’s solutions for tomorrow’s problems” (The European Banking Federation, 2015).

The big winners seem to be fintechs and third parties who can improve their customer relationship and have access to APIs from banks without taking in charge the necessary cost of implementing protocols and IT tools.

The next two sections will discuss some arguments that the parties of the AFAS case could have used if the PSD2 was implemented (an analogy could be made for startups Vs Banks).

Arguments for AFAS (startups/fintechs)

a- Competition / Single market

In order to encourage competition, the Commission tried to modernize PSD by adding some amendments to it. As a result of PSD2, new players will have better opportunities to enter the banking market and propose innovative, cheaper and user friendly solutions to consumers. The statement made by the European Banking Federation (EBF) on payment service agreement is one of many initiatives made by big Banks to influence the national courts and their interpretation of the directives in this field. If TPPs cannot have access to the data of customers and initiate payments, it will limit their ability to compete with the established banks.

As the PSD2 seeks to improve competition in the Single Market, the court would probably have ruled in favor of AFAS if the PSD2 had been implemented.

b- Contracts / Unfair commercial Practices

One of the main arguments of ING bank was that its general terms and conditions do not allow their users to provide data to TPPs. Nowadays, customers are considered as responsible citizens who are supposed to be aware of the terms and conditions of a contract including electronic contracts (Trzaskowski, Savin, Lundqvist & Lindskoug, 2015). However, it is almost impossible for the “average consumer” – defined as reasonably well-informed and reasonably observant and circumspect (The European parliament and the Council of the EU, 2005) – to fully understand all the implications of the different points within these terms especially those which require technical background. Also, the annex of the COUNCIL DIRECTIVE 93/13/EEC states that a term may be regarded as unfair if it is “irrevocably binding the consumer to terms with which he had no real opportunity of becoming acquainted before the conclusion of the contract” (The Council of the European Communities, 1993) .

As a result, these contracts might be considered as unfair given that they contain unfair conditions that are used to “mislead” the customer. Accordingly, the violation of the terms and conditions argument becomes flawed.

c- Data protection

PSD2 aims to reduce the risk of fraud and protect confidentiality by securing the payment process. One of the fundamental notions to reach this goal is strong customer authentication. Basically, the idea here is to use at least two of the three types of security elements to validate any transaction:

  • Knowledge( something known by the user, e.g. PIN code)
  • Possession (something only the user possesses)
  • Inherence (a proper characteristic of the user e.g. Voice recognition, fingerprint…)

Moreover, these providers will only have access to necessary data explicitly consented by the user under this new directive. The credentials will not be accessible to TPPs but will be transferred efficiently through a secure channel using APIs. Thus, the fears of banks concerning security become unsubstantiated and cannot be taken into consideration for the final judgment if PSD2 is implemented.

Arguments for ING Bank (banks)

a- Broad language leading to different interpretations

As a directive, PSD2 shall be implemented in all EU countries but not necessarily with the same wording. In fact, a directive is defined as a formal and usually mandatory executive order or official pronouncement on a policy or procedure. In that sense, countries have to implement it but they have the freedom to adapt it to their legal framework as long as the main objective behind the directive is fulfilled. This situation may lead to different interpretations of PSD2 due mainly to the broad language used in some parts. An example is the 58th article in which “all information” is not clear enough and the sources of data that should be disclosed are not enumerated. ING bank could exploit these ambiguities to push for the adoption of national laws in favor of banks thereby possibly affecting the outcome of similar cases.

b- Commercial practices / Conduct risks

Last year, one of the most interesting cases was the one involving Wells Fargo. The bank was asked to pay more than 185 million$ as fines after violating the Unfair, Deceptive, or Abusive Acts and Practices (UDAAP) defined by the Dodd-Frank Act of 2010 (The 111th United States Congress, 2010) . The US bank created accounts for customers without their request.

This is an example of “conduct risk” that European banks have to take into consideration given the complexity of their ecosystem. With PSD2, many believe that banks will converge to an API-based architecture. Consequently, new conduct-related risks will emerge especially regarding financial operations (loan, deposit, investment products, mortgages…). Allowing TPPs to have access to their APIs would oblige EU banks to trust the judgement of TPPs when it comes to controlling the integrity of the transactions. This situation would raise ethical issues such as the possibility for gambling third parties to initiate loans for users using their app. The article 3(3) of the “Unfair Commercial Practices directive” (The European parliament and the Council of the EU, 2005) defines vulnerable consumers as a clearly identified group of consumers who are particularly vulnerable to the practice or the underlying product because of their mental or physical infirmity, age or credulity in a way which the trader could reasonably be expected to foresee. In this particular case, the simple consent of a “vulnerable consumer” shouldn’t be enough to permit the initiation of transactions. This risk can be used by ING bank to justify its position in the AFAS case.


We are entering a new era where all the cards are redistributed in the banking industry. We are likely to see new startups becoming the next big actors of payment services especially with the future implementation of PSD2. It appears that if this directive is implemented, the decision of the national Dutch court would have been in favor of AFAS as the arguments of ING bank are either not enough substantiated or not based on the “average consumer”.

Despite its position in this particular case, ING bank is already taking strategic choices and adapting its future business model and value proposition to this new paradigm by launching “ING Innovation Studio”; an accelerator to support fintechs; in an attempt to promote cooperation with innovative startups.


Payment services in the EU. (2016). Retrieved from http://eur-lex.europa.eu/legalcontent/EN

Rechtbank Midden-Nederland. (2014). ING BANK N.V. v. AFAS SOFTWARE B.V C/16/372291 / KG ZA 14-481.

The 111th United States Congress. (2010). Dodd–Frank Wall Street Reform and Consumer Protection Act.

The Council of the European Communities. (1993). COUNCIL DIRECTIVE 93/13/EEC. The European Banking Federation. (2015). EBF statement on EU payment services agreement. 1.

The European parliament and the Council of the EU. (2005). DIRECTIVE 2005/29/EC Unfair Commercial Practices Directive.

The European Parliament and the Council of the EU. (2015). DIRECTIVE (EU) 2015/2366 OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL. Official Journal of the European Union, 93.

Trzaskowski, Savin, Lundqvist & Lindskoug. (2015). Introduction to EU Internet Law.