The Low Code/No Code did not appear in the early 2020s, it is a trend that has already been taking shape for several decades and in 2021 the market is still in the creation phase. There are a very large number of well-known players such as Salesforce, OutSystems, K2, Mendix, ServiceNow, Appian, AgilePoint, or Microsoft PowerApps. However, the VSE/SME market is in the process of being structured with players offering high-quality products such as Webflow, Airtable, Zapier, Notion, Integromat, Bubble, or Parabola.

According to ResearchAndMarkets’ studies, the Low Code/No Code market was worth +4 billion dollars in 2017 and will probably exceed 27 billion dollars in 2022. These figures and orientations are confirmed by Forrester, which announces 15 billion dollars in 2020. These trends are the sign of a real market that is firmly established for tools aimed at large companies that wish to launch projects quickly, but also at small companies that are undertaking a digital transition, all dominated by a context of a shortage of developers and a technology market that has become too complex [1].

The fact is that these development environments are little known, as are the different tools they offer and how to use them. We are therefore in the “pedagogical” phase of the business that will, over the next ten years, put in place the different solutions and the market structure to support their economic development. A survey conducted by the Forester Institute in 2019 confirms this trend.

Low-Code/No-Code environments are used for :

• Rapidly develop simple services for reduced functional perimeters
• Establish software universes for “peripheral” extensions to the functionalities of business solutions (scheme adopted by SalesForce, Microsoft, SAP).
• Develop a “specific” engineering service to support the growing demand for personalization/customization.

Introduction

Since the emergence of software modeling techniques and more particularly since the creation of UML (Unified Modeling Language) in 1997, one vision has occupied the minds of software engineering specialists: to produce applications without writing a single line of source code. At the end of the 1990s, it was the use of modeling techniques to formally describe all the facets of a software application (interface, behavior, data model, network architecture, etc.) that prevailed. This should de facto allow the most automated source code generation possible. The principles defended by software engineering, therefore, advocated concentrating the effort on software design while sparing oneself the purely technical difficulties linked to the compilation, programming errors, or the slowness caused by profound architectural modifications.

This ultimate goal was never achieved. For if in theory there is no obstacle to this type of approach, in practice the complexity factors of modern software applications make this type of approach inoperative. The multitude of models that have to be produced to describe all situations and all cases have made development through modeling more tedious than direct writing of source code. As early as 2004, a number of players have been pointing out the ridiculous nature of the verbosity of UML and software development where everything is a model [2].

From the 2000s onwards, a very large number of solutions allowing end-users to build their own software have been very successful. For example, Adobe DreamWeaver, a website editing software offering a “What You See Is What You Get” (WYSIWYG) interface, allowed neophytes of HTML, CSS, and Javascript to graphically build and publish simple web sites (static pages and media). In the same way, a large number of so-called visual programming solutions were created. For example, some of the tools offered within MatLab, LabView, or the Scratch programming language developed by MIT have enabled programming beginners to start building programs within a predefined framework and for a range of very specific uses. These solutions have also met with increasing success in the fields of industry, education, video games1, and many other citizen domains [3].

These successes, confined to specific disciplinary fields, are now widening their scope. This phenomenon is also supported by a very favourable context:

• The inexorable expansion of digital technology and the exponential growth in the quantity of mini-software, mini-services developed and available in the form of mobile applications and websites.
• Standardization of our usage, user interfaces, software architecture styles, and APIs that allows developers to focus their attention on the purpose of applications (the what) rather than the tools to be implemented (the how) for a large number of applications offering standard services.
• The shortage of developers in an increasingly complex technological context.

We are therefore witnessing a shift in vocabulary. We no longer speak of modeling, meta-modeling, end-user programming (EUD), code generator (total or partial), component programming, graphic interface builder (i.e. UI Builders). This universe of methods and tools is now grouped under the names of Low Code/No Code (LCNC). Nevertheless, the fundamentals remain the same and we will detail these approaches, how to benefit from them, as well as their advantages and limitations.

1.https://docs.unrealengine.com/en-US/ProgrammingAndScripting/Blueprints/Editor/index.html

Origins and History

In the Beginning: WYSIWYG Interfaces!

The promise associated with Low Code/No Code is not new. As early as the 1970s, programs have favored the display of rendering rather than the visualization of code fragments that can only be understood by computer scientists. In this field, Bravo (1974) is probably one of the first WYSIWYG (What you see is what you get) editors. Since then, these solutions have continued to expand, from text editors to tools for designing websites that are nothing more than solutions providing advanced functionalities without coding.

Major contributors include Low Code/No Code, GeoCities in the 90s, as well as complete CMS (Content Management Systems) such as Drupal, Joomla, Dotclear, and WordPress that appeared in the 2000s. For almost 20 years, they have been enabling non-technical profiles to create and maintain a website. The popularity of WordPress has also pushed companies to design powerful tools to build sites based on this CMS in a simple drag-and-drop. We can specifically mention “Page Builders” such as BeaverBuilder [4] or Elementor [5], which are very popular with small businesses and marketing departments.

Figure 1 Two examples of a No-Code interface: On the left is the Elementor web page builder for the CMS WordPress. On the right the BotNation chatbot configuration/construction interface.

The Low Code/No Code trend is to be seen as a simple extension of these first steps in the field of web page construction to a more varied set of solutions. Applications in these new environments include chatbot creation [6], [7], [8], web data scraping [9], [10], business process automation or RPA2, and even mobile application design. For about 50 years, the principle has remained the same: a software environment editor provides a tool enabling a non-technical and/or technical individual to easily create a solution that, at first sight, involves a very high level of technical complexity.
The example of Web Elementor is very symptomatic of these new work environments. This web page builder is the “star” for building sites using WordPress and has been growing steadily since its release in May 2016. It has more than 2 million active installations and is helping to build a freelance world with little technical expertise to contribute to its deployment. WordPress is therefore the application software base that has served as the basis for this Low Code tool to develop a completely functional and economic model in just a few years.

2.Voir état de l’art Berger-Levrault sur la RPA : https://www.research-bl.com/2020/06/08/rpa-report/

Next: Major Technological Advances

Other more conceptual technological advances have also contributed to this movement. A first factor comes from the fact that over the last 15 years, an increasing number of solutions have used the creation of APIs to set up digital “services”. The universe of developments has therefore created a layer that has made life easier for other developers and users while establishing a logic of “microservices”. A second factor has been the advent of documentary databases, sometimes also wrongly referred to as “NoSQL” databases. Instead of storing data in rows and columns, these databases generally use a document-type structure similar to JSON (JavaScript Object Notation), a format popular among developers. These databases have enabled unparalleled flexibility while retaining the critical functionality of tabular and relational databases [11].

The principle of no-code appeared several decades ago, as technology progressed it allowed first the creation of simple applications such as web pages and then “richer” software such as mobile applications or automatisms.

Low-Code / No-Code (LCNC): General Principles and Definitions

The ambition of Low Code/No Code (LCNC) approaches is to enable the rapid development of simple applications by minimizing the writing of source code as much as possible. To achieve this, a large number of solutions rely on visual creation interfaces integrated into a software development environment. Each stage of the application lifecycle is automated and simplified as much as possible for users, from creation to commissioning. This enables the rapid delivery of usable software solutions.
One of the most important induced effects of the LCNC is that it removes the traditional development silos since this process leads, in theory, to a wide range of stakeholders (business, UX, marketing, technical) being able to produce all or part of the solutions.

Continuum Low-Code / No Code (LCNC)

The difference between low code and no code is quite blurred. A simple way to conceptualize it is to imagine a continuum where both ends would be no-code and full-code. On the one hand the ability to develop applications without coding and on the other hand the ability to code the development of classic applications by writing the vast majority of the source code. We, therefore, propose the following continuum to clearly define the LCNC phenomenon.

	FULL-CODE	SOME-CODE	LOW-CODE	NO-CODE
DESCRIPTION	Create applications by coding the majority of the application	Create applications by coding but reusing/configuring as many off-the-shelf components as possible	Create applications essentially without coding and integrate parts of code created elsewhere	Create applications without coding
WHO	Expert developers	Expert developers	Low-skilled, non-developers	Non-developers
HOW	Tedious but precise development	Rapid development when the right components exist	Rapid development	Rapid development
WHAT	Anything is possible	The choices are constrained by the components used	Limited to the possibilities of the low-code platform and the APIs that can be used to integrate custom code	Limited to the possibilities of the no-code platform

Once this continuum has been established, two phenomena are notable in this full-code to no-code progression: 1) the increasing speed of development made possible by the LCNC approaches and 2) the growing limitations induced by the use of LCNC environments which are by nature partitioned.

The use of an LCNC approach means making strong trade-offs between :

1) development efficiency

2) limitations in terms of choice of architecture, technologies, user interfaces.

The details that characterise the LCNC approaches can be summarised as follows:

• No Code: Using these environments allows you to create applications without coding. That is to say that the platform encodes the whole application automatically according to the configurations and settings defined by the user. These solutions are initially designed for non-developers who do not know or do not need to know programming languages in order to use and develop software. The suppliers of these solutions have integrated all the functionalities considered essential by users to create their applications. These environments are similar to popular blog platforms and “all-in-one” web design solutions (e.g. WordPress, SquareSpace, etc.).
• Low Code: Such a platform will be able to automatically generate the application under the same conditions. However, it theoretically allows integrating parts of code made elsewhere when necessary. These environments are slowly becoming a means for developers to help them design applications faster and with a minimum of manual coding. Existing platforms offer pre-configured feature suites and tools that complement the needs of developers. In very simple terms these environments allow visual blocks of existing code to be dragged and dropped into a workflow to build applications quickly. This process can completely replace the traditional method of manually coding an entire but simple application. This allows developers to produce more efficiently and quickly, without the need for repetitive coding. They can focus on the essential differentiating elements of their application.

Reasons for the Current Emergence of the LCNC

The vectors that are driving today’s growing demand for LCNC universes are well known and show marked differences from the 1990s. Indeed, the number of companies that need partially customized solutions (or wish to establish specific additional functionalities) is constantly growing. This movement is combined with an equally growing shortage of qualified developers. Finally, the generalization of the use of software tools in all sectors of activity is producing a multiplying factor for the two previous vectors of growth in demand. This set of constraints has brought to the forefront the adoption of so-called “low code” development platforms.

It is also necessary to add a much greater maturity of the technical fundamentals of the environments offered today. It has been acquired over the last 20 years through progress in research and technological innovations in both software and hardware. Moreover, the generalization of development tools and methods that tend to become more uniform is a facilitator of the adoption of these tools. Moreover, the creation of computer code loses its appeal over the decades and ultimately represents a brake on industrial developments in the field of information technology. The last decade has been marked by “platforming”, the use of “ready to use” frameworks, and the growing importance of architectural needs (urbanization of developments).

Finally, it is the development of solutions exploiting AI that is the most advanced field in the use of Low Code/No Code environments. This field of application is made up of mathematical complexities, models, the need of organizing massive data flows, and the frequent interlinking of several models (in cascade) to satisfy a need. It is therefore often an “aggregate” of pre-stressed models that are produced with very little code production. AI is a perfect illustration of an extremely important underlying phenomenon that goes relatively unnoticed for the moment: the exponential increase in complexity of our developments. Mobility, ubiquity, modeling, personalization, dematerialization, connected objects, data lakes, hybrid hosting, dozens of languages, changes in habits (sales, work, transport, consumption, travel) have built a universe of indescribable complexity. We are reaching a tipping point in the global complexity that humans (even in groups) can no longer grasp, let alone master, in order to conceptualize digital solutions. This is favorable terrain for the deployment of these Low Code/No Code universes.

Finally, it is necessary to analyze the application uses. Business applications generally evoke solutions such as CRM or ERP. However, these systems represent only a small part of the digital ecosystem with which employees interact on a daily basis. In fact, they use dozens of applications designed to meet specific use cases. For example, marketers use many tools to track their customers, automate email marketing, perform data analysis and reporting, manage social network planning, track content paths, manage campaigns and events, and more. Each of these tools raises its own issues: cost, training requirements, security, user adoption, IT administration. According to Blissfully’s “SaaS Trends 2019” report [12], an employee uses an average of eight applications per day. Organizations with 500 to 1,000 employees use an average of 151 applications, and this figure rises to 203 applications for companies with more than 1,000 employees. This is one of the reasons why Low Code/No Code tools have become a major asset. Whenever there is a need for new applications, organizations are then able to develop highly customized, scalable, secure, and cost-effective solutions in-house.

The growth in demand for specific tools, the shortage of developers, the increasing maturity of technologies, and the rationalization of development standards justify the emergence of LCNC approaches in 2020.

The LCNC can now cover a very large number of software service developments. And this trend will continue to grow, in particular thanks to the possibilities offered by AI.

Low-Code Manifesto

The overall framework of the “contract” carried by the LCNC movement is laid down by the “Low Code Manifesto”[13] which sets out the following 9 main rules or principles of the Low Code movement :

While the principles (in green) Cloud-Based, Community, Governance & Control, Innovation, Openness, Agility, Multi-users are fully aligned with the standards of “modern” software development, the first two principles are the real differentiators of the LCNC that we detail below.

Model Driven Engineering (Model Driven Development)

The metamodel is therefore comparable to the grammar of a programming language in the sense that it is used to describe a set of correct models. Thus, as it is possible to consider all the words and sentences accepted by a language, it is possible to consider all the models conforming to a metamodel. These models form part of what is called a model space incorporating a graph. The latter is composed of a set of models (nodes) and a set of updates or deltas (arcs) linking the models. As the syntax of a language is not sufficient to build correct programs, the conformity of a model is often not sufficient to qualify a model as valid. It is the semantics of a programming language that determines whether a syntactically correct program is executable. Similarly, a mechanism is needed to specify which of all possible models are valid [14].

Another important issue is that of model transformation [15]. It is essential in order to make models operational for code generation, documentation, and testing, as well as their validation, verification, and execution. It is possible to schematize the design chain as a series of stepwise transformations that refine models from the abstract to code generation. These transformations mainly use dedicated model transformation tools that specify the transformations through a set of relationships expressed in a language. Despite the diversity of these tools, the OCL (Object Constraint Language) has emerged as the common component of all approaches.

The Model-Driven Engineering approach is at the heart of the OCL solutions. Although the LCNC approaches do not require knowledge of programming, they do require a basic understanding of the syntax and tools they offer. The learning curve is therefore not zero. De facto, the creator of a Low Code/No Code software does not develop but builds the model of his application. The latter is generally interpreted or compiled to generate the source code that will actually allow the software to run.

Make no mistake, Low Code/No Code solutions only mask the source code, which is produced indirectly.

Collaboration and Multidisciplinarity

Collaboration between the different business stakeholders is essential since silos are disappearing and in the future, a marketing manager will be able to produce functionalities (in theory) [16]. The handling of a software creation tool remains a long-standing scientific objective which manifests itself under the name End-User Développement, a discipline in its own right at the crossroads of modeling, software ergonomics, and design [17]-[19].

Indeed, the visual development environment of an LCNC platform is specially designed to solve communication problems and thus strengthen collaboration between team members, regardless of their field of expertise or technical acumen. With a robust visual language and iconography, everyone can visualize the problem to be solved, the need to be addressed, and the tools and resources available to build a solution. Visual language is the lingua franca of the entire lifecycle, from problem definition and solution exploration to application creation, testing, and deployment.

Working with a common visual language, contributors can share a screen to try out ideas, try to solve difficulties, or refine an interface. All parties can understand the nuances of the discussion and make meaningful contributions because they see everything before their eyes. There is no need to interpret code or translate PowerPoint.

Because the learning curve for visual language is theoretically short, team members quickly feel empowered to contribute to aspects of application development that are outside their core competencies. A business analyst or product engineer, for example, may actually create an application themselves or add, subtract or reorganize the components that make up the application. Conversely, an experienced developer can bring a fresh look at a business process or customer interaction scheme and come up with innovative ideas for optimizing efficiency or user experience to create greater business impact. What’s more, everything happens in real-time in a truly open and collaborative environment.

When salespeople and developers speak the same language, thanks to the visual model, no translation is required, everyone understands the issues as they are presented and the solutions as they evolve, iterations are fast and everyone remains invested in a clearly defined process from the start of the project.

Important Limitations

Of course, the LCNC movement does not offer an ideal solution for all cases either, and there are many limitations at present. The main limitations at present are as follows:

• Some platforms allow the integration of existing tools, but these options are generally minimal, and few environments offer them. This is therefore a major technical limitation.
• To use a Low/No-Code platform, one must question the intellectual property of the software created. With current platforms, we often only have a right to use. For software publishers, this is therefore a limitation that is difficult to accept.
• In the event of a change of tool, migration to another Low/No Code system is very difficult (or even impossible). What is created with this type of solution will only be very partially exportable and transferable elsewhere, as no LCNC standard exists to date.
• The possibility of editing the source code and adding functionalities that would appear necessary is very often impossible.

These environments are very similar to the Computer Aided Software Engineering (CASE) of the 80-the 90s (Access, PowerBuilder, Clarion, WinDev). The freedom to create and increase productivity comes at the price of dependence on a specific environment with all the disadvantages that are already known with CASEs.

Furthermore, the success of this environment is also explained by the maturity of hosting offers in the cloud, especially “Platform as a Service” (PaaS). In addition, no code / low code platforms benefit greatly from the trend towards opening up many web applications to third party environments, facilitated by the Application Programming Interfaces (APIs) that they make available to all. By this means and via the “consumerization” of their services, these platforms are part of the transformation of business process management and document flow management.

According to Alain Faure and Laurent Sollier of the IT architecture firm Octo Technology, “the pressure induced by the digitalization of activities is pushing companies to erase the division that has long been made between IT services and business services”. However, even if the market and demand grow strongly, a good number of companies will continue to be confronted with problems that will not find standardized answers and that will require IT development in the strict sense of the word. “Twenty years ago, enterprise systems were implemented by hundreds of consultants and experts at a cost of several million dollars,” explains Simon Chan, founder, and CEO of DigiVue Consulting. “Today, thanks to no code / low code, the same systems can be implemented by smaller players at a fraction of the cost and in a much reduced time.

LCNC environments are very similar to the CASEs of the 80s and 90s. The freedom to create and increase productivity comes at the price of dependence on a specific environment!

LCNC : Threat or Opportunity for Software Publishers?

Bonitasoft, Appian, Pega, and Salesforce via Einstein Automate join forces to offer solutions for the design of critical applications exploiting hyper-automation. This is the next stage of automation, based on the use of new technologies such as artificial intelligence, RPA [20], and machine learning to robotize processes. In some cases, it may involve the creation of a digital twin to visualize how functions, processes, and performance indicators interact to generate value.

Among the tools that are advancing, intelligent business management software is gaining popularity. It appears that by 2025-2030, it will be necessary to “build as many applications as in the last 40 years” (Quote from Appian Technology’s CEO). In order to overcome this difficulty, many companies are favoring the “robotization” of software development processes in order to concentrate existing developers on much more productive business and functional value creation tasks.

In this context, players such as Mendix (owned by Siemens) and OutSystems (a major player at Volkswagen, Mercedes-Benz, Schneider Electric, Honda, Intel, HPE) are primarily targeting manufacturers to offer them an ad-hoc production approach. This movement will, in the long term, short-circuit the software publishers. If this trend is confirmed, it is quite logical and conceivable that this process will continue for a large number of software publishing domains, and more particularly sectors such as management.

LCNC solution providers target manufacturers as a priority to offer them an approach enabling them to produce their software solutions themselves. Financial and insurance specialists have already begun this transformation.

If this trend is confirmed, this short-circuit movement will eventually affect a large number of areas of software publishing, and more particularly whole sections of the management sector.

As of today, finance and insurance specialists seem to be really mastering these platforms. Retail groups are beginning to take up the concept, such as Cdiscount, which has taken advantage of the deployment of its Pegasystems CRM to create associated applications using the publisher’s bricks and mortar. Another classic example frequently mentioned concerns HR teams who, in order to welcome an employee, must collect information from several sources and then enter it into one or more applications in order to create the employee’s accesses, prepare his medical visit, send him his welcome letter, put him in touch with several company departments or send information to the payroll management service. A robot can do all these administrative and repetitive tasks and an LCNC platform opens up the possibility of customizing this task from start to finish for each company at a lower cost.

Specifically for software sectors with a high business or regulatory content, the extremely rapid and constant progress in automatic natural language processing (NLP) techniques will combine inexorably with the CNCL movement. Indeed, hyper-automation will apply to the massive integration of text corpora, which are themselves now widely available in digital formats directly accessible to machines. The junction of the two movements will open up a breach in many of the “protected” areas of software publishing to date.

For the time being, we can note that the use of these LCNC platforms on a large scale requires not only deploying them but also building a whole methodology. This will require the training of seasoned, much more expert developers capable of setting up these “new development factories”.

Solutions & Market Trends

The main market trends are as follows:

Gartner’s “Guide to Evaluating Codeless Development Technologies” published in February 2019 predicts that “by 2024, more than 65% of the world’s software going into production will be created using codeless configuration platforms”.

In 2020, half of the developers plan to use a Low Code platform (Gartner 2019)

By 2024, 75% of large companies will use at least four Low Code platforms for IT application development (Gartner 2019)

Forrester predicts that the market for codeless development platforms will grow from $3.8 billion in 2017 to $54 billion in 2024.

Renowned publishers have been relying for years on the contribution of development platforms that reduce the increasing complexity inherent in coding. Historically, the first LCNC (Low Code/No Code) tools were mainly aimed at application architects. The main aim was to simplify the work of developers through interfaces and declarative programming languages.
However, these tools have now become increasingly popular with players such as Pegasystems, Mendix, Appian, and Salesforce. For example, SalesForce offers in the new version of its CRM a No Code tool called FlowBuilder [21]. FlowBuilder allows processes to be graphically created, deployed, and executed. Such a configuration historically reserved for software experts and/or developers is now fully accessible for non-technical profiles such as human resources, marketing, and sales.

More broadly, startups and Business Process Modeling (BPM) specialists have simplified their products to offer similarly no-code environments. The WordPress CMS, available since 2003, has been followed by Shopify, Zapier, Airtable, Notion, Bubble, or Coda.

Despite this craze, the low code concept did not finally find a name until 2014 in Forrester analysts’ articles. It was around the same time that the notion of “citizen developer”, imagined in 2009 by Gartner, began to make sense. The two consulting firms perceived several phenomena. On the one hand, developers trained in fourth-generation programming languages began to use visual environments to build their software. On the other hand, employees wishing to build business tools have sometimes assimilated these languages and have found their happiness in this UI (User Interface). The lack of programmers, the slowness of evolutions or developments, and the growing need for agile application design now ensure the success of the LCNC approach.

Although the principles of the LCNC have been in existence for a long time, the trend really took hold from 2010 onwards and is now becoming more widespread.

There are a very large number of commercial and open-source LCNC environments, including 2 French universes.

A first quick observation is that there are many so-called LCNC solutions on the market. As shown in the map below, it is nevertheless important to distinguish between solutions that allow a product to be built in its entirety (Product Builders) and generic components (Base Blocks) which are components of an information system that can be configured in Low Code/No Code.

Application platforms provide runtime environments for application logic. They manage the lifecycle of an application or application component and ensure the availability, reliability, scalability, security, and monitoring of application logic.

In 2020 the most mentioned actors in the field are the following:

A Low-Code Application Platform (LCAP) is an application platform that supports rapid application development, deployment, execution, and management in a single step using high-level declarative programming abstractions, such as model and metadata-based programming languages. They also support the development of user interfaces, business logic, and data services, and improve productivity at the expense of portability between vendors, compared to conventional application platforms. Gartner considers no-code application platforms to be part of the LCAP market [22].

Business Solutions

Without going into the precise details that make up many commercial environments, it is possible to describe the characteristics of some of the most recognised:

• Salesforce – Lightning. Salesforce’s aPaaS solution, called the App Cloud, is a toolkit that includes the Force.com proprietary application creation platform and Heroku’s proprietary application creation platform. While Force.com operates in Salesforce’s data centers, Heroku’s is hosted at AWS. A gateway, Heroku Connect, has been developed to ensure data synchronization between Heroku and Force.com. This is a perfect illustration of the ‘consumerization’ and platform station of solutions. It was in 2015 with Lightning that Salesforce truly entered the low code world. Inaugurating a graphical overhaul of Salesforce’s SaaS offering, Lightning above all introduced a different way of programming, evolving Visualforce, Salesforce’s UI, as well as Force.com and its proprietary language (Apex), which is strongly inspired by Java. Its Process Builder and Visual Workflow modules allow Lightning to instantiate business bricks and assemble them into a workflow. The user can use pre-integrated components from Salesforce or those provided by third parties via the AppExchange marketplace. SaaS bricks can be configured 80 to 90 percent without code. However, for the implementation of complex and specific business rules, code and Force.com are required.
• Appian – Quick Apps. This company, created in 1991, is positioned at a crossroads about its competitors. It is mainly known for its business process management (BPM) and case management solutions. It has extended its expertise to low code in a fairly natural way. Laurent Chailley, Appian’s Sales Director for France, believes that the expertise of these two “historic” businesses (process automation and data management) enables Appian to industrialize and securely deploy an application on the scale of an international group. Appian concentrates its R&D in Restone, Virginia, and has customers in Sanofi and BNP Paribas. Appian promises to create apps in 15 minutes. In November 2020, the company released the third version of Quick Apps Designer, a graphic design tool that requires no technical skills. It guides the user through the major steps of implementing an application (data management, forms, processes…). The editor offers various solutions: on-premise, hybrid, or public cloud modes. The latter choice uses AWS’ infrastructure.
• Mendix. Founded in 2005 in the Netherlands, Mendix has now moved to Boston, USA. The editor offers a modeling tool based on a visual development environment. It claims more than 3,400 client organizations worldwide but does not specify the proportion of users of its free version. The latest version of Mendix focuses on the rapid development of smart apps for the IoT by offering ready-to-use connectors with AWS IoT, IBM Watson, and LoRaWAN. Applications built with Mendix can be deployed on AWS, IBM Bluemix, or Microsoft Azure infrastructures. The use of Cloud Foundry and Docker technologies ensures their portability to all types of Clouds, whether public, private, or hybrid. Mendix offers a rather rich application shop with hundreds of components and ready-to-use applications.
• OutSystems. The company was born in Portugal in 2001 and has since migrated to the United States, precisely to Atlanta. The publisher is, like Mendix, what we call a pure player. Its PaaS would have enabled the rapid development of more than 100,000 applications throughout the world. Its platform, based on a metadata model, automatically generates Java or .Net code by dragging and dropping components. The web or mobile application thus created can be completed with code written in C#, Java, JavaScript, SQL, CSS, or HTML. Once development is complete, the application can be deployed on a Cloud (private, public or hybrid) or an internal server (on-premise). The OutSystems Forge library offers connectors for popular cloud services such as Zendesk or Twilio. OutSystems is a member of CISQ (Consortium for Software Quality). GFI recently acquired Roff, a specialist integrator of OutSystems.
• ServiceNow – Application Development. As a specialist in business processes, this company was naturally interested in low code, thus extending its historical target (IT departments) to business management. Its service portal, Application Development, allows, among other things, and via an in-house development environment (Studio), to create a web or mobile application using a set of predefined components and templates. The Studio offers workflow type functionalities allowing the integration of existing systems and forms. The application can then be deployed by ServiceNow’s PaaS. The editor offers a series of APIs and integration tools. In 2015, it launched its CreateNow store, as well as a developer program targeting both experienced programmers and low code enthusiasts.
• Microsoft – Power Platform. Power Platform’s ambition is to offer an integrated suite of tools enabling companies to develop applications from A to Z without having to go through IT. Microsoft brings together in this platform three pre-existing offers: Power BI, PowerApps, and Flow. PowerApps is a low code development environment for designing Web or mobile applications without programming knowledge via a 100% graphical interface while integrating it with third-party software and data sources. Flow is considered as an RPA (robotic process automation) tool for designing, deploying and orchestrating business processes. The strength of Power Platform is clearly the “bouquet” of some 250 connectors covering both Microsoft (Office 365, Dynamics 365) and non-Microsoft software (Adobe, GitHub, Salesforce, ServiceNow, Slack, Zendesk…). This is a decisive step taken by Microsoft in terms of office automation and integrated collaborative workspaces.
• Amazon Honeycode. Honeycode is AWS’s new (June 2020) low (or no code) application development solution to counter Google Cloud and Microsoft Azure. Honeycode is a spreadsheet model assisted by templates and many integrated functions. Clients can use a simple visual assistant to create highly interactive web and mobile applications using a powerful database built by AWS. This database enables tasks such as tracking data over time, notification of changes, routing approvals, and facilitating interactive business processes. Application complexity varies from a task tracking application for a small team to a project management system that manages a complex workflow for multiple teams or departments.

Open Source Solutions

It is perfectly possible to produce an LCNC platform by exploiting relatively comprehensive open source technologies. A production “workshop” could very well be established with the following elements:

• Acceleo is an open-source code generator for Eclipse used to generate textual languages (Java, PHP, Python, Javascript) from EMF models (Eclipse Modeling Framework) extracted from meta-models such as SysML or UML.).
• Actifsource is an Eclipse plugin allowing code generation via graphical templates and the use of templating methods.
• DMS Software Reengineering Toolkit that allows the definition of DSLs (domain-specific languages) that serve as high-level templates for translation in all text languages.
• GeneXus is a cross-platform code generation platform based on knowledge representations mainly oriented towards the creation of management software. Its use is based on descriptions in a declarative language with multi-language programming generation.
• OpenXava is an open source low code platform that allows the generation of complete management applications in Java.
• OSBP is a software factory provided in open source by the Eclipse Foundation. This environment combines Low Code and No-Code with classic development tools. Coding is replaced by the use of a descriptive model.
• Spring Roo is an Open Source Java code generation framework for addressing server-side considerations, particularly for separating different layers of software.
• RISE is a free modeling and code generation software suite for database and persistence aspects.
• Wavemaker is a Low Code platform that enables the production of large areas of server-side business applications by generating code that is largely based on market standards (Angular, Typescript, Java).
• Generative Objects. This is an open-source alternative to guarantee democratization of access to low-code via a GO-Low-code platform.
• Bubble. Bubble Group was founded in 2012 by Emmanuel Straschnov and Joshua Haas in New York. For the two founders, if coding is essentially similar to old-fashioned typesetting, then Bubble is the word processor for writing applications. It is a platform for creating web applications for computers or smartphones, without writing a single line of code. If the use of the online software is free, the cloud hosting of the application is paid for, in the form of a monthly fee.
• Convertigo. It is one of the few French low-code players with an international dimension. Its particularity consists in promoting open-source Low Code and relying on market standards to reduce technical adherence as much as possible. Convertigo is also a No Code player in a rapidly growing market. We asked Olivier Picciotto a few questions.

Some Remarkable Anecdotes

Unqork

New York City used a code-free platform to quickly implement a new anti-Covid web portal. To achieve this very quickly the city used Unqork’s LCNC environment. “There is no longer an army of engineers,” said Gary Hoberman, the founder of Unqork, who worked with the New York Department of Information Technology and Telecommunications to launch the new portal. Officials are using the website to try to gather information about where people with symptoms of the disease have potentially come into contact with others.

The Unqork platform gives clients the ability to quickly design web applications by configuring various datasets and widgets using a visual interface, rather than producing code. Hoberman sees his team as creating generic ‘Lego blocks’ that a company or government can customize to suit their objectives. New York City had used Unqork to develop other applications, including a food delivery service through which the city recruits taxi and limousine drivers to bring meals to people who cannot leave their homes and cannot afford deliveries from grocery shops or restaurants. The New York Food Bank, a non-profit organization, estimates that approximately 1.2 million New Yorkers are food insecure. There is also a web portal for donations of medical equipment, such as ventilators or personal protective equipment, as the city’s hospitals are overwhelmed.

In October 2020, Unqork also obtained $207 million in Series C funding, bringing the company’s value to $2 billion. The round was led by funds and accounts managed by BlackRock and joined by Eldridge, Fin Venture Capital, Hewlett Packard Enterprise, Schonfeld Strategic Advisors, and Sunley House Capital Management, a subsidiary of Advent International. Existing investors including CapitalG, Alphabet’s independent growth fund, Goldman Sachs, Broadridge Financial Solutions, Aquiline Technology Growth, and World Innovation Lab (WiL) are also participating. The company will use the new funding to continue to execute a number of ongoing initiatives, such as expanding its global sales force, investing further in technology, training, and go-to-market partnerships with the world’s leading service companies and systems integrators, including Capco, Cognizant, Deloitte, EY, KPMG, and Virtusa.
Unqork has been adopted by leading organizations in financial services, insurance, utilities, healthcare, and other industries. Clients include Aon plc, District of Columbia, Goldman Sachs, Liberty Mutual, Montgomery County, New York, Nippon Life, Pacific Life, Rethink Food, Vault, and more.

Bubble

Creating a start-up involves recruiting engineers to build the technological architecture of the desired platform. Recruiting such profiles can quickly become a headache because it is expensive and difficult in the face of strong market demand. In this context, it is also possible to program yourself but you still have to learn how to code. This learning process can quickly become tedious, to the point of making some entrepreneurs throw in the towel. New York-based start-up Bubble has therefore set itself the goal of simplifying the technological tasks of entrepreneurs. Founded in 2012, the company was previously self-financed, but after seven years of existence, it now intends to move up a gear to offer its technology to as many people as possible.

In June 2019, the American start-up announced a first round of financing of 6.25 million dollars from the Californian fund SignalFire (Uber, Lime). Business angels such as Ali Partovi, an investor in Airbnb, Facebook, and Dropbox, via Neo Ventures, and Farzad “Fuzzy” Khosrowshahi, creator of Google Sheets, both cousins of Uber’s boss, Dara Khosrowshahi, as well as Frederic Kerrest, founder of Okta, and the singer Nas, also participated in the operation.

Launched by Josh Haas and Emmanuel Straschnov, Bubble is a platform that makes it easy to create a website or a mobile application without having to learn how to code or call on a developer. Concretely, the graphical interface allows entrepreneurs to visually program the software solution they want to build while freeing themselves from the constraints of computer coding. Indeed, the user has access to a library of bricks of elements and actions that he can combine visually to build his site easily and quickly. Bubble makes it possible to program the equivalent of a social network like Twitter or a marketplace like Airbnb in a week’s work!
Bubble boasts more than 270,000 users who have created more than 230,000 applications and has a turnover of 2 million dollars in 2018.

Zapier

How to increase sales during the dull, grey winter season? A good strategy is to motivate buyers with personalized contextual messages. In cold winter weather, weather-based campaigns can be used to motivate customers to spend through tools such as a coupon code, free shipping, a gift card, or even additional loyalty points earned after placing an order. The challenge is to only target customers whose weather forecasts meet certain conditions? It is understandable that launching such a weather-based campaign may seem complicated and time-consuming, which is antithetical to reacting quickly to the weather.

Voucherify, a company specializing in the management of loyalty tools, promotes “weather marketing”! This approach exploits a powerful automation capability to analyze real-time weather data to trigger advertisements and customize marketing messages based on local weather. The company has focused on scenarios that can be set up in a matter of days to enable use as close as possible to seasonal changes. In order to establish an experience and set up loyalty campaigns (coupons, gift cards) based on global and local weather, using little or no code, it relied on the use of five API-first platforms. Each time the implementation took only a few hours, including the ideation stage.
One example is the promotional campaign for a coffee shop in Berlin. At the beginning of the winter season, users receive two promotional codes by SMS which they can only use if it snows (the first code is active if the temperature is above -15°C, the second if the temperature is below -15°C). The coupons are automatically deactivated or activated daily, depending on the weather forecast for Berlin. This check is carried out every day at 7 a.m. via Zapier automation. The coupons can only be used once per customer. The general architecture of the application is as follows:

Generative Objects

This project demonstrated the power of the French solution Générative Object. The LCNC tool of Generative Objects made it possible to build a solution for managing legacy, life insurance, and donation files.

The product offers a database, alerts, reliable and standardized data entry, automated calculations, automatically pre-filled and configurable documents, accounting allocation of income and expenses, and RGPD compliance.
With this example, Generative Objects demonstrates that their solution makes it possible to build management applications that have nothing to envy to those developed using traditional methods.

LCNC at Berger-Levrault

As we have developed throughout this study, the LCNC approach is not new, although it is now emerging commercially. The LCNC derives its concepts and technology from code generators and modeling environments. The Research and Technological Innovation Department has been using LCNC solutions for several years for a very large number of projects. Here are a few examples:

• Modeling and Code Generation: Our semi-automated migration projects (GWT to Angular, PowerBuilder, Access) rely on parsers and code generators as well as Model Driven Engineering techniques that are at the heart of LCNC solutions. For example, we use the Moose3 platform to analyze existing code, model an expected result and generate the related code.
• Graphical language for configuring a planning optimizer: Expressing business constraints and optimization functions to set up a planning optimizer requires strong skills in programming but also in mathematics. However, these elements are highly dependent on the business to be addressed, and very often users and POs (product owners) have the ability to express them in natural language. To allow them to set up a planning optimizer without writing code, we have developed a solution that uses a graphical DSL (Domain Specific Language4) to generate the mathematical formulations and the necessary code that will then be executed on a solver. This prototype illustrated below, uses Model Driven Engineering techniques, notably through the ADOxx5 meta-modeling platform.

3.https://moosetechnology.org/
4.https://en.wikipedia.org/wiki/Domain-specific_language
5.https://www.adoxx.org/live/home

Figure 6. Left: An example of a DSL or Domain Specific Language created specifically to configure scheduling optimization algorithms. Here the DSL is used to specify the constraints to be considered (right-hand column), to describe the data (on the map) and their interactions. The DSL then generates a mathematical intermediate formulation (right) which is translated into source code used by a solver. The transition from the visual to the mathematical environment, to the source code, is achieved using Model-Driven Engineering (MDE) techniques in the ADOxx environment.

• Interoperability Connector Generation: There is nothing more similar than two interoperability connectors. For each new software interoperability need, it is nevertheless necessary to develop a connector to adapt to the message format, the protocol, and the security constraints of each party. We are currently working on a low code solution to generate interoperability connectors using Model Driven Engineering techniques.
• Connectors for the IoT platform: The IoT platform developed by DRIT for CARL Software requires very frequent logical and programmatic upgrades. This is due to the context of sensors in the field, which changes regularly. To achieve this, we have deployed a Low Code solution called Node-Red, which enables a set of hardware peripherals, PLCs and online services to be wired graphically. Node-Red6 provides a browser-based editor that facilitates the rewiring of hot data streams without interruption and in seconds. A good example of fully operational interoperability and data processing using Low Code principles.

6.https://nodered.org/

• Automation of accounting processes: Recently we have been using LCNC solutions to configure the behavior of an RPA. This solution allows us to automate the tallying of accounting balances for our financial management products. We have experimented with the use of UIPath, an ABR solution that is based on the principles of LCNC. As shown in Figure 8, the robot is configured graphically using predefined components. If required, custom components can be added by writing code.

These examples and prototypes put three aspects into practice:

• The diversity of LCNC solutions that can be used for some of our business problems. Each prototype uses a different LCNC solution that meets a particular need.
• The important role of Model-Driven Engineering, a science that has long been exploited in critical industrial sectors such as aeronautics, space, and automotive [23], [24].
• The potential of these tools to rapidly redevelop/configure software solutions without the need for technical development or expertise.

Conclusions

Operational Context and Implications

Low Code/No-Code platforms have always existed and have “solved” to varying degrees the problems of industrialization in the face of the “complexity” of the moment. Operating systems have masked hardware problems, the cloud has solved the problem of hosting, site builders have solved the problem of web pages, the C compiler has solved the problem of generating assembler code, etc. In the end, the LCNC is still 10 years behind the needs of companies, i.e. it is still up and running but never catches up with the sometimes disproportionate expectations of the industry.

In the end, the LCNC still lags 10 years behind the needs of the business, i.e. it is still up and running but never catches up with the industry’s sometimes disproportionate expectations.

LCNC techniques can now progressively satisfy certain needs of software publishers who create applications. It is reasonable to believe that software publishing will, in turn, undergo massive “industrialization” by repositioning the focus on :

• The uses, which are becoming far too numerous,
• Optimization, certainly integrating environmental considerations,
• Responsiveness, to adapt solutions much more quickly to ongoing changes,
• The absorption of complexity, which is slowly slipping away from even the largest of team capacities.

It is therefore very important to understand the very strong complementarity that exists between several movements that are taking place simultaneously at the moment in the software development sector :

• The multi-tenancy. It plays a key role in the Cloudification of solutions since it significantly improves the so-called elasticity of infrastructures. More simply, the capacity to absorb transparently the increase in the number of users (volume of data and business transactions combined). It is fundamentally the data modeling and architectural organization that establishes this capacity in software solutions, but this translates into a profound and complex transformation of the methods of production and operation of the solutions.
• Micro-services. They play a central role in the dissemination, updating, and adaptability of solutions. Micro-services install native means in applications to facilitate CI/CD (Continuous Integration / Continuous Delivery) strategies. In simple terms, this makes it possible to deliver functionality, sub-functionality, data updates, or technological troubleshooting in an industrial, (semi-)automated and almost transparent manner to users. This is a radical transformation that changes the way we design our solutions and it is the design and functional transformation professions that are impacted.
• The Low Code/No Code. This movement literally shifts the value chain upstream and little by little places industrialized and packaged production capacities in the hands of marketing and business experts. This leads to a double movement. On the one hand, a horizontal movement implies an increase in the skills and flexibility needs of the product marketing professions and business expertise (including pre-sales or technical sales engineers for example). On the other hand, the dispossession of developers, whose profession will move vertically towards architecture, complex environments, and automation processes, require far fewer people but much more qualified personnel (experts).
• The IA-Code. The latter, which appeared 4 to 5 years ago, tackles the business and functional core of software applications. The names of its use are numerous, such as test automation, code analysis, reverse engineering, and in fact hiding domains that research has established more than 25 years ago. However, AI now provides the beginnings of solutions for problems that have remained very theoretical until today. Soon, “software factories” will make it possible to recover the business knowledge contained in software to produce tests in an automated way, to analyze production systematically (quality, security, protection of privacy), or to capitalize at a high level on the know-how that is integrated into the code. The most important thing is that the functional and business dimension will no longer be trapped by low-level technology (Java, PHP, HTML, etc.) and will be accessible in natural language in models that can evolve, be updated, and reused. Tomorrow’s question will be: to which platform should I generate my application? How can I express the evolution of my business?

The alignment of these four major evolutions over the next 10 years is likely to totally disrupt the software development business. The trend towards massive industrialization will make it increasingly possible to focus on the “what” (what to produce, what services to develop) rather than on the “how” (what technology, what hosting).

Dangers of Outrance Automation

Despite the simplicity of low code platforms (and this applies to RPA as well), maintainability is not guaranteed. Based on standard components, an integrated development environment allows applications and robots to be designed quickly, which can result in forgetting the need for thoughtful design. Thoughtful design is crucial to ensure good maintainability and to avoid cascading costs in the future.
Low code applications and RPA robots often depend on the environment in which they are run. All it takes is an (often unavoidable) tool update, the arrival of a new product, or an exceptional event to disrupt this environment. If you are unable to restart a falling robot or application, an entire department can be penalized.
On the other hand, the power of these platforms means that they must be used responsibly. This introduces here the notion of governance. RPA and Low Code require a framework to define development standards, evangelize good practices, verify and validate robots and applications before they go into production so that they do not create more problems than they solve.
Major accounts that automate a large number of robotic processes have already implemented this governance model. Indeed, RPA is demonstrating its full potential when it comes to scaling up. The benefit of robots taken collectively is greater than the sum of their individual values. A Talan customer was thus able, by implementing a governance model, to produce some 200 robots in two years with a maintenance cost that is still negligible today compared to the gain brought by these robots.
This governance must also avoid excessive automation, which would destroy a large number of jobs, particularly in back-office functions. As in the field of artificial intelligence, it is indeed a question of “increasing” the number of employees to free them from tedious tasks and give them increasingly powerful tools to maximize productivity.
Between low code, no code, and RPA, the marriage of reason seems inevitable. However, experience has shown that strategic governance is the key to reinventing the business and giving full scope for team creativity.

Bibliography

1] ” Convertigo-White-paper-FR.pdf “. consulted on January 24, 2021. [En ligne]. available on: https://www.convertigo.com/wp-content/uploads/2021/01/Convertigo-White-paper-FR.pdf.
[2] A. E. Bell, ” Death by UML Fever : L’autodiagnostic et le traitement précoce sont essentiels dans la lutte contre la fièvre UML. “, Queue, vol. 2, no 1, p. 72-80, March 2004, doi : 10.1145/984458.984495.
3] A. Peters et A. Peters, ” How A Man With No Coding Experience Built An App That’s Bringing Solar Power To Yemen “, Fast Company, May 08, 2017. https://www.fastcompany.com/40417060/how-a-man-with-no-coding-experience-built-an-app-thats-bringing-solar-power-to-yemen (consulted on January 26, 2021).
[4] ” BeaverBuilder: WordPress Page Builder Plugin “, Beaver Builder. https://www.wpbeaverbuilder.com/ (consulted on January 20, 2021).
[5] ” Elementor: WordPress Website Builder “, Elementor. https://elementor.com/ (consulted on January 20, 2021).
6] ” Flow XO – Easy to use chatbot platform “. https://flowxo.com/ (consulted on January 20, 2021).
[7] ” Intercom: Chatbots customizable “. https://www.intercom.com/drlp/customizable-bots-biz utm_source=google&utm_medium=sem&utm_campaign=9363026333&utm_term=%2Bchat%20%2Bbot&utm_ad_collection=95042838956&_bt=418883716507&_bg=95042838956&utm_ad=418883716507&offer=drcustombotsbiz&utm_campaign_name=go_evg_acq_trial_nb_chatbot_bld_emea_en&utm_ad_collection_name=sup-p_chatbot_bmm&utm_ad_name=drcustombotsbiz_text_21q1&gclid=CjwKCAiAo5qABhBdEiwAOtGmbmweUU2MY9NUyPC7FQahF3RVNAmRbnDYVhnwe9_GZ6KZn0x6wZQKLxoC_xQQAvD_BwE (consulted on January 20, 2021).
[8] ” Botnation “, Botnation AI. https://botnation.ai/fr/?gclid=CjwKCAiAo5qABhBdEiwAOtGmbq0CqlAuYyECcRrUjDBfdRSjdVNIDpAYmqFdyQQ8ScHmPlmy3owvJRoCG-kQAvD_BwE (consulted on January 20, 2021).
[9] ” Import.io – Data Extraction, Web Data, Web Harvesting, Data Preparation, Data Integration “, Import.io. https://www.import.io/ (consulted on January 20, 2021).
10] ” Octoparse : Web Scraping Tool & Free Web Crawlers “. https://www.octoparse.com/ (consulté le 20 janvier 2021).
[11] ” eBook: The Field Guide to No-Code Application Platforms “. https://www.unqork.com/resources/ebooks/ebook-the-field-guide-to-no-code-application-platforms (consulted on January 24, 2021).
12] ” 2019 Annual SaaS Trends Report – Researched By Blissfully “, Blissfully. https://www.blissfully.com/saas-trends/2019-annual/ consulted on January 26, 2021).
13] ” The Low-Code Manifesto & 9 Principles of Low-Code App Development “, Mendix. https://www.mendix.com/low-code-guide/the-low-code-manifesto/ (consulted on January 20, 2021).
[14] T. L. Calvar, ” Exploration d’ensembles de modèles “, p. 157.
[15] B. Combemale, ” Ingénierie Dirigée par les Modèles (IDM) – État de l’art “, 2008.
[16] ” Maximiser l’efficacité : How TIME works smarter, not harder “. https://blog.airtable.com/maximizing-efficiency-time-video-production/ (consulté le 26 janvier 2021).
[17] G. Fischer, ” End-User Development and Meta-design: Foundations for Cultures of Participation “, End-User Development, Berlin, Heidelberg, 2009, p. 3-14, DOI: 10.1007/978-3-642-00427-8_1.
18] G. Fischer, E. Giaccardi, Y. Ye, A. G. Sutcliffe, et N. Mehandjiev, ” Meta-design : un manifeste pour le développement des utilisateurs finaux “, Commun. ACM, vol. 47, no 9, p. 33-37, sept. 2004, doi : 10.1145/1015864.1015884.
19] A. I. Mørch, G. Stevens, M. Won, M. Klann, Y. Dittrich, et V. Wulf, “Component-based technologies for end-user development”, Commun. ACM, vol. 47, no 9, p. 59-62, sept. 2004, DOI: 10.1145/1015864.1015890.
20] T. Hamelin, ” AUTOMATISATION DES PROCESSUS MÉTIER : ÉTAT DES LIEUX ET APPORTS DU RPA “, p. 14.
[21] ” FlowBuilder “, Salesforce Lightning Flow. https://trailhead.salesforce.com/fr/content/learn/modules/business_process_automation (consulted on January 20, 2021).
22] Gartner, ” Magic Quadrant for Enterprise Low-Code Application Platforms “, August 2019.