The problem we are facing
The outbreak of COVID-19 in Europe and necessary national measures taken to tackle the spread of the virus may cause significant disruption to the provision of education, training and mobility opportunities for learners, teachers and educators across the European Union (EU). Most governments around the world have temporarily closed educational institutions in an attempt to contain the spread of the COVID-19 pandemic. These nationwide closures are impacting over 72% of the world’s student population. Several other countries have implemented localized closures impacting millions of additional learners. (source:Unesco) Covid-19 pandemic spread is affecting schools and universities all around Europe. All Italian schools and universities have been locked down since the 5th of March with the aim to limit the spread of the virus. Subsequently, almost all European educational systems decided to follow Italy and closed their schools by 16th March 2020. The last country to announce such measures was the United Kingdom, where all the schools closed by March 20th-23th. Since then, lessons, exams and degrees dissertations have been postponed or held remotely. Lessons were moved from classrooms to e-learning platforms, while remote exams still remain a big question point due to the uncertain effectiveness of solutions given, in terms of process and technology performance. Most relevant effects on the process are:
- Extra time spent on additional required tasks.
- Troubles on planning and attending of the assessment activities.
- Lack of options in term of examination modalities.
- Lack of quick interactions between the professor and the student to resolve doubts during exam execution.
- Slow capability in reacting to unforeseen events and changes of schedule.
Effectiveness on technology performance are:
- Internet connection instability during the exam session.
- Unfamiliarity with the platform.
- Difficulty in finding a unique platform that meets all needs.
- Migration to external platforms with consequent loss of some academic informations (students' personal data, courses).
- Use of un-tested services.
- Digitization on handwritten work.
Moreover, the remote examination platforms do not respond effectively to a problem that has a strong weight in the success of an exam: the control over students’ cheating behaviours by the student during exam’s execution. Remote examinations are related to an increasing of students’ cheating behaviour cases due to ineffectiveness of control system. From this comes the strong need to deter, detect and prevent cheating behaviors. But today, it is not possible to simultaneously control environment, interactions and use of forbidden tools without a complex mix of several proctoring technologies:
- Face recognition and Voice recognition
- Plagiarism and authorship validation
aiLearning is an AI-based solution developed with the aim to simplify remote examination
The project is meant to ease both students’ and teachers’ lives. aiLearning is a software system created to overcome the difficulties brought up by remote learning process. It responds to the necessities of new tools to help students and teachers keeping contact during Covid-19 and conduct examinations effectively in a safe environment. The focus is based on the exams and, more specifically, on enabling universities to test their students remotely providing an all-in-one platform where students can be tested on a subject both in a written or oral session. aiLearning supports oral exams, multiple-choices exams, written exams with open answers and written exams carried out offline on paper. The answer will be captured by the system and immediately sent to the teacher’s profile, so that the professor can check the evaluation, modify it or confirm it. The teacher creates the exam’s event on the platform, followed by all the information such as the list of the enrolled student, the exam questions and the correct answers. The student answers by sending a vocal message, writing the answer or checking out the right one if it is a multiple choices-exam. Chances to cheat on the test are knocked out by adopting several features in order to drastically limit the chance of putting in place cheating behaviors. An AI algorithm also supports the teacher during the test’s validation, rating the answer based on its relevance, the language used and the student’s ability to synthesis.
The solution impact to the crisis
Since social distancing can’t ruin the magic of a conversation, aiLearning provides the possibility to take an exam by speaking with a bot. Each teacher can share simultaneously all questions with all students using a single interface and without needing any witness. All students can interact with the bot providing to it the best answer for each exam question. Checks on student voices can certify who is the real speaker, unmasking cheating students. Since remote education requires more time with respect to the located learning, aiLearning is aimed to free teachers time supporting them with the activities related to exams correction. This time can be better involved to support students with their learning path.
The value of the solution after the crisis
The time consuming aspect of oral exams is not a problem related only to Covid emergency. In many universities or schools, there is often a short time-boxed window for exams and so many students to examine. aiLearning can be used as solution of this issue also after the crisis providing a valid alternative for simultaneous and time-saving exams thanks to the innovation of the solutions. Once the emergency will be overcome, with a hypothetical reopening (partial or total) of the schools, aiLearning can play a role in carrying out the examinations on site or remotely. Thanks to the following benefits:
- Process innovation;
- Allowing simultaneous sessions;
- flexibility on location and timing of the session;
- time saving.
Moreover, aiLearning can continue to be a good learning companion also after our home doors will be opened.
aiLearning: AI for remote examination
aiLearning platform has all functionalities that are needed to create, schedule, execute and evaluate an exam. These functionalities answer to all needs coming from both figures involved within an exam: teacher and student. aiLearning provides user friendly interfaces, one for the teacher and one for the student, designed and tailored to their needs, that the team collected during this period with a set of surveys addressed to several Italian and European universities. The cheating behaviour is limited by a vocal and face recognition system that checks the student identity at the beginning of the test and periodically, during the whole duration, by a built-in solution that avoids some functionalities like changing browser tabs/windows during the exam and copying/pasting contents. Moreover, providing questions one by one, limiting the time to answer each of them, the oral exam modality aim at guarantee the correct exam execution. The teacher is also allowed to access a live streaming monitor in real time, in order to detect any suspicious behavior. Finally, a more complex proctoring solution, based on the use of a double cam monitoring system, with face tracking and object recognition, will be rolled out as a second step of the platform evolution.
aiLearning provides to the teacher the possibility to significantly reduce the preparation and correction time. It allows the teacher to dedicate more time to teaching and research. The truthfulness is preserved thanks to Machine Learning algorithms, providing speaker and face recognition capabilities. How does aiLearning help the teacher? Providing a dedicated interface with four main functionalities:
- Exam creation
- Exam launch, with identification checks
- Exam execution
- Exam evaluation.
aiLerning solution provides a teacher interface with a page to create the exam. The teacher can upload:
- Exam information: name, code, exam type, date and time
- List of enrolled students
- Exam questions
- Correct answers: this functionality is optional. The teacher can choose it if he wants to activate the AI-based support for exam evaluation
The teacher interface provides easy functionalities for uploading files with all information needed to create the exam. After the uploading step, aiLearning automatically sends an email to the student with all exam information and with the link for the exam access.
Before starting the exam, aiLearning solution foresees two types of identity checks, based on advanced AI algorithms:
- Speaker recognition: at the beginning of the exam the student has to say some basic sentences in order to verify his identity. Thanks to a Deep Learning-based approach, aiLearning solution provides a level of similarity between the voice of the student and the one provided at the registration to aiLearning platform. This activity is performed simultaneously for all students, reducing drastically the identity checks time. Through aiLearning platform, the teacher can visualize in real time the similarity raised from the algorithm and can, eventually, proceed with manual checks.
- Face recognition: using the camera, a Machine Learning algorithm, based on Neural Network, compares the student face with the picture provided at the registration step. The check is done simultaneously for all students: results are shared in real time with the teacher, who can proceed with further checks. Once all checks are easily performed, the teacher can start the exam.
During the exam execution the webcam is always switched on, for student monitoring. Video checks are not enough: aiLearning solution blocks some functionalities, such as changing browser window/tab and copy-paste content option. These functionalities ensure that students cannot adopt cheating activities. If something goes wrong, aiLearning notify the teacher, who can interact with the students, he can manually verify the situation and eventually close the exam. The teacher has the possibility to visualize in real time two types of dashboard:
- Exam focus: with all information related to the status and progress of the overall exam
- Student focus: with the answer of each student to the provided exam questions.
During the exam execution, all answers submitted from the students (that cannot be changed) are available to the teacher in real time. This functionality is available for three types of exams: oral, written with open answers, written with multiple choices answer option.
At the deadline of the time window, the teacher can proceed with the exam closing. All final answers will be directly and immediately available within the dedicated dashboard. If the AI functionality has been activated, aiLearning provides a fist evaluation of the single student answer by applying Machine Learning algorithms. The evaluation is available for three types of exam: oral, written with open answers, written with multiple choices answer option.
- If the exam has multiple answers option, at the uploading phase, the teacher must indicate the right one/ones. Starting from this information, for each submitted answer, aiLearning will check if the student answered correctly. At the end, according to the number of right answer and based on the weight that the teacher could provide to each question, aiLearning provides a suggested mark.
- For written exams with open answer options, aiLearning leverages advanced solution of Natural Language Processing and Understanding in order to analyze the text input, to remove noise and to get keywords and key-concepts expressed by the student. Each student's answer is compared, through Machine Learning algorithm, with the expected answer provided by the teacher: the more similar they are, the higher the mark suggested by the algorithm will be.
- For oral exam, the evaluation works as for written exams, with an additional step: the recorded student voice is transformed into text by applying speech-to-text algorithms. If further manual corrections on printed paper are needed, the teacher can export and print all students' submissions. The teacher can review all students' answers and decide to change the mark suggested by the AI module. The definition of the final mark can be done using the dedicated interface. The complete list of students and related marks can be finally exported.
aiLearning provides to students all functionalities for preparing and taking exams without stress. aiLearning is also a learning companion for the exam preparation. Indeed, thanks to Artificial Intelligence algorithms, aiLearning can provide feedbacks on the preparation level for each exam, reliving the feeling of a conversation with a classmate, emotion that the social distancing has taken away. Moreover, the student can enroll the exam and easily check his/her identity. There is only “one problem”: cheating activities are forbidden! How aiLearning does help the student? Providing four main functionalities:
- aiLearning platform registration
- Onboarding process
- Exam execution
- Exam closing
aiLerning foresees a webapp interface dedicated to students. For using it, the first step is to register to the platform providing:
- Personal data (for example, name, surname, email address, student code).
- Recent picture
At the end of the registration phase, the request is put “on-hold” in order to be verified from a human operator (for example, a university assistant). Registration is easy, quick and it represents a one-time step, it must not be repeated before each exam.
The scope of the onboarding phase is to verify that the student provided real information about himself, since these information are used to verify his/her identity at each exam. This step involves also a human operator who, during a video call, has to:
- Check that the picture provided at registration phase corresponds to the student (this information is needed for face recognition before starting the exam)
- Get a voice recording of the student, allowing speaker recognition check before starting the exam
- Check the identity card for verifying that the person is the student enrolled at the university.
After these checks the assistant can close with success the request of registration. Onboarding is a one-time step. aiLearning team will define a knowledge transfer process for universities in order to support them for this onboarding step.
The exam starts automatically at the scheduled time. The student can access the dedicated webapp in order to see exam questions. Each question will be provided one per time for reducing the possibility of cheating activities: if the student submits a question, he cannot modify it anymore. The order of question is defined by the teacher. Depending of the exam type, the student can provide the answers by:
- Sending a vocal message (for oral exam). In this case the student can autonomously record the audio and the he/she can upload it within the interface. The audio will be processed with speech-to-text algorithms in order to be transformed into text, verbalized, stored within the aiLearning platform and made available for the teacher.
- Typing the answer content (written and open answer exam). The textual answer is processed and made available in real time within the teacher interface.
- Selecting the right option (for multiple-choices exam).
- Uploading a picture of a paper sheet (for written exams done offline, i.e. scientific topics). This functionality is available within the aiLearning mobile app. In order to avoid untrusted behaviour, the student must upload the pictures during the dedicated exam slot. Uploading after the exam deadline are not admitted. During the exam, each student has the possibility to have a live chat with the teacher for asking questions and clarifying doubts.
The exam can be finished for three different reasons:
- Automatic submission: the exam is closed (scheduled time reached or closing from teacher).
- The student can submit the final exam before the deadline with all provided answers.
- The student can retire, without submitting any answer. In this case, no information are provided within the teacher interface.
Microservices and Docker
aiLearning is a composite of several interfaces provided that are used at different moments of the user journey and by different type of users (student, teacher, university accountant) The architecture based on microservices allows us to parallelize and speed up the development phases, to keep the system highly maintainable and testable, loosely coupled and organized around business capabilities. The use of microservices helps us to easily manage several programming languages and frameworks. This simplifies the resources onboarding phase and the adoption of new technologies over time. Containerization, through Docker, provides individual microservices with their own isolated workload environments, making them independently deployable and scalable. Finally. traffic management, authentication and access control, throttling, monitoring and API version control will be handled by an API Gateway: a”front-door” for the business logic.
Serverless paradigm using Lambda
The use of the application is strongly related with the exam period that is usually concentrated in specific time frames, so the system won’t be stressed evenly along the year. The concept of serveless, coming with Lambda computation, helps us to optimize the cost related to the infrastructure during the quiet period and to scale up easily and fastly during the exam sessions. The idea is to deploy all our AI modules as lambda functions that are activated on-the-fly.
All aiLearning interfaces are provided as web applications, easily accessible through a common browser. After the hackathon, the team has been focused on the design and implementation of the teacher interface. Some screenshot of the interface, fully developed in Vue.js, are available here
Where we are
aiLearning idea is born on 24th April during a virtual meeting among 7 friends currently located in two different countries (Italy and France). The AI algorithm, developed with Python, currently manages the following steps:
- Speech to text conversion in order to turn the voice command input into textual data. This step is performed by using SpeechRecognition Python package: a library for performing speech recognition, with support for several engines and APIs, online and offline.
- Natural Language Processing (NLP) and Understanding (NLU) steps have been developed for text mining solutions in order to analyze students' answers. For this important step, aiLearning leverages the main functionalities available within nltk Python package. Using classification, tokenization, stemming, tagging, parsing, lemming and semantic reasoning, the textual input is processed with the aim to remove noise (for example, stopwords) ang get important information.
- Machine Learning algorithm defines the level of correctness of each answer thanks to similarity models based on different weighting factors used for information retrieval solution. Within the current solution, the algorithm has been developed by leveraging the most important functionalities of Machine Learning Python library: sklearn (for example, defining the tf-idf matrix).
The current aiLearning solution simulates an Art exam, just few examples:
aiLearning team also develop a user-friendly interface for the teacher with all functionalities described before.
Here below we have made an analysis focusing on the worldwide major players.
Meetl, based in India, was founded in 2009 by Ketan Kapoor, Tonmoy Shingal. The company has raised a total of $4.4M in funding over 3 rounds. Their latest funding was raised on Aug 13, 2012 from a Series A (source: crunchbase). In 2018 Meetl did the exit: it was bought by Marcer with a deal value of 42.7 MIL $ (source: inc42). The major market of Meetl is concentrated in Asia and USA, not in Europe), with 4000+ clients, mostly offering HR recruitment tools to companies (B2B). Meetl is a software-as-a-service (SaaS) platform which uses data science for talent assessment. The major functionalities are: candidate authentication (OTP, Email and ID proof), ai-based proctoring (candidate authentication, safe exam browser, student behavior detection), exams preparation formats (different types of questions and formats), exams execution formats (typing, audio recording, video recording and math equations). The ai-algorithm is applied only for the proctoring functionality. Even though it is a complete tool kit also for universities, there are still some important functionalities missing like: NLP-speech to text for oral exams, exam execution without ai as well as no real-time evaluation. Furthermore, a front cam controls cheating behavior, however a student can place a device/paper just at the same level of the screen overcoming the front cam checks.
Talview, based in California, was founded in 2012 by Sanjoe Jose, Subramanian Kailasam, Tom Jose. The company has raised a total of $6.8M in funding over 6 rounds. Their latest funding was raised on Aug 14, 2019 from a Series A round (source: crunchbase). It is based mainly in the USA, Canada, Asia and Australia with 1000+ clients. Talview is an Instahiring platform especially addressed to companies supporting the Talent Acquisition field. Proview is defined as an industry specific solution of Talview, focused on the educational system. Proview offers a proctored online exams experience for candidates. The major functionalities offered are related to proctoring activities to secure the examination environment. In details: face and vocal recognition, speech to text (NLP), Proctoring, store video recorded, browser policing (navigation check & disable copy-paste), real time communication via chat box and results of the exams. Talview is mainly focused on HR solutions. Its subscriptions model is available for you to purchase in three models: Per Transaction, Per Recruiter User or Per Employee (Enterprise Account). For the education system, recently has been introduced a safe exam kit for universities to face the Covid emergency. It provides an advanced proctoring kit, but does not resolve the exam execution and evaluation.
Disamina was founded in 2012 and it is based in India. Relevant activities started to be traced in 2020. No relevant information is available for the market and the number of clients. Disamina is an online assessment software which can be used by educational institutes (school, college, university, training center) and organization to conduct an online examination. The main functionalities are: No added software installation, variety of questions patterns, video recording, detailed analyses of results. Disamina model is a customized product subscription plan.
We focused our analysis initially on the education system (primary, secondary and tertiary). The primary source that we have used to find the information for the European, North American and Asian education systems is UNESCO. Our analysis begins with the European educational system.
The total number of students (source: Unesco):
- Primary schools 23,8 MLN
- Lower Secondary schools 31,8 MLN
- Upper Secondary schools 26,2 MLN
- Post-Secondary (non-tertiary) 1,7 MLN
- Tertiary 28,9 MLN
The estimation take in consideration also the number of professors, which, following the above category selection is:
- Primary schools 2,8 MLN
- Lower Secondary schools 3,14 MLN
- Upper Secondary schools 2,29 MLN
- Tertiary 2,38 MLN
Primary – Secondary schools
Following a pricing evaluation weighted for the primary/secondary schools providers’ benchmark, we have defined a year unitary cost per student of 3 €. Spaggiari Group, major schools’ provider in the Italian market, confirmed this evaluation. Assuming that an elementary school has 300 students the total cost per year for this institution is around 900 €. Price of aiLearing for primary/secondary schools: 3 € per student / year
To price the service for the universities we have focused our attention on the work made by each professor. The average salary of a university professor is around 35.500 € /year (source: European University institute) with an average of 1600 working hours, which means 22 € / hour. Each course has an average of 100 students and each professor follows approximately 1 to 2 subjects, 1-2 exams are done for each subject. Moreover, some students might need more sessions to pass an exam, therefore we assume that professors might need to prepare, execute and evaluate approximately (100*2*2*2). Following this logic each professor needs to manage 800 tests per year. Considering a default estimation, we assume that a professor using an aiLearning platform to manage the exams can save 20 minutes per test. Therefore, each year each professor can save up to 16.000 minutes, which can be re-invested in more profitable activities for the university such as research. According to this enormous benefit brought by this technology, we identify as fair a cost of 49 € /year per professor. Therefore each university should pay only 49 € per professor to cover 800 tests and save up to 16.000 minutes. Price of aiLearing for tertiary schools: 49 € per professor / year
Once our pricing is defined, we can detail the market size. As assumption, please consider that we are just taking in consideration the implementation of our technology in the European educational systems. For this first analysis, we are excluding its applications in other sectors (private education, HR recruitment process, journalism, etc…) and in other continents (as North America, Asia, etc…). Now that we have a fair evaluation of the pricing, we can also detail the European market size. As previously mentioned, we assume a fair price of 3 € per student per year. If we sum the number of students for primary, secondary and post-secondary schools, we arrive at a total number of European students: 83, 7 MLN. The European market size for primary, secondary and post-secondary is: 251 MLN € (83 MLN*3). Assuming that the pricing for the universities is 49 € per professor/year and that the total number of universities’ professors is 2,38 MLN, the European market size for the universities is: 117 MLN (2,38MLN*49). The European market value just considering the educational system is overall: 368 MLN €.
Two main customers streams have been identified: Universities and primary/secondary schools. For universities, we propose a B2B yearly subscription plan per professor. For primary-secondary schools, we can exploit two alternatives:
- Full aiLearning package as B2G yearly subscription plan per student
- API integration within existing platforms as B2B fee based on usage
You can find the full analytical version here
Our plan is to go live with aiLearning MVP after 2 months of development and V1 after 4 months of development After 2 months, we have already planned to test the MVP with some universities. The goal is to prevent production issues as well as to obtain the users feedback before the real usage. It is a tight deadline, nevertheless we commit to delivering the Version 1 for August 2020 in order to support September exams sessions for Italian and European universities. Starting from a basic need we have split the development costs in relation to the three major requirements for our product:
- Learning companion app
- Advanced AI
We have estimated the man-days essential to develop the main functionalities, based on a prioritization, that the MVP should have to be used by the educational system. The following table highlight each functionality per man-day needed with a differentiation of BE/FE activity:
The total development man-days needed to develop the platform are 182. On this, we should apply a contingency rate, which is equal to 254. We have selected a range of experts that are needed to develop these functionalities in 4 months (see table below). Each figure is valued daily, based on an Italian average salary range, in relation to the years of seniority requested: Considering on average 36 working days per figure (with contingency) on the last column we can find the total cost estimated for each role.
The total development costs for the implementation of the platform are therefore: 58 097 €.
Learning companion app
For the development of the learning companion functionality (AI assistant helping students in the exam preparation) we have estimated to involve just the BE expert and the FE expert with the following engagement:
Considering a total cost for the development for the Learning companion app of: 7 506 €
The R&D activity will be crucial for the development of the artificial intelligence algorithm. For this reason the majority of the functionalities previously listed as well as the basic AI functionalities, will be released to the pilot universities starting from the 2nd month in order to enrich also the AI algorithm with the inputs of professors and students. Doing so we will be able to reach its readiness in the 4th month of development. Following this path, we assume to request a full time engagement of 2 senior data scientists (4/5 years of seniority):
The total costs for the development of the AI algorithm are 35 538 €.
V1 Production costs
The total costs for the creation and implementation of the V1 are 116 800 € The internal team in charge of the development of the platform, the learning companion app and the AI algorithms will be composed of 11 figures: 1 fullstack, 2 devOps, 1 BE expert, 1 BE junior, 1 FE expert, 2 Data scientist, 1UI/UX designer, 1 Project manager/Product owner and 1 architect.
In parallel to technical development, two other departments are fundamental for the growth of aiLearning’ business: Legal and Sales. For the two we have forecast 4 figures that will manage: prospect, partnerships, customer relationship and new market development. A buffer of 1 Fullstack figure for the university support is required. Therefore an additional 64 000 € payroll wages is added.
For the firsts months our goal will be to create a strategic (free) partnership with two universities to test our MVP and enrich the ai algorithms, before the official “go-live” in August. Doing so we need to have a server structure able to cover twice 20 000 students (avg number of students per university). We have estimated a server cost of 0.30 € per year per student. The server costs coverage are (40 000*0.3) 12 000 €
Equipment cost of 100 €/month per person 6 400 €
On the total costs logic calculated for the V1 we have applied also a recovery cost % for extraordinary expenses. The total costs requested to develop and distribute the minimum viable product on the market is 280 000 €
In the regional phase, we must establish our brand awareness and spread our product. We have identified that we will be able to work with 15 universities during this period. Moreover, we forecast to open the primary/secondary school market with 50 units.
Based on the price of 49 €/professor per year and considering that 15 universities have on average 24752 teachers, the revenues estimated for the regional phase are 758 500 €. From primary/secondary schools, giving a price of 3 €/students and 500 students per school, the revenue estimated for the regional phase of 150 000 €
The total estimated revenues for this phase are 908 500 €.
The server costs will be around 105 000 € following the same logic as above.
Following the logic for the V1 increasing the number of people we will need an office. We will adopt the smart working approach so we need an office with a capacity of the 70% of the team dimension. A total cost for services and third party assets is estimated to be 338 600 € With the recovery costs percentage we will have a total production cost of 388 600 €.
In order to serve these universities we will need to increase the internal team. Considering an already existing team of 10 engineers, a part of those will focus on new functionalities, bug fixing, maintenance and product improvement. We will need to add 5 more figures to support universities over the technical design, integration and delivery of the service (for both professors and students).
For the commercial, marketing and legal departments we forecast a need of 6 more figures to support the market expansions in all the European countries. The total staff costs will be 556 000 €.
SCALE UP TO EUROPE
In the scale up phase, we need to speed up the usage across European countries of aiLearning platform. We did a pessimistic evaluation considering that we will be able to acquire just 25 new universities and 100 new schools compare to the regional phase. Nevertheless, our goal is to establish an aggressive execution thanks also to the European community support.
Based on the price of 49 €/professor per year for universities and 3 € per student for primary/secondary schools, total revenues estimated for the scale up phase are 3 534 000 € The server costs will be around 270 000 € following the same logic as above. A total cost for services and third party assets is estimated to be 995 000 € including also a contingency cost.
In the second year, the team will be increased significantly to a total staff number of 35. The idea is to invest part of the net earnings acquired the 1 year to increase the growth path of the company as well as provide a valid support to all the universities. In this phase, we will take in consideration also the opportunity to implement new aiLearning features for the company segment (HR recruitment activities).
The total staff cost will be equal to 1 316 000 €.
The EBIT is going to be 1 223 000 € and the NET earnings after taxes 942 000 €.
To reach the break-even already in the first year we need to have at least 23 universities.
The company has been evaluated using the DCF method assuming a constant growth in the cash flow dynamics. Since we have no historical data which could help us for an effective estimation of medium and long term flows, in order to compute the Terminal Value of the company we preferred to consider only the results obtained in the Business Plan for the first and second year. Starting from the Business Plan and the estimated results for the second year (Scale Up), we used the resulting Cash Flow in order to input the following formula and evaluate the company:
- TV = Terminal Value
- FCFF = Free Financial Cash Flow
- g = Growth Rate
- wacc = weighted average cost of capital
- T= 1yr Scale Up
Assuming an unlevered point of view - today it is not possible to estimate which will be the debt/equity ratio of the company - the WACC will be equal to the Ke (equity return). For the estimation of the Ke, we started studying the sector to define a target Ke higher than the average. It has been done to guarantee a premium to the venture capital investor. This value has been defined at 25%.
has been calculated starting from the EBIT estimated during the first year of Scale Up lowered by 35% considering contingency problems which could affect results. Finally the resulting value has been considered net of taxes (in Italy, taxes are around 23%). Please find below the computation:
About g parameter - average Cash Flows growth rate – the computation started planning the market share target we would like to achieve inside the two main markets identified during the pricing model definition: "primary/secondary/post-secondary education" and "tertiary education"; In particular, considering the following 5 years, we estimate to reach a market share of 0.27% for the "primary/secondary/post-secondary education" and 2.92% for the "tertiary education". Based on this estimation and starting from the forecasted results for the second year (1yr Scale Up) with the same ROS, we will reach an average cash flow growth rate of 5.01% per year. On the basis of these data, today the value of the company is:
The pitch desk is available here
The necessities in order to continue the project
The aforementioned analysis led us to identify the following needs:
We identified professional figures needed to roll out the development during a period of 2 months for a first MVP and a total of 4 months for a V1 containing an advanced AI module. In detail, we would need 11 technical experts:
- 1 devops
- 1 backend expert
- 1 backend junior
- 1 frontend expert
- 1 UX/UI designer
- 1 NLP expert
For advanced proctoring solutions we are open to discuss partnerships.
The total costs requested to develop and distribute the minimum viable product on the market is 282 000 €
Relations with the **educational institutions as Schools and Universities with the aim of creating the network necessary for the testing phase and subsequent adoption of our product. The institutions will be involved during an onboarding phase from June to September focused on 2 product deliverables:
- MVP: by the end of June. 2 universities involved.
- Version1: by the end of August. 14 universities involved.
A key point of our solution is the research on AI and machine learning that can be done in collaboration with universities through outsourced research thesis. aiLearning scalability, flexibility and low-fixed costs allowed us to open discussion also with non-profit institutions in order to boost the spread of education worldwide.
Since our product targets institutions, the government could support and promote our relationship with them. Moreover it can help us in identifying and addressing all critical aspects related to digital identification.
Born in such a critical situation that strongly impacted the whole world, aiLearning aims to become the solution to disrupt the way education is provided, making it accessible to anyone, anywhere.