There is a growing global awareness that harnessing the potential of information and communication technologies (ICTs) can foster innovation, progress, and economic development. Through this process, a transformation takes place that aims at country growth, employment, improved life quality as well as simplification, and greater active democratic citizen participation. In this rapidly evolving scenario, cloud services for Public Administration seem to be one of the most cost-efective means to overcome system limitations, providing those features of efectiveness, eficiency, transparency, participation, sharing, cooperation, interoperability, and security needed for today's challenges. In this research, we present an approach based on distributed systems and ontologies for creating virtual assistants that leverage artificial intelligence and machine learning to save time and public money by providing better services, increasing efectiveness, eficiency and transparency. We developed a framework to guide cloud consumers in selecting cloud technologies through opportunity and risk analysis to reach this objective. The results demonstrated reduced support costs, improved tracking of provided services, and simplified work for public oficials with the goal of re-balancing bargaining power between the small organization and the cloud service provider.
relevant in our lives in the future.
With this in mind, the giants of technology have
develToday, technologies related to cloud systems and artifi- oped platforms capable of interpreting expressions in
natcial intelligence (A.I.) are areas of considerable interest ural language and ofer virtual assistants such as Google
and subject to huge investment and development. With Assistant; Amazon Alexa; Siri; Cortana, and IBM Watson.
the progressive development of new ways to produce Such platforms are capable of performing human
diaknowledge, these technologies define new approaches logue, answering questions on topics of diferent nature,
to business activities [
However, despite this growth of interest, the level of little support in decision-making and in monitoring the maturity of these solutions, in many cases, has stopped whole procedure. at the simple Proof of Concept, while in other cases there The “Sàvio” model overcomes all these limitations, as were only theoretical developments never put into oper- the proposed solution guides the buyer to choose the ation. This research aims to analyze cloud-related issues best solution for its needs by analyzing risks and opfor public government, developing a prototype model portunities, through a process and control model that that supports the public government in the wise approach is adequate and adapted to the context of the specific to the cloud services market. Hence, the name "Sàvio" public government. Sàvio Agent is based on IBM Wat(from the Latin sapius "to be wise"), a cloud supply man- son, the same technology used in the City of Markham – agement model that integrates not only best practices Canada’s high-tech capital – where are using the artificial and proven strategies but also artificial intelligence, to intelligence-driven virtual agent "IBM Watson Assistant obtain a system adapted to the reality of public admin- for Citizens" to ofer 24-hour customer service for resiistration. Sàvio proposes a scheme for governability of dents looking for COVID-19 information. The solution cloud services for the public government through: use machine learning to deliver reliable, consistent, and accurate information via online text chat and voice calls • the development of a back-end web app that takes – anywhere, anytime. This article discusses a combined into account specific contractual indicators and and innovative approach to the use of cloud and A.I. syssupports the public government in their detailed tems for the public government. Moreover, the model detection and tracking; presented is easily expandable and applicable to diferent • the development of a virtual assistant/chatbot to domains, both public and private.
intelligently govern the conversation scenarios The document is structured as follows: in section II, envisaged as part of a cloud provider for the pub- we explain the state of the art and the reasons why we lic government, and also ofer a targeted help chose the technologies used in the project; in section system; III, we introduce the Sàvio project, describing the model, • the tracking of Service Level Agreements (SLA). the case study, and the application. At last, in section IV we summarize our work, draw conclusions, and outline The Sàvio model allows: possible future works.
• reduce support costs; • trace the details of cloud services/products supply simplifying the work of public government oficials; • re-balance the bargaining power between small organizations and cloud service/solution providers by setting up an automated reputational center that allows any consumer, through cooperation, to report issues concerning their cloud service provision and assess the risk parameters resulting from the adoption of a specific cloud solution.
on the development of systems or products for interpreting natural language expressions based on supervised learning techniques, without further benefits. In addition, the systems proposed to the public administration are limited almost exclusively to ofering Software as Service (Saas) solutions-oriented mainly to electronic procurement (eProcurement) and therefore limited to the simple management of the assignment of works, services, and supplies (these are simple management systems tailored to customer needs). Such systems seldom integrate virtual assistants/chatbots and in the few cases where
transformation. As defined by the National Institute of Standards and Technology (NIST), cloud computing is a pool of computing resources such as servers, storage, networks, applications, and services (NIST, 2012). These resources are available on-demand with little or no interaction with the cloud service provider. Cloud computing is growing rapidly as it can be used in any industry without many obstacles [23], some of which involve contractual implications.
According to Gartner [24], the utilization of cloud computing is growing and governments have started to capitalize on the cloud. Worldwide end-user spending on public cloud services is forecast to grow 18.4% in 2021 to total $304.9 billion, up from $257.5 billion in 2020. The large adoption of Cloud technologies has demonstrated the efectiveness of this new paradigm to simplify data centers management. The Public Administration will use multi-tenant services, shared and managed by diferent organizations. Cloud computing for the public government is enabled through Government Application Stores (e.g. Fedramp marketplace in the U.S.A., AgID in Italy; paper (Sàvio BO). This survey led to the identification of Gov.uk in the U.K.) where services can be purchased, SaaS solutions mainly oriented towards e-Procurement, used, reviewed, and reused across the public sector. The understood as a service composed of integrated functions objective of the Government Application Store is to: pro- to support the public government in the computerized vide an open and transparent market in terms of costs for and telematic management of the entire process that the public sector; allow the public government to have goes from the collection of requirements, programming, information on the accreditation status of the service and awarding, and testing of a service/work. the characteristics of the service; allows public govern- Although these solutions have been announced as inment users to easily find, compare, purchase, deactivate teroperable, they do not include architectural compoand change services. nents such as those found in Sàvio. The Supervisor allows
Cloud consumers are confused when they choose to se- interacting in real-time with other deployments of the lect a cloud service provider (CSP) to build cloud services. same product in the Back Ofice, ofering to the commuThe Cloud Services Broker (CSB) is an entity capable nity of consumers useful information about the service of solving choice problems. The market surveys clearly provider, while the Agent communicates with end-users show which are the main CSP players: in 2020 according through a database-driven chatbot, to ofer them a tool to Gartner, we have Amazon Web Services (AWS), Mi- to support, guide and facilitate access to services. crosoft, Google, Alibaba Cloud, Oracle, IBM. Using the The study also analyzed the development of chatbots search engine www.mendeley.com and setting diferent for public government, based on the Osservatorio Agenda keywords for the search such as "Cloud Broker", "Cloud Digitale [31] of the Polytechnic of Milan. In particuService Broker", "Cloud quality problems Service Broker", lar, it emerged that most chatbots and virtual assistants "Cloud Services Broker for Cloud Services Provider Se- designed for a public government ofer first-level only lections" etc. It appears that a lot of research is based on information to users without actually interacting with the selection of cloud service providers based on quality back-ofice systems. and cloud brokering. The contractual relationship be- An example of a chatbot for public government is the tween Cloud service providers (CSPs), their customers, Roma Capitale project called ’Romolo’. This virtual assisand, above all their customer’s end users, are generally tant, managed with artificial intelligence tools, based on defined in a standard. Within cloud computing contracts, NLP (Natural Language Processing), conveys information there may be conditions such as free vs paid services; and FAQs to accompany citizens in their use of services US versus EU jurisdictions; IaaS vs Software-as-a-Service in the territory. The Municipality of Milan launched its (SaaS). information chatbot on WhatsApp. A more advanced
Because cloud computing is a distributed model, data experience is represented by the Municipality of Siena may be stored and processed on multiple data centers and [32], which allows users to make requests in addition in multiple jurisdictions. The general principle is that in- to having immediate and advanced information (such as formation risk owners will remain responsible for the in- certificates). formation risks that the department owns or guards. The Other chatbots are used in the City of Markham [33] only resources the public government owns are the in- (Canada’s city) to answer COVID-19 questions. The Mariformation in the service, the information assurance, and copa County Clerk’s Ofice deployed the virtual assistant the associated reputational risks. Risk management, the to improve eficiency for its employees and residents [ 34]. relationship between small public government and large The Government Technology Agency of Singapore (Govcloud service providers, and the governance of service Tech) and Smart Nation and Digital Government Ofice delivery are certainly the biggest issues to be addressed. (SNDGO) [35] have been exploring the use of virtual as
At present, the following research areas exist about sistants (VA) and A.I. to improve government services. cloud aspects, application interoperability, and chatbots: The chatbots, just listed, do not allow contributing to the governance of the provision and/or service. Our solution • Systems that analyze and aggregate web services fulfills these prerogatives by providing valuable support [25]; for internal users within the organization. The analysis of • Chat-bot for Public Government [26] the Sàvio model was conducted based on the AGID guide• Tender documents for public government- lines [36]. In particular, the tenders issued by Consip [37] oriented cloud services [27]. were analyzed, from which, in addition to the details of We proceeded to verify the design of calls suitable the cloud services, regulatory ideas emerged that led for the chatbot implementation after conducting a pre- the study towards the definition of a detailed technicalliminary analysis of the documents published and peer- regulatory scenario. Concerning the Sàvio model, nareviewed [28] [29] [30]. Further investigation was car- tional and international regulations were contextualized, ried out with the help of the AGID Italian marketplace to paying attention to the issue of the responsibility of all identify solutions similar to our solution illustrated in the actors in the field of cloud services between PA-SupplierProfessional. The study also emphasized the importance (QoS) metrics, and the mechanisms for auditing service of academic and post-academic training of the resources delivery and QoS, and compensating clients for under needed to carry out tender/contracting activities. This performance. The AUP, sometimes called a ’fair use polaspect is often ignored in the tender process. icy’, is a policy to protect CSPs from the actions of clients,
In the following paragraphs, we will describe the ap- and in the case of enterprise clients, their end users, by proach used to manage the cloud services provisioning detailing prohibited uses of the contracted cloud service. integrated with database-driven virtual agents. We will Privacy Policy details the CSP’s policy for handling and also present the service created to support public admin- protecting personal data, in line with data protection law istration in managing shared risk. More specifically, we requirements. The migration phase to a cloud infrastrucare going to describe the approach used for building a ture can be greater than the cost of managing an existing database-driven virtual assistant in IBM Watson. In ad- infrastructure. The consumer or the public administradition, we will introduce the additional service created tion calculates whether the migration costs and the fully to support the Public Administration in the adoption of operational costs in the cloud environment will be able to cloud solutions that can integrate the functions of digital ofset the costs of the existing infrastructure. The return marketplace services. on investment (ROI) of cloud infrastructure adoption can be calculated through the formula ROI= 3. Sàvio Sàvio is a framework that guides the cloud consumer in the adoption of cloud technologies through opportunities Bt are the tangible benefits, Bi are the intangible benand risks. Our solution ofers the following advantages: efits and TCO is total cost of ownership (TCO) that is decrease assistance cost; track the cloud services deliv- obtained from the sum of initial costs (Ci), recurring costs ery, simplifying the work of government oficials and (Cr) and termination costs (Ct). If the On-Permise solure-balance the bargaining power between small organiza- tion is used, the TCO can be converted into production tion and cloud service provider by creating an automated cost considering the following factors: cost and average reputational center. local hardware duration; current costs for servers and The key cloud players are: networks; total invested capital; structural costs (costs of electricity, rent and management of the premises); cost • cloud consumer (CSC): who uses cloud computing of human resources for infrastructure management; cost and signs a contract with the cloud provider; of application migration; cost of staf training; cost of • cloud provider (CSP): subject responsible for mak- partners and third party tools; costs not mentioned but ing the service usable to interested third parties; resulting from monthly invoices. It is possible to estiCSP provide cloud services; mate the efort in moving software solutions to the cloud • cloud broker: intervenes between Cloud Con- through the WideBand Delphi Techniques [39]. sumer (Public Administration) and CSP by ofering brokerage services, aggregation of necessary 3.1. Sàvio Model cloud services with existing resources and arbitrage; Organizations need to approach the cloud with responsi• cloud auditor: perform audits about privacy, per- bility. The cloud consumer or the Public Administration formance, security, regarding the services pro- must consider and negotiate Service Level Agreements vided by the CSP expressing opinion on the mer- (SLAs) to avoid the worst scenarios such as: increase in its; costs upon renewal of the contract; interruption of com• cloud carrier: cloud broker that provides con- mercial operations by the service provider or supplier nectivity, transport and interconnection tools be- without any migration plan to other, more economically tween the cloud consumer (PA) and CSP. advantageous platforms; commercial disputes between the service provider and the cloud service provider; drastic reduction in the quality of services. The present research analyzes the governance aspects of the provision of cloud services and builds a model of process and control of the supply: the Sàvio model. This model requires that SLAs are described by measuring a set of target values (SLO: service-level objective) and performance, reliability and result indicators (SLI: Service Level Indicator).
The contractual relationship between CSPs and other actors, including the public administration, is typically defined in a standard form which includes the following components: Terms of Service (TOS); Acceptable Use Policy (AUP); Privacy Policy; Service Level Agreement (SLA). The TOS set out the provisions that define and regulate the overall relationship between a CSP and the client. The SLA [38] details the level of service to be provided, often in the form of specific quality of service
the provision of cloud services: pre-contractual analysis and initialization of the Sàvio model; onboarding, pre-production and production; contractual termination.
In the initialization phase of the model, Sàvio analyzes the contract and identifies the SLA indicators. The preproduction, on-boarding and production phase includes constant monitoring of the SLA indicators. In this phase, failure to comply with the SLA determines contractual actions consisting of: application of contractual remarks (R), application of contractual penalties (P), suspension of contractual efect (S) additional actions (request for damages, early termination of the contract, etc.) In the Sàvio model, any action is promoted by the cloud consumer, or by the Public Administration, in relation to a single objective (for deadlines not met).
Common metrics for SLAs are the mean time between/to failures (MTBF), the mean time to repair/recovery (MTR), and Mean Time to Failure (MTTF: average time to failure that measures the average time of occurrence of a system failure or malfunction or the time average uptime). MTBF = MTTF + MTR. Assuming that the first fault occurs at time t1, it will take a time t = MTR for the repair to take place and a further time t = MTTF for the second fault to occur at time t2, therefore: MTBF = t2 - t1 = MTR + MTTF. We will therefore have that: = ∑︀ (+1− ) the establishment of an automated reputational center.
In this context of application cooperation, the good per
In a Sàvio model MTBF just gives a yardstick by which formance of the single cloud service supply becomes a given company’s SLA can be compared against another very important for any service provider. Sàvio identify in the Sàvio supervisor. The probability that the cloud critical Shop Identification - Unit Time: Average Numsystem is in an operable and usable state when the service ber of Failures (ANF) - Shows an average of failures ocis requested at a random time is the availability: curred in a given time period in numbers; Number of = + supplier errors - number of contractual reports small customer supply index (SCSI): it is an index imposed by
After the study of the Sàvio model, the architecture of Sàvio which compares the number of small customers the software system was designed through interoperabil- supplied with the number of large customers. This inity schemes. Sàvio software is a system that manages an dex obliges suppliers to serve a predefined number of application and a chatbot with user-generated content small public government based on the number of large as well as an advisor system for monitoring the risks, organizations served. Sàvio also identify of critical softquality of services, costs of cloud services adopted by ware modules - Units of time: Availability; MTBF; MTTR; the Public Government. The chatbot guides users to the Failure rate (FR):1/MTBF. Sàvio applies decision-making correct representation of the problems that afect the techniques to execute identification of critical solutions significant parameters for the consulting system. Three based on the attributes identified above and calculates a Sàvio software modules have been built: Sàvio supervisor, cloud service provider ranking based on user feedback Sàvio back ofice (BO), and Sàvio agent (figure 1). Sàvio using the PROMETHEE method[40]. Supervisor is the external interoperability application Sàvio back ofice is a web app that allows to govern component that allows you to automatically establish, service/product provision, also tracing the events of the by means of information flows from the N deployments supply in a knowledge base. Sàvio Back Ofice (BO) conof Saviò back ofice, the reputation of the cloud service / sists of: Sàvio Back Ofice Core that contains the business product provider according to predetermined parameters. logic of the application and implements all the functional
The prerogatives of Sàvio Supervisor are: to moni- and interface services of the solution; interacts in an autor supplies; re-balancing the bargaining strength be- thenticated and protected manner with Sàvio Supervisor tween cloud consumer or Public Administration, espe- and Sàvio Agent; Sàvio Back Ofice Web User Interface cially small ones, and service / product suppliers through (WUI) that designs the user interface and manages the in the UK www.digitalmarketplace.service.gov.uk, in interaction with it by exploiting the interface services Italy https://cloud.italia.it/marketplace/, and in the USA of the core component; Sàvio Back Ofice Archive, the https://www.fedramp.gov/. The open data available in database with which the Core component interacts to the marketplaces have been analyzed and then imported store and consult information. Sàvio Back Ofice core into Sàvio’s database, after classifying possible anomalies. is based on Spring Model-View-Controller (MVC), one Specific entities have been built in the Sàvio archive for: of the most popular open source frameworks for devel- recurring themes (thematic areas) such as Taxes, Citizenoping high quality java applications. The heart of the ship, Commerce, Culture, Elections, Family, Public Works, framework consists of an Inversion of Control container Work, School, Social Services, Sport, Public Construction; that manages the entire life cycle of the objects in the management of defects, problems or failures that could application context, from configuration, to finding depen- compromise the correct functioning of cloud products or dencies and creating individual instances, all through De- services. The user of any system can identify a defect, pendency Injection. Sàvio Back Ofice Core implements have a doubt and report it to request help or a solution. REpresentational State Transfer (REST) ful Application The user forwards the reports by telephone or forwards Programming Interface (API) based on the OAuth2 secu- them to Sàvio Agent. The Sàvio BO implementation rity protocol. An OAuth token is required for any REST made it possible to define all the management functions API call. Application Programming Interfaces (APIs) are relating to the governance of supply and reporting. The foundational to a modern digital ecosystem. These stan- deployment of Sàvio BO was done on IBM BlueMix using dards govern how APIs are to be developed across the the Bluemix CLI and Cloud Foundry CLI. The Application Government of Canada (GC) to better support integrated cooperation between Sàvio BO Core architectural comdigital processes across departments and agencies. Sàvio ponents was implemented through RESTful API authenBack Ofice WUI is based on the Vaadin framework inte- ticated with OAuth 2.0 Authorization Framework. An grated into the Spring (MVC) pattern. Vaadin Framework OAuth token is required for any REST API method made is a tool to build good-looking Web apps without work- available by Sàvio BO Core. The Sàvio Agent chatbot ing with low-level Web technologies. The framework was implemented on the IBM cloud platform where we itself contains all the logic to create the modern Web app defined: intents, entity, dialog tree and web-hooks. The while you concentrate on the UI itself, using a familiar ifrst Intent is related to frequent greeting forms such as: component-based approach, almost like you’re building "Hello", "Good morning", "Good night", "Good evening". a traditional desktop app. Sàvio Back Ofice Archive is Then two options are proposed: "I am using your services based on PostgreSQL. and I have problems", "Other issues". In the study of Sàvio
Sàvio Agent is a chatbot that uses the public govern- Agent, it was very important to define a guiding node: ment knowledge base that was built with Sàvio BO. Sàvio it is the way to restart the dialogue in case the agent Agent invokes Sàvio BO core API services through HTTP recognizes the #Restart intent. In this context, the first calls with post method from one or more dialogue nodes. level of interoperability with Sàvio BO Core is defined This mechanism is activated when Sàvio Agent processes through the activation of Webhooks. a node that has a call-out enabled. The chatbot collects This mechanism allows you to call a service RestFul the data during the conversation with the user and saves API with authentication token in method post HTTP. them in the context variables, subsequently transmitting The webhook calls the resource exposed by Sàvio BO the data as part of a HTTP post request to the URL of core passing the parameter containing a JSON structured the restful API of Sàvio BO core (listener). The listener as follows: ""description": "value" where "Value" is the performs a predefined action using the information trans- value of the context variable assigned during the dialog mitted to it in the definition of the Sàvio Agent call-out. interaction with the user. Sàvio Agent sends a callout to Subsequently, the chatbot optionally returns a response Sàvio BO Core and waits for the details of the topic, then to the user also based on the call-out response. proposes the answer with multiple options or as one URL referring to a website for more information. Sàvio Agent 3.2. Case study research and applications also allows users to forward reports to the back ofice. The user is initially directed to the FAQs registered in The Sàvio model has been applied to public adminis- the Sàvio BO Archive. Sàvio Agent performs a callout on trations and local authorities. Local authorities are nu- the resource "/rest/v2/ entities/savio_defect/search". The merous in every national territory and do not have chatbot allows the user logged in to also forward new human resources with the skills to manage the re- reports to the back ofice. lationship with service providers. Often, cloud ser- This technology approach ofers the ability to quickly vice management and service contracting are delegated and easily extend system functionality and integrate with to the service provider. The study analyzed a num- multiple interoperable back-ofice systems. In fact, the ber of government digital cloud marketplaces such as: virtual assistant embeds interoperability nodes with back ofice management systems. To add new actions, it is not necessary to modify the dialogue flow, but it is suficient to use the management functions of Sàvio BO. Supply governance and feedback management with Sàvio Supervisor, would have an extremely positive impact in terms of supporting all organizations in their role as consumers of cloud services. Sàvio Supervisor can support each country’s cloud services market for constant monitoring and control of the cloud services market.
provision of cloud services for consumers and the Public Government; an interoperability model between diferent distributions of the same product and the guidelines that can be used in the design phase of database-driven chatbots. To do this, we have implemented an interoperable software architecture, a database-driven chatbot, designed to meet specific requirements of highly useful use cases.
Sàvio is the result of the integration of the following innovative solutions: back-ofice solution with Spring MVC and VAADIN web interface; interoperability solution based on ResutFul API; database-driven chatbot solution developed on IBM Watson Assistant. This architecture has been enhanced by introducing the OAuth 2.0 framework and developing core software components for managing users and authentication tokens.
The guidelines for the design of database-driven chatbots have numerous possible areas of applications. Chatbots are also promising to reduce assistance costs and make the execution of recurring tasks easier and more immediate. Tests have shown that the proposed solution can be easily implemented in a cloud environment and provides the following benefits: to the end-users to get immediate and guided access to a help system; to the cloud consumer to have precise control over the individual steps of the cloud service; to the market surveillance authority of the individual country suggests a solution for the constant monitoring of the cloud market.
Thanks to the use of standard and open source products, the experience can be easily replicated to other clouds infrastructure. The final results contribute to the advancement in knowledge of the Cloud computing market and related software as a service.
As future work, we intend to focus on the challenges implementation of risk assessment tools and timely monitoring of the cloud market, starting from the empirical analysis of the development tools adopted, and the concepts identified in Sàvio Supervisor. 11–18