AI in SimpleOne ITSM: How to automate zero and first line support with AI BPA

Automation usually starts with routine tasks. They are the first to be entrusted to technology to free people for more complex work. But generative neural networks are changing the rules of the game: they can handle even what previously seemed impossible for machines. This piece will discuss how AI BPA systems help automate zero and first line support in [ITSM](). ## What tools are needed to automate ITSM using generative neural networks Support automation requires not just neural networks, but a built system that can integrate them into workflows in a flexible and efficient manner. It's important to understand what tools are needed to make this integration deliver real business value. To automate zero and first line support in enterprise systems, you need to use AI BPA (Artificial Intelligence Business Process Automation) class systems - AI-enabled business process automation systems. But choosing the right AI BPA system is not enough. For it to truly become a driving force of automation, it is important to consider the key requirements and parameters that will ensure its stable and efficient operation. ## Requirements for an AI BPA system **Neural network orchestration** AI BPA systems must support the orchestration of generative neural networks so that they are not dependent on a single vendor. The rapid evolution of the neural network market, where new models with higher cognitive performance, better performance, and lower cost are constantly emerging, makes this approach particularly relevant. Being tied to a single neural network or a particular way of working with it creates risks. After three months, it may be necessary to switch to a new model to meet market requirements. **Abstraction layer and universal protocol** The AI BPA platform should provide an abstraction layer so that refinements do not strictly address a specific neural network, but can communicate with different networks via a universal protocol. This will allow neural networks to be updated quickly and securely. **Version management and monitoring** The AI BPA platform requires version and configuration management of large linguistic models, as well as monitoring their performance and resource utilization. It is important to ensure that the neural network provides acceptable SLA and response times, can handle the required load, and outperforms the previous version in terms of cognitive performance. **Business process design tools** The information system should include a variety of tools for developing business processes using artificial intelligence. **Visual workflow designer** The visual workflow designer should provide built-in components and specialized blocks that support intelligent automation. It is important to provide the ability to save business process integration templates and patterns for later copying and replication across departments and different use cases. **Tools for tracing and debugging** Tools for tracing, debugging, testing processes and logging actions are required to quickly identify and eliminate errors at any stage. **Development of intelligent widgets** The system should provide customers with the ability to develop their own intelligent widgets that interact with AI and simplify the interaction interface between the user and artificial intelligence. **Autonomous AI agents** There should be functionality to create and execute operations by AI agents that work autonomously to accomplish specific tasks. **Corporate supportability and risks of custom integration** The solution must provide high maintainability, stability, performance, and cost-effectiveness. Rapid custom integration with commercial LLM comes with risks such as model update failures, increased logging complexity, and lack of monitoring. The result can be an unstable and expensive solution that won't work for large enterprise organizations. Businesses need reliable enterprise-systems that are easy to maintain, provide stable operation and guarantee a certain level of service. ## Criteria for selecting an AI BPA system For effective automation using AI BPA, the system must meet the following key requirements: **Basic tools** Key requirements include a workflow editor, business rules and an internal API. The key focus should be on two aspects - embeddability and deep integration of AI functionality into the platform toolkit. **Reliability components** Connectors to various neural networks, load balancers, backup connection configuration tools, logging management and connection security systems. These elements ensure reliable and efficient operation of the system at the enterprise level. **Neural network management** For AI BPA to work effectively, it is important to have a complete API and tools for managing neural networks and models. A universal approach is fundamental, allowing the use of a single script to address different neural networks and their functionality. **Knowledge repository management** To implement the first line of support, the platform should support the creation, management and updating of vector knowledge repositories, which allows systematizing and efficiently using the accumulated data. **Query processing** When processing each request, the platform accesses the vector repository, which allows the neural network to use relevant data to solve user issues. This includes typical incidents, frequently asked questions, known errors and described problems. This approach provides automated advice in the context of a particular call. **Possibilities of chatbots** These mechanisms enable chatbots that accept queries, access the knowledge base, extract relevant context from it, and generate a response that matches the query. Instead of directly citing sources, a user-friendly answer is created based on data analysis. The bots are able to dialog in the customer's language and explain related issues during the communication process. For a successful AI BPA implementation, it is important to ensure integration with various data sources and the availability of tools for flexible customization and management of neural networks. This approach allows the platform to effectively automate processes, providing more accurate and personalized interactions with the user. ## AI BPA on the SimpleOne platform **Ainergy's AI BPA on the SimpleOne platform** is fully compliant with key enterprise-level requirements. [Low-code platform]() provides fast and convenient realization of task processing scenarios, and AI BPA allows to integrate AI functionality into business processes.

The product integrates deeply into all platform tools and provides LLM capabilities for various automation scenarios. The embedded AI BPA system automates processes **on any support line**, significantly increasing their efficiency. The platform supports integration with popular repositories such as Confluence, SharePoint and file repositories. It enables vectorization of data for neural network operations, as well as regular updates and addition of new data in an AI-accessible format. ## Zero Support Line and traditional approaches to automating it Zero-line support is an automated interface that attempts to meet the user's needs without involving a human first line. Traditional zero line options include a self-service portal. The user fills out a typed form, generating a request for a specific service, product, or service, and selects data from directories. The structured request is routed to a team of implementers, bypassing the first line of support. Alternatively, the zero line can be represented by scripted chatbots programmed to solve problems through a pre-defined decision tree. The chatbots, relying on a manually populated knowledge base, use keywords to search and direct the user to relevant articles. Such systems solve a significant portion of standard queries without passing them on to the first line of support. ### Challenges of this approach in the B2B segment The problems with traditional approaches in the B2B segment are that they require significant effort to structure and maintain processes. Scripts need to be constantly updated and new requests need to be added. This process is complex and time-consuming, which is only justified in the mass segment. For example, when it comes to a simple service accompanied by numerous and often repeated requests. In the corporate segment, the number of requests is much smaller, they require deeper elaboration, and the variability of requests is much higher. ### Problem solving with generative AI The problem can be solved with generative AI, which can make more complex inferences than traditional chatbots, support clarifying questions, and provide better user interaction. In corporate support, customers expect a higher level of service than those who, for example, are trying to figure out how to change a SIM card on a telecom provider's website. They are used to being presented with an article, a form to fill out, and the difficulty of contacting an operator. In a corporate environment, users rely on a personal manager or high-level specialists to quickly resolve their issue. Chatbot scenarios are inconvenient for them as they are time-consuming. An intelligent chatbot allows users to get answers to their questions from the database faster and more conveniently, without going through long trees of scripted solutions. It simply formulates a query and receives an answer or no answer, such as: "I will now switch you to a first-line operator". This approach reduces the time a corporate user interacts with support, increases their satisfaction and improves the quality of resolving requests, providing a better user experience. For the company, this is beneficial because it is not necessary to build a complex process to maintain the entire structure of possible solutions to a request. Long customization is not required: detailed classification of requests is not necessary, data can be poorly structured. The main thing is that the information should be up-to-date and consistent. In this case, the AI chatbot will be able to find a suitable solution if it is contained in the database. Support costs are minimized in terms of manual labor, making this approach easier, cheaper and faster to implement. ## How AI automates support in SimpleOne ITSM

#### 0 support line In ITSM SimpleOne, AI automates zero line support by answering user queries using AI agents if the information is contained in a knowledge base. Users receive information about known bugs, such as: "We are aware of this issue, it will be fixed in the next version. In the meantime, you can use a workaround," indicating the appropriate solution. The AI agent can also be configured to accompany its detailed, personalized responses with links to relevant articles from the knowledge base. In this case, the user can go and read the primary source, assuring that the AI agent is providing information based on real-world materials officially published for users. If the user's question is not fully articulated, the AI agent can refine the request in several iterations, clarifying the details and closing the need completely without passing the request to a first-line support specialist. #### Between 0 and 1 support lines At the junction between zero and first line of support, AI functionality can categorize requests that could not be resolved automatically and automatically route them to the right group of performers. This process can work either in fully automated mode or in assisted mode, where AI helps a person select the right service or determine which group to route an appeal to. If the process is already proven and clear routing and classification rules are in place, automation can relieve the burden on the first line of support. #### First support line The first support line handles only those referrals that could not be categorized or resolved automatically. It also performs those procedures that cannot be automated. If there is an internal specification that requires human execution of a request, the zero support line neural network passes the request to the first line, classifying it as a task for the first line executor, which is responsible for executing this type of requests. It is possible for the zero support line to pre-classify and process the request without involving the first line. However, the first line performs the control function - checks the classification results and, if necessary, confirms the correctness of the request processing. The first support line, receiving requests for which there is no solution in the knowledge base, closes these non-typical requests with the help of AI-assistant. It finds similar requests that were closed earlier and offers solutions based on the information already found. Complex queries are solved by highly qualified specialists of the second support line. #### Keeping the knowledge base up to date Keeping the knowledge base up-to-date in SimpleOne is automated due to the presence of a back loop of intelligent knowledge accumulation. AI agents analyzing solutions on the first and second lines automatically offer information to replenish the knowledge base with new solutions and identified errors. They analyze the solutions that the first and second line experts described when closing the requests not solved on the zero line. In this way, the system can automatically extract from the latest solutions similar appeals that were closed in the same way and automatically create articles for the knowledge base. These articles are published after approval by the knowledge base manager. The point of human control is important. Knowledge base managers receive drafts of articles with a proposal to publish them. Depending on the quality and compliance with corporate standards, they can either approve the articles or make necessary edits. This ensures a constant flow of new material and updates to the knowledge base, with the entire process automated and under the control of the responsible person. ## Implementation example To successfully automate support processes and improve service quality, many companies are turning to innovative solutions. One such example is the implementation of Ainergy's AI BPA platform for international cloud provider Serverspace. SimpleOne helped Serverspace, an international cloud provider, solve customer support issues related to overloaded support staff, language barriers, and manual request processing.