AI represents a transformative technology that is fundamentally reshaping how businesses operate, offering unprecedented opportunities to enhance efficiency, automation and decision-making.  AI’s potential to impact countless scenarios across many businesses create opportunities that are boundless and will transform the way we do business.

These limitless possibilities can be a bit overwhelming, creating complexity and confusion in evaluating the latest AI features. That is why Acumatica is very intentional with our approach to AI and innovation. First and foremost, we thoughtfully consider how to use AI to help solve business problems. This AI-first approach addresses business problems from the ground up: integrating intelligent capabilities thoughtfully into our platform and focusing on the real-world needs of small mid-sized businesses.

Our Chief Engineering Officer Miten Mehta recently shared how we are focused on delivering practical AI solutions in the first of this AI blog series.  Practical AI in ERP to Drive Powerful Results.  Acumatica’s AI strategy is centered on transforming the user experience, boosting operational efficiency, and offering users actionable insights from ERP data through advanced machine learning (ML) and AI. This strategy is structured around three pivotal pillars: AI Advisor, AI Automation and AI Assistant. In future blogs we will provide more details on each of these areas.

Today I want to share with you how we are delivering AI automation through the AI Studio component in our xRP platform – making it easier for businesses to integrate AI into their workflows through customization – putting the power of technology in our customers’ hands and making AI work for them.

 AI Studio – Putting Practical Power in Your Hands   

AI Studio will serve as a dedicated platform for designing, managing, and executing customized pipelines. It enables users to script precise actions, engineer prompts, and select appropriate technologies for specific tasks, ensuring expected outcomes. Through a suite of tools, AI Studio will support registration of tasks, definition of execution pipelines, and tailored execution contexts—enabling users to build complex, prescriptive AI workflows.

AI Studio is a part of Acumatica’s low-code/no-code customization framework, enabling users to define AI workflows for specific, prescribed scenarios. These workflows address complex processes that require detailed instructions, eliminating the need for users to manually provide these instructions every time. The prescribed scenarios can include anything from generating support case closure notes based on the case activity history to creating a website description for an item using a predefined template.

AI Studio – How it Works  

The flow of information, or interaction flow that takes place within AI Studio involves five separate stages – Action Invocation, Prompt Generation, Prompt Execution, Response Processing and Entry Update.

• Action Invocation: The process begins with the invocation of an action tied to a business entity. The action can be triggered manually by the user, automatically through workflow automation, or by an intelligent agent. This invocation initiates the AI action flow.

• Prompt Generation: AI Studio gathers the execution context, which includes data from the associated business entity and additional information from related generic inquiries, if needed. The execution content is then used to populate placeholders within the predefined prompt template, generating the final prompt for the action.

• Prompt Execution: The generated prompt is passed to the AI Service Gateway and the gateway identifies the appropriate AIservice based on prompt configuration and executes the request.

• Response Processing: Once a response is received, it is returned to AI Studio for analysis. AI Studio determines the next step based on the response, which may include things like requesting human interaction for result review or correction; using the response to generate content for a subsequent prompt; or directly processing the results to update the invoking entity.

• Entity Update: In the final stage, AI Studio extracts relevant data from the response. A script is generated to update the entity or execute actions based on the response’s result and executed over the entity the action was invoked from.  

AI Studio Components

AI Studio has three primary components that help users configure and manage these prescribed scenarios.

• Prompt Configurator: An administrative tool for managing prompts within the system. It allows administrators to perform a myriad of tasks including registering and managing prompts and configure prompt settings (i.e. link a Generic Inquiry (GI) to provide additional context; configure how data is extracted from the results and updated back to the invoking entity).

• Data Update Service: A two-step approach to process JSON files that will enhance reliability and flexibility when handling varied prompt outputs.

• Data Masking Service: A tool to enhance privacy and security by preventing sensitive information from being passed on to external AI and AI services.

AI Gateway  

The AI Service Gateway is designed to facilitate the interaction between external or internal AI services and Acumatica’s Assistants or XRP components. Acting as the interface between the Acumatica ERP application and ML services, this gateway performs several essential functions, including:

• Provisioning: For external AI services, the gateway provides a registration point where API plugins can be configured, and connection credentials securely specified. For internal AI services, it serves as the entry point for service provisioning, ensuring seamless integration for the ERP users.

• Licensing: The gateway enforces and monitors license restrictions, ensuring compliance with usage terms and preventing unauthorized access.

• Governance: It oversees the enforcement of usage policies, managing aspects such as concurrency, quality of service, and resource allocation to ensure consistent and efficient operations.

• Security: The gateway implements security controls, such as managing sensitive data when interacting with external services and preventing prompt injection attacks when users or APIs interact with assistants or agents.

• Telemetry – It collects and centralizes ML-related telemetry data, enabling insights into service performance, usage patterns, and operational analytics.

AI Services  

AI Services invoke the AI action, sending it back through AI Gateway to AI Studio for response analysis and then an instantaneous determination and completion of action.  Acumatica develops and hosts a suite of internal AI Services as well as enables integration with externally hosted services.

Internal AI Services 

Acumatica’s internal AI services are designed to enhance ERP capabilities with services that include:

• Fine-tuned LLMs – Tailored to Acumatica-specific information and telemetry, these models will power intelligent assistants and deliver contextual insights.

• Pretrained ML Models – Focused on automating common ERP operations, improving efficiency and accuracy in routine processes.

These internal AI services, provisioned and accessed through the AI Service Gateway, have seamless access to tenant-specific ERP data, telemetry, and the globally shared Acumatica Retrieval-Augmented Generation (RAG) vector database. This integration enables robust and personalized data-driven solutions.

External AI Services  

Acumatica will not restrict ERP users to internally provided AI services but will also enable integration with externally hosted services. These external services can include customer-hosted solutions, custom-developed agents, or publicly available services. Access to these external resources will be provisioned and securely governed through the plugin system hosted by AI Service Gateway.

Looking Ahead

Today we gave you a glimpse behind the curtain on how we are seamlessly delivering AI capabilities and making it easier for users to define AI workflows for specific scenarios.  In future blogs, we will share with you the amazing capabilities that will address the real-world challenges small and mid-sized businesses face through practical technology that you can use and understand – making AI work for you.