Discover how DHL Group increased connectivity across the enterprise and automated multiple processes end-to-end
What are AI Agents?
AI Agents are specialized assistants, powered by Gen AI to perform specific tasks. AI agents can be built using text-based prompts and generative AI. Once added to a business process, they can operate without human intervention within the process flow, as they are fed information and perform actions.
How do AI agents work?
Once AI Agents are created and tested, you can deploy them in your business processes. The process feeds data to them, the Agents can take actions or make decisions based on predefined rules or logic. Actions could be triggering other systems or processes, or handling information itself. This means humans don’t have to manually complete the task as it is handled by the Agent, such as reading and summarizing a document, or hiding personal, sensitive data before forwarding a document.
Agents work as part of a business process and perform tasks or trigger actions that would otherwise have to be manually carried out by a person. Some ways that AI Agents benefit businesses include:
- Accuracy: AI agents are programmed to perform tasks accurately, which avoids the risk of potential human error.
- Always-On: Continuously working in the background within a business process.
- Efficiency: AI agents significantly boost productivity by performing unprompted tasks to make users’ tasks easier.
- Data insights: They can read and summarize vast quantities of data, provide data summaries, and put data into context, such as checklists and highlights.
- Cost-effective: The build once, use anywhere composition of AI agents makes them cost-effective to deploy, and thanks to their accuracy and risk reduction, they augment the output of human users.
- Easy to build: The building process is so easy that you do not need any technical knowledge. So you do not need a data scientist or prompt engineer; just type your description, e.g. “I need a helper to translate Spanish documents into English.”
As AI Agents are specialized to perform specific tasks, there are numerous examples of what AI Agents can do, including:
- Content creation: Create new emails, documents, text, copy, knowledge base articles, & replies
- Categorize: Automatically detect and classify incoming requests, tickets, & cases
- Summarize: Summarize or outline long meeting notes, transcripts, articles, & research
- Prioritize: Automatically find and apply a priority to requests, tickets, & cases
- Obfuscate: Identify and anonymize personal or sensitive data for GDPR & compliance.
- Translate: Remove the skill obstacle of foreign languages by translating into any language.
- Organize: Organize disparate data into logical groups to simplify analysis.
- List Maker: Get things done by seeing clear lists of information.
- Analyze: Detect sentiment, tone, and logical groupings, & SWOT
On a low-code platform, like Bizagi, it’s very easy to create an AI Agent. It can be done in three simple steps:
1. Define title and general description: First, name and describe your agent, and create a prompt for Gen AI to define what you want the agent to do. Provide as much information as possible: GenAI needs the full context for the best results.
2. Use the assistant to create and test the prompt: Next, Bizagi’s prompt assistant will help you to test and create the prompt, identifying the different variables, helping you make adjustments to get it right. There's no need for a prompt engineer or data scientist, you just add the name and the description of your agent. The assistant generates the prompt and it identifies the variables needed for that prompt, then you can test it by manually adding values on each variable to see what the output is and determine the best outcome.
3. Apply the AI Agent to any processes you want: Finally, add the agent to your process model and you are ready to go! The agent will start receiving and acting on data immediately.
Are Gen AI assistants the same as an AI agent?
No, GenAI assistants and AI Agents are not the same. The primary difference is that AI Agents are used to make daily tasks more efficient, rather than GenAI assistants that provide data insights. This means they are used by different business personas.
For example, Agents are used by workers within their business daily processes to make them more efficient; such as an insurance agent needs to read information in lots of claims documents, so they can create an Agent to provide a summary of each claim so they don’t have to read the whole document.
Alternatively, a GenAI Assistant would be used by a manager or knowledge worker to help them understand what is going on with their data. They can create queries and provide business insights to gain insights and help make strategic decisions based on information.
Additionally, you have to feed GenAI assistants information and ask them to take actions: GenAI assistants are responsive, rather than proactive and their decisions require human prompts as and when information is needed.
For example, in Bizagi, you have Ada, the Gen AI assistant, who knowledge workers can ask within their apps to help with tasks, such as presenting data in a more digestible way. The user can then take action on that data. Whereas a translator AI Agent deployed in Bizagi business processes will translate without being prompted so that the user can immediately read and understand documents in a foreign language.
Both Assistants and Agents are fuelled by Gen AI and generate responses based on information. They are not predictive and cannot be trained or learn from past events: they will only execute the task they are programmed or asked to perform.
You can use two AI Agents to enhance the process of handling customer complaints. Prior to using AI, once the bank receives the complaint email from a customer, it is ingested by an RPA bot and a complaint case is initiated in Bizagi. The process includes two humans: a Data Analyst to read the complaint, categorize it, and obfuscate any private information; this takes about 10 minutes. The Analyst then passes on the obfuscated complaint to the Lending Manager to read, and create and send a response, which would take about 20 minutes: half an hour in total.
With the addition of an AI agent, Bizagi passes the complaint email over to an AI Agent to obfuscate the data and generate a new email that hides the private information. This is passed to a second AI Agent to read the obfuscated email and generate a suggested email response. The email response is then passed to the Lending Manager, who reviews the suggested email from the AI Agent, makes any necessary changes, and sends it to the customer. This will only take a matter of minutes, saving the bank significant time and money.
Bizagi AI Agents simplify tasks and drive efficiencies across the business, boosting productivity and improving user experiences. You can deploy multiple assistants in your business processes. They are easy to build using text-based prompts, so you do not need a prompt engineer or data scientist. Plus, they are reusable components, so once an AI Agent is created, it can be reused as many times as you need across different business processes.
Bizagi AI Agents fit into your existing business model. Bizagi users don’t need to add anything to their plan as it is part of the BPU consumption. Additionally, Bizagi is not charging for execution in the development environment, so you can run multiple prompts for free to tune the agents to perform accurately before deploying to a production environment.
Find out more: https://www.bizagi.com/en/platform/ai-agents
