AI chatbots for ERP: Assessing the benefits and tools
Generative AI and other machine learning bots are coming to ERP in a big way. Here's an overview of the terminology, the pluses and minuses, and the top tools and vendors.
Organizations are increasingly using AI-powered chatbots to automate interactions with their back-end ERP systems. It is easy to see why. Chatbots can significantly improve the customer and employee experience and simplify user interaction with complex ERP systems and the various applications integrated with them.
Rich Sernyak, U.S. SAP practice leader at PwC, said the consulting firm predicts that every enterprise application will eventually enable chatbot interfaces. It's the first step in bringing generative AI to existing applications and workflows. The bots will make it much easier to interface with the data in applications.
At the same time, it is important to consider vendor-neutral chatbot platforms, according to Sernyak. Such tools can provide additional value when the data and workflow being transformed go beyond any single application.
Recent innovations in machine learning and generative AI techniques, such as large language models (LLMs), are dramatically simplifying development of chatbots in general and ERP chatbots in particular. More sophisticated companies might customize these generative AI and bot platforms for their unique workflows and use cases. They might also turn to the chatbot capabilities already being implemented in the top ERP platforms.
It's also worth noting that these tools come with assistive capabilities that run the gamut from older interactive chatbots to new digital assistants or copilots designed to work more in line with existing application UIs. In practice, enterprises must decide what type of experience to prioritize to provide the greatest value and ease of use to customers and employees.
Importantly, none of these tools are perfect replacements for human agents or experts. The underlying LLMs driving the most advanced chatbot experiences have a propensity to hallucinate (give out wrong or inaccurate information) -- and that's in the best cases. In the worst cases, chatbots interpreting safety information could confidently and authoritatively make dangerous recommendations. Enterprises deploying chatbots and other digital assistants need to consider additional quality assurance testing and safety measures to identify and mitigate the impact of these AI hallucinations.
What are ERP chatbots?
An ERP chatbot is a software application that provides a conversational interface to a back-end ERP system. For employees, this might mean asking for scheduling information, retrieving supporting documents or generating a summary of data buried in multiple sources in the system. For customers, it could be looking up procurement information, estimating a shipping schedule or confirming the status of a return.
The massive popularity of the OpenAI ChatGPT service has helped inspire ERP vendors to add similar capabilities to their ERP tools and platforms. Sometimes, they might do it via APIs into cloud services, such as OpenAI services running on Microsoft's Azure public cloud. In other cases, vendors customize or fine tune other LLMs to work with their platforms.
"This is one of the fastest-growing areas for conversational AI," said Priya Iragavarapu, vice president of data science and analytics delivery at AArete, a global management and technology consulting firm. Most ERP platforms now provide either native chatbots or integrations with third-party conversational AI tools to enable intelligent virtual agents, and the capabilities vary across vendors, she said.
How are chatbots integrated into ERP?
Chatbots are only as good as the underlying data they are trained on and interact with. Poor-quality, missing and uncalibrated data will limit the accuracy of results and increase the risk of hallucinations.
"Chatbots and generative AI rely heavily on accurate data," said John Collins, change leadership and transformation lead at Easology, a vendor of software for mobile devices. Many companies have struggled with this for years, Collins said. Data cleanup initiatives take time and often reveal issues with how the data is created, modified and validated.
In addition to cleaning up data, it is also important to ensure that data from different sources is calibrated or aligned. In classic ERP systems, it was sufficient to create an overarching enterprise schema and then work with various data, development and business teams to align their data with the schema, according to Collins. Chatbots and LLMs expand the ability to query the unstructured data in procurement documents, maintenance manuals and data sheets on safe handling of materials. Consequently, it is worth investigating how these capabilities can be harmonized with the reams of structured data stored in the ERP.
Once the enterprise has achieved satisfactory accuracy, the integration process between chatbot capabilities and ERP systems will go much smoother. Teams can integrate chatbot functions into the ERP at this stage by using pre-built connectors, niche tools, ERP-specific chatbot templates, conversational apps or partner marketplaces.
The benefits of AI chatbots in ERP
The biggest reason for using chatbots in ERP is to improve the user experience and increase productivity. Conversational experiences make it easier to find appropriate information regardless of how complex the ERP is. They can also surface real-time information more quickly without the user having to navigate complex menus and buttons that made sense to the developer.
The traditional ERP UI development model involved building templates to assist users.
"Users want to converse with their tools rather than have to click multiple buttons and follow a bunch of steps to get the information they are looking for," Iragavarapu said. AI chatbots will enable users to do just that by presenting information more precisely and succinctly, depending on their needs. Additionally, prompt engineering and prompt management techniques can allow users to save commonly asked questions. These tools can also help automate the process of modifying parameters, such as employee or customer name or time interval, to speed up repetitive tasks.
The generative AI models baked into chatbots can help democratize access to complex ERP systems.
"Generative AI is very good at interpreting natural language, making it easier for users to interact with data and for administrators to configure and customize ERPs by describing what they want," said Bret Greenstein, partner and generative AI leader at PwC. This capability lowers the threshold for data query and software development skills.
Top chatbot platforms to consider adding to your ERP system
Iragavarapu finds it helpful to break down the types of chatbot platforms to consider based on the specific integration approach the enterprise wants to pursue.
Even though ERP vendors have developed different chatbot tools, their capabilities are available to development teams in different ways. For example, SAP Joule and Oracle Digital Assistant for ERP and SCM are baked directly into the ERP platform. In others, such as Microsoft Azure Bot Service, ERP-specific chatbots are provided in a marketplace that can use the domain expertise of various partners.
There are also tradeoffs in the way enterprises implement ERP chatbots, she said. Chatbots from ERP vendors integrate tightly with the vendors' systems but also lock you into their technology stack. In contrast, vendor-neutral chatbots allow more deployment flexibility across systems but may require more custom integration work.
Each tool also comes with various capabilities in natural language processing (NLP) and dialog management, developer experience, scalability and cost. No-code bot builders promise faster deployment, while custom development tools support increased flexibility. It is also important to evaluate pre-built content, security features, supported channels and NLP accuracy with ERP terminology.
"Rather than defaulting to your ERP vendor's chatbot, assess platforms based on your specific use cases, integration needs and enterprise requirements," Iragavarapu said. It's also important to consider time to value if a vendor platform can accelerate your deployment. The best platform will depend on your industry, size and business needs.
Here is Iragavarapu's breakdown of the ERP chatbot tool landscape:
- Pre-built connectors. Many chatbot platforms like LivePerson, Kore.ai, and Inbenta offer pre-built connectors or adapters for common ERP products from SAP, Oracle and Microsoft. They make integration easier.
- Vertical applications. Some chatbot vendors provide vertical options tailored for ERP use cases like supply chain, manufacturing, HR and IT service management. They have predefined workflows, dialogs and content.
- Chatbot templates. Platforms might offer templates or starter bots for scenarios like FAQs, lead generation and customer support. They can be customized.
- Low-code bot builders. Tools like Flow XO, Cognigy, and Hubtype let users build chatbots with minimal coding. Users can connect ERP data sources and define dialog flows visually.
- Conversational apps. ERP vendors like SAP and Infor offer conversational apps that can plug into their platforms with predefined interactions.
- Partner marketplaces. Products like Microsoft Bot Framework have a marketplace with bot templates and other tools built by partners.
George Lawton is a journalist based in London. Over the last 30 years he has written more than 3,000 stories about computers, communications, knowledge management, business, health and other areas that interest him.
