In a previous post, we wrote “the era of Artificial Intelligence is here” as it relates to topics of conversation from industry conferences this past Spring. This post serves as a primer on AI in regulatory software and provides guidance on the questions government regulators should be asking relevant to AI.
Having an approach to AI
At the broadest level, the first question for software vendors is their approach to AI. There are typically three responses to this (aside from saying they don’t have an approach):
- #1: We built AI directly into our software
- #2: We bought a third-party AI tool to integrate with our software
- #3: We can integrate our software with third-party AI tools
For each approach, there are different considerations that we’ll discuss later on. With Oracle, we support approaches #1 and #3. AI is built directly into Oracle Permitting and Licensing (OPAL), and our AI APIs allow customers to integrate our Fusion Cloud applications to third-party AI tools.
Understanding AI strategy
The Next question to ask is: Does the software vendor have an AI strategy? Or put another way… is there a guiding principle for how AI is being used with their regulatory software?
At Oracle, we are focused on deploying AI that enhances what users – both agency staff and customers – can do without removing the “human-element” from what is being done.
Examples of this in Oracle Permitting and Licensing include:
- Generative AI can author a communications message while allowing the agency staff the ability to edit the message prior to sending it.
- Retrieval-Augmented Generation (RAG) AI can allow an agency to upload its municipal code or other ordinances, and then provide answers to questions based on the uploaded files through a conversational interface.
- AI Agents allow users to make changes across multiple transactions at the same time (i.e. reschedule all of next week’s inspections for the following week) but prompts them with “are you sure you want to make these changes” prior to them taking effect.
While keeping the user involved in these processes is important to ensure that the AI is working as planned, we’ve learned from conversations with customers that it serves a dual purpose in helping alleviate concerns that AI will replace people. To that end, AI is not replacing people but rather reducing the amount of time and effort people spend on simple, redundant tasks, allowing them to shift more of their attention to complex tasks that AI cannot easily streamline.
Finally, a list of detailed questions to discuss with software vendors on the “tactical” considerations for AI. Where appropriate, we have included a response for AI delivered as part of Oracle Permitting and Licensing in bold.
- What is your company’s history with artificial intelligence?
From the beginnings with “classic” AI to machine learning, generative AI, and agentic AI, Oracle has been delivering AI across our cloud applications and infrastructure for 15+ years.
- Did your company build the AI features of your regulatory software? If not, how are those features integrated, and who is responsible for testing/fixing features that may be impacted by software upgrades?
Oracle built the AI delivered in Oracle Permitting and Licensing. We test features as part of the development of software upgrades and are responsible for fixing features that are impacted by those upgrades.
- What is the cost for using AI with your regulatory software? If there is a cost, how will it increase over time?
Oracle believes AI will provide significant benefits to the government regulatory industry. We do not want the decision to use AI to be financially-driven, and as such, there is no cost for the AI features delivered in Oracle Permitting and Licensing.
- How often will new AI features be added to your regulatory software?
Oracle is investing substantial development efforts to deliver AI features as part of our quarterly software upgrades.
- What is required to enable AI in your regulatory software?
Some AI features in Oracle Permitting and Licensing are enabled with a simple toggle switch in your setup configuration menu. Others are delivered through system configuration that lowers the technical complexity to deploying them.
- Do you provide AI training?
Yes, Oracle has a comprehensive training program on artificial intelligence, including industry-recognized certifications that can be obtained.
- Can customers build their own AI features?
Yes – Oracle provides customers access to AI Agent Studio where they can build and deploy their own AI features.
- What options are available for large language models (LLMs)?
Oracle provides customers with access to several leading LLMs. Should you want to choose your own LLM, we provide the option to do so.