Despite the many potentials of gen AI, it’s important for utilities to remember that gen AI is nascent — it’s still very new and it’s developing extremely quickly. Utilities should stay on top of developments in AI capabilities and the changing risks and benefits that arise from those developments.
Risk assessment and risk mitigation are critical components when integrating any new technology and should be embedded at multiple stages of utilities’ gen AI journeys. This will help ensure that the technology not only enhances capabilities but also maintains high standards of security, reliability, and ethical operation.
Implement guardrails for gen AI use
As in DC Water’s case utilities should set up guardrails for their employees so they can interact with AI tools safely. These guardrails may address concerns like inaccurate outputs and hallucinations, data privacy and security, biased decisionmaking, intellectual property infringement, potential misuse, and job loss.
For example, guardrails for AI use might include provisions related to how employees should cross-verify AI generated information to ensure accuracy.
A gen AI roadmap for water utilities
It’s important for utilities to consider how they will leverage gen AI systematically and responsibly. Utilities can create a roadmap for gen AI adoption in their organizations and implement change management frameworks to keep things on track.
Here, we present a simple roadmap for inspiration. While the roadmap is organized in a linear fashion, many of the stages could overlap with other stages, or require looping back to earlier stages as the journey develops and challenges arise.
The gen AI roadmap in detail
1. Awareness & education
Educate the utility’s leadership and staff on what gen AI is and its potential impact on the water industry. Communicate and discuss gen AI risk as a part of the engagement process, and transparently discuss potential downsides and how they can be mitigated.
Action items: Host informational sessions, provide resources and guardrails for experimentation and self-learning. List key capabilities of AI models and provide examples of how they can be used to save time, money, or other resources. Discuss case studies of AI in similar industries.
2. Establish responsible AI guidelines and principles
Develop a framework of guidelines and principles that ensure the ethical, transparent, and accountable use of gen AI within the utility. This framework should address key areas such as data privacy, AI fairness, security, and regulatory compliance.
Action Items: Establish a cross-functional team that includes members from legal, compliance, operations, and IT departments to develop, maintain, and update the AI governance framework. Collaborate with internal stakeholders and external experts to define a set of ethical AI principles that cover consent and privacy, fairness and non-discrimination, transparency and explainability, security and safety, and accountability. Create usage policies for AI tools that include guidelines on data handling, model training, output monitoring, and corrective measures in case of failure or unintended consequences.
3. Needs assessment & strategy
Engage employees across the organization to identify and prioritize areas within the utility where gen AI could have an impact. Even small efficiencies across roles, like time savings and error-reduction by automating data entry, can have a significant cumulative impact and possibly lead to staff freeing up their time to work on larger challenges. Assess and solicit feedback from staff about perceived risks related to gen AI tools being used in the organization.
Action items: Conduct workshops to map current utility challenges against gen AI capabilities and risks to assess their potential value and drawbacks. Have employees look at their own workflows and processes to identify repetitive tasks that gen AI tools may be well suited for. Create guidelines to help employees experiment with their own ideas to bring greater efficiency to their own role. Each employee could be encouraged to experiment with areas that they feel will help make their job more efficient as long as they stay within the organization’s gen AI guidelines. Prioritize opportunities based on their alignment with your utility’s goals — but also monitor and keep track of current and new gen AI capabilities, so you can incrementally integrate new capabilities as you decide on what tasks to take on next.
4. Stakeholder engagement
It’s important to gain support from all levels of a utility, from executives to field operators, for AI implementation. Utilities should also gather input and buy-in from all relevant stakeholders, including customers and regulatory bodies.
Action Items: Organize meetings with stakeholders to discuss potential AI uses and address any concerns about changes and impacts. Also take these opportunities to showcase what is working from early experiments to help educate stakeholders on the use of the gen AI capabilities.
5. Technology partnership
Select the right technology providers that align with the utility’s needs. Choose partners not only for their technological capabilities but also for their ability to support risk mitigation strategies. You may wish to look at large enterprise versions of gen AI services, like Microsoft’s Copilot, Open AI’s ChatGPT or GPT builder, or Google’s Gemini as well as bespoke, specialized AI tools that leverage gen AI models for specific applications.
Action Items: Evaluate different AI platforms and vendors, request demonstrations, and select partners based on technology compatibility and support offerings. Assess vendors’ experience with similar risks and their approach to issues like data security and ethical AI use. Ensure the AI solutions you’re implementing are well-tailored to the specific needs of your utility. That may mean utilities work closely with vendors to tweak AI models and applications to water-sector domain knowledge and data, as well as knowledge unique to your utility.
6. Pilot projects
Start small with pilot projects to test the effectiveness of AI solutions in controlled environments. Use pilot projects as a means to specifically test risk mitigation strategies under controlled conditions.
Action Items: Implement AI in both small-scale projects, such as automating customer queries or optimizing maintenance schedules, and small-scale tasks, like summarizing PDF documents and manuals or automating data cleansing and entry, then monitor the results. Monitor for any unforeseen operational disruptions, data breaches, or other risks, and evaluate the effectiveness of the mitigation strategies in place as well as the efficiencies the projects bring to the organization.
7. Training and development
Roll out proven AI solutions across the organization.
Action Items: Develop a detailed implementation plan, update internal systems and processes to integrate AI, and ensure all staff are familiar with new workflows. Develop contingency plans and disaster recovery strategies to handle potential AI failures or data integrity issues. Share smaller efficiency improvements and task enhancements with staff to help encourage and inspire similar advances.
8. Implementation and integration
It’s important to gain support from all levels of a utility, from executives to field operators, for AI implementation. Utilities should also gather input and buy-in from all relevant stakeholders, including customers and regulatory bodies.
Action Items: Organize meetings with stakeholders to discuss potential AI uses and address any concerns about changes and impacts. Also take these opportunities to showcase what is working from early experiments to help educate stakeholders on the use of the gen AI capabilities.
9. Monitoring and evaluation
Continuously assess the performance of AI implementations. Regularly assess not only performance metrics but also risk exposure and mitigation effectiveness.
Action Items: Establish KPIs to measure the impact of AI, such as increased efficiency, reduced costs, or improved customer satisfaction, and adjust strategies as needed. Explore advanced AI capabilities as they develop, such as predictive analytics for infrastructure management, and consider expanding successful AI applications to other areas of the utility.
The outlined roadmap provides a pathway from initial awareness to full-scale gen AI implementation, ensuring that each step is carefully considered and that risks are effectively managed.
Read our full whitepaper
This article is taken from Qatium’s recent whitepaper on: Water utilities & AI: how utilities can thrive in the new generative AI reality. If you’d like to read the full whitepaper, you can do so here.
