Focusing City Government Resources on Using Generative A.I. to Serve Residents
I was recently invited to attend a City of Boston webinar on the use of Generative AI by staff, indeed the use cases were compelling and included helping with drafting communications, policy, and analysis. See more here.
Nonetheless, the transformation at the City side looks like it's just beginning, and the more powerful use cases will require more technical resourcing, and applying generative A.I. to transform entire functions within the government, such as citizen services or community engagement, thereby enhancing their productivity and effectiveness.
Key to this is:
1-Establishing a Data Governance Model
Before a city government can devise a generative A.I. strategy, it must first establish a data governance model. The advancement of generative A.I. highlights the need for a centralized, organization-wide strategy for collecting and governing data. We often see departments at the city level operating as silos, having a data governance model ensures that a ‘system’ of data tools and processes exists. Generative A.I. can aid in data excavation, programmatically assisting city governments in cleaning and analyzing large and unstructured datasets. City staff should prioritize the creation of centralized datasets and collaborate with data scientists to identify new and creative sources of data for refining generative A.I. models. In cases where data gaps exist, generative A.I. can manufacture synthetic data, enabling city governments to explore novel datasets and derive valuable insights. Supported by a governance model, this makes it possible to do securely and effectively.
2-Choosing LLMs
Given that a single or a small number of LLMs can serve various purposes, selecting the right one becomes a critical decision for city governments. While City Staff may not need to delve deeply into the technical evaluation of each LLM, they must understand the broader impact and recognize that choosing an LLM is also choosing a strategic partner. When selecting an LLM provider for a specific use case, a ‘score-card’ including three factors could help guide the decision-making process:
i-Degree of data confidentiality needed: City governments must carefully evaluate the options offered by LLM providers in terms of data security, whether it’s in public, securely managed, hybrid, or fully private environments. The choice should align with the city’s data security requirements, especially in light of the increasing threats.
ii-Internal resources required for implementation: Full-on implementations are resource-intensive, demanding a team with cutting-edge expertise in addition to operations engineers and IT professionals, which for many cities will be challenging to mobilize. Adapting a model in a cloud environment requires fewer resources, typically a small group of full-time data scientists or machine learning operations engineers. However, this model would be fully managed by an ML provider and automatically updated, limiting the city’s ability to customize it as needs evolve, and probably the most viable route.
iii-Implications for citizen engagement: City governments need to consider the impact on citizen engagement when integrating customer-facing services into a third-party LLM platform. While using platforms like ChatGPT can provide convenience, it may also risk alienating users (who has not been alienated by a chatbot!) and reduce direct citizen engagement. Building in-house, customer-facing chatbots using open-source tools may avoid these pitfalls but might miss out if third-party LLM platforms become popular sources of citizen engagement and interaction.
Conclusion
AI use — is not just about placing guidelines for staff but more so building the systems and tools for the future. City governments must develop a future-proof A.I. strategy to thrive in the rapidly evolving landscape. Rather than viewing generative A.I. as a mere tool, City Staff should embrace it as a revolution that will impact every aspect of urban governance and service delivery. While the future cannot be precisely predicted, city leaders must proactively prepare for what lies ahead
