Governing Reimagined: AI's Role in Modern Public Services
Explore Amarillo’s innovative AI assistant, Deloitte’s comprehensive report on AI in government, and the practical performance of top AI chatbots
In this month’s AI-focused edition of TL;dr, we explore the innovative applications and assessments of AI technology in government and everyday tasks. We explore Amarillo’s groundbreaking AI assistant “Emma,” Deloitte’s analysis of how generative AI can transform government work, and a comparative evaluation of top AI chatbots’ performance in practical skills.
Meet Emma, Amarillo's AI assistant and 'digital human'
Route 50 reports Amarillo, Texas, has introduced “Emma,” a generative AI-based digital assistant designed to communicate audibly with residents in multiple languages. This capability is crucial for Amarillo’s diverse population. “Over the decades, the city opened its doors to tens of thousands of refugees from Africa, the Middle East, Asia and Latin America, becoming what The New York Times dubbed the ‘Queens of the Texas High Plains,’ comparing Amarillo to the New York City borough.” Developed in collaboration with Dell Technologies, Emma aims to enhance access to non-emergency city services and information through natural voice conversations. This project is so cool I’m offering four takeaways instead of the usual three.
Multilingual Communication for Diverse Populations: Emma can converse in over 60 languages, facilitating communication for Amarillo’s diverse population. Nearly a quarter of Amarillo’s residents do not speak English as their primary language and at least 62 languages and dialects are spoken across Amarillo schools. “The whole premise behind Emma is that instead of typing or searching for information, you can just have a conversation with her in the language you speak and get access to nonemergency city services and information,” says Rich Gagnon, Amarillo’s assistant city manager and chief information officer.
Cost-Effective and Efficient Service Delivery: Emma’s implementation is significantly less expensive than hiring additional staff and helps streamline city services, Gagnon said. Emma could remove some of the mundanity of current staff’s roles, allowing them to focus on other tasks, he added.
Citizens Helping Emma to Learn: The public is actively helping Emma do a better job. The city wants to make sure the software understands residents correctly before Emma makes a formal appearance on the city website, expected in October. “We’re bringing in residents to ask the same questions to Emma because one of our learnings is that you and I won’t ask the same questions the same exact way,” Gagnon said.
Looking Ahead to Future Enhancements, Building Trust: Gagnon said Emma’s current iteration is only the beginning, with future versions potentially teaching languages, summarizing city council meetings, and proactively engaging residents. “Our approach is to be a conversational city,” Gagnon said, emphasizing that building relationships through conversations and integrity might help improve the public’s trust in government.
Generative AI and government work: An in-depth analysis of 19,000 tasks
Deloitte’s analysis of 19,000 tasks explores the potential impact of generative AI on government work. The study identifies criteria such as creative difficulty, context variability, and accuracy to determine suitable tasks for AI automation. The report also discusses the dual-wave adoption of AI, with immediate impacts on knowledge-based roles and delayed impacts on more physically-oriented roles. Here are three takeaways.
Task Suitability Based on Criteria: Tasks with moderately high creative difficulty, moderate context variability, and moderate accuracy are good candidates for gen AI. Examples include preparing speeches, summarizing laws, making reports and recording regulatory compliance.
Dual-Wave Adoption Impact: Immediate benefits are seen in knowledge-based roles, with a delayed wave for physically oriented roles as AI technology advances. “Consider workers in government shipyards, highway maintenance divisions, or sanitation departments. The bulk of their day-to-day work is physical in nature, but they still often need to receive work orders, track tools, or record maintenance fixes,” the report states. “Embedding gen AI in maintenance management, inventory tracking, and other systems that these workers use every day can improve both the ease and efficiency of their work.”
Strategic Use of AI for Efficiency, Effectiveness, and Efficacy: “Work is made up of more than just single tasks and adapting workflows to use a set of different automation tools, each taking on the tasks to which they are best suited, can increase how effectively government accomplishes its mission, not just how quickly,” the report states. For example, to make an argument in court, government lawyers may need to do several tasks: Gather evidence about previous cases; analyze those cases for relevant evidence; and make a judgment and argue that judgment in court. Each of those tasks requires different skills, making them amenable to different types of automation. “Getting the work done would require not one monolithic AI tool but several smaller ones—working together with and supervised by human judgment.”
The Great AI Bot Challenge
This May 25 Wall Street Journal article examines the performance of five top AI chatbots — OpenAI’s ChatGPT, Microsoft’s Copilot, Google’s Gemini, Anthropic’s Claude, and Perplexity — through a series of practical tests. While the evaluation focused on everyday skills, including health advice, financial planning, cooking and current events, we’re focusing on work writing, creative writing, and summarization, which likely have the most application for government work. Each bot’s responses were judged on accuracy, helpfulness, and quality.
Work Writing Capabilities: Tone and detail matter in work-related writing. The tasks included writing a job listing for a prompt engineer and an office-appropriate birth announcement. The race between Perplexity, Gemini and Claude was close, with Claude winning by a nose. Final tally:
Credit: Wall Street Journal Creative Writing Capabilities: While Copilot crashed in work writing, it soared in creative writing. Given the prompt, “Write a wedding toast for Shara and Chris as told by the Muppets,” Copilot responded with, “Gonzo: ‘Ah, love! It’s like being shot out of a cannon into a pile of rubber chickens!”Claude was the second best, with clever zingers. Final tally:
Credit: Wall Street Journal
Summarization: Copilot provided the best summary of a Paul McCartney Wikipedia page with a skimmable outline and fun facts, while Claude struggled with handling web links. Perplexity emerged as the category winner for its consistent and accurate summaries, including extracting subtitles from a YouTube video. Final tally:
Perplexity emerged as the overall winner when looking at all nine categories, showcasing a strong balance of concise and informative answers, while ChatGPT, despite its recent update, did not lead the pack.
In Other’s Words
In his May 27 Wanderland column for The Dispatch, Kevin D. Williamson writes why Marjorie Taylor Greene should be expelled from the House for propagating the lie the FBI wanted to assassinate former President Trump when it served a search warrant at his residence in the documents case.
(A) free society cannot endure government by lie. Government by lie is not the same thing as a government that includes the usual share of lying politicians. Government by lie is what you get when lying becomes, in effect, a political creed of its own, a party and a movement unto itself. Lying is unpatriotic. It is dangerous. It is corrosive. … But Marjorie Taylor Greene is not only a social-media troll—she was that, but, thanks to the miracle of democracy, she is something more than that today: the honorable representative for the good people of Dalton, Georgia, and the surrounding area. But the House can expel her. It expelled George Santos for much less serious offenses.
Five days earlier, Nick Catoggio, another columnist for The Dispatch, put the MAGA world freakout over the FBI raid on Mar-a-Lago in the documents case in a broader context.
Indifference to the truth, not stupidity or malice, is the hallmark of the propagandist. There might be something sinister about the FBI preparing for deadly force at Mar-a-Lago or there might not be. What does it matter? The real question is, what does that have to do with helping Trump win?
I’ve always believed that that’s where Trump ended up psychologically in 2020 on the question of whether or not the election was really rigged. Initially, he seemed to be lying about it; by January 6, he appeared to sincerely believe that he’d been robbed. The likely truth is that he wasn’t sure and concluded that it didn’t matter a whit to his purposes. He wanted to stay in power and to do so he needed to convince a critical mass of Americans that he’d been cheated.
To ask “but was he cheated?” is beside the point.
We’ll leave you with this palate cleanser from Atomic Habits author James Clear, on managing your attitude.
When I go into a situation, I try to keep a mindset of baseline optimism.
When starting the day: It’s going to be a good day.
When meeting someone new: I’m going to like them.
When trying something challenging: It’s going to work out.
Problems will inevitably arise and I’ll deal with them as the situation demands, but my preference is to assume I’m on a winning path until proven otherwise.
Onward and Upward.