AI Upskilling: The Important of AI Literacy and How to Upskill Your Staff
Learn about the City of San José's approach to upskilling their staff on AI and how you can set up an AI upskilling program at your agency.
By Angela Zhao
Organizations in all industries are racing to integrate AI into their applications and the workforce is scrambling to keep up. As educational AI courses proliferate to grow the workforce, the City of San José is seeing the value of AI Upskilling firsthand. As one of the co-creators and co-instructor for the City’s inaugural AI Upskilling program, I’ll draw on insights from the City of San Jose’s approach to show why AI Upskilling is crucial and share tips on how to successfully upskill staff at your own organization.
Consider the impact from the City’s AI Upskilling program, which has been built and delivered in partnership with San José State University. This program has led to a 10-20% efficiency gain (100 to 250 hours saved annually) for each participant, translating to over 5,000 hours saved so far. 65 staff members from 19 (of 25) departments received 10 hours of targeted training over 10 weeks to create and share custom GPTs to help them with their work. Over time, training 15% of the City’s 7,000 staff members could save an estimated 300,000 hours.
Motivation
Why should we care about upskilling our staff, and why should we care about upskilling them with AI skills?
The benefits of upskilling staff are well-documented. Per the University of Pennsylvania, upskilling staff helps to improve staff productivity, employee satisfaction, and employee retention. Upskilling is particularly needed in times of technological upheaval when jobs and the skills needed for them are changing rapidly.
Additionally, the government has been tasked with doing more with less. Unlike the private sector, an increased demand for services does not correlate with increased funding or budget for staff. From 1960 to 2017, the US population has increased by 81% while the federal workforce has increased from 1.8 million to 2.1 million civil servants. The government's role of providing crucial services means that staff cannot reduce their responsibilities. Instead, staff put in overtime to meet the needs of residents. Teaching staff how to use AI to be more productive at their job means that staff can go home at 5 p.m.
Finally, many staff will likely use AI with or without their manager’s support. Upskilling your staff early on means that you can teach them how to use AI in a way that aligns with organizational priorities and values without having to break poor habits that have already formed. Additionally, a more positive first impression of AI means staff are more willing to use the AI integrated into organization tools.
Upskilling staff with AI is similar to upskilling staff with email. These skills are essential for workplace success and are here to stay.
Setting your staff up for success
There are several elements that influence the success of any program. The first is manager support. Employees are much more likely to use AI with manager support. According to Irrational Labs, “When managers endorse AI, usage reaches 79%. Without that support, it drops to 34.4%.” Additionally, anecdotal evidence from the City suggests that having both senior leadership support and direct manager support produces the greatest results for effective AI use.
The second is ensuring that instructors have experience teaching and experience using AI the way upskilling participants would use AI. In the City, what we’ve done is partner with the local university, San Jose State University (SJSU). One of our instructors is someone who designs courses professionally, so having her expertise in teaching and designing the course was invaluable. All three instructors have prior teaching experience.
Our Curriculum for AI Upskilling
The goal of this course was to train staff with basic AI skills. We wanted to ensure they could learn how to use new AI tools fairly easily, how to get an LLM to respond to their prompts with the most efficiency, and how they can apply AI tools (not just LLMs) to their tasks to make them easier.
The course focuses on creating custom GPTs, which are reusable versions of ChatGPT trained to be experts at a particular task. This includes adding specific system instructions and knowledge files to the custom GPT. These custom GPTs can be shared with anyone who has a free account, and have proven to have a high impact in time savings and work efficiency.
The curriculum was broken up into three stages: basic AI knowledge and creating custom GPTs, advanced AI use, and student showcases. Each week was one hour of course time and one hour of optional office hours (for staff to drop into with any questions). The office hours were crucial for filling in gaps in knowledge or for any questions staff had.
This pilot program has created a robust foundation for responsible scaling through (a) an established curriculum balancing efficiency with safety, (b) documented best practices for ethical AI adoption, and (c) internal champions who prioritize responsible implementation.
You can see the curriculum below:
The first four weeks are focused on having students create their custom GPTs:
Intro to course and prompting tips
Intro to custom GPT
Effective Prompting Guide
Custom GPT Troubleshooting
The second four weeks are focused on advanced topics:
AI Risks and Responsible AI Use
ChatGPT actions
ChatGPT for Data Analysis
Sharing ChatGPT and additional AI Tools
The last five weeks are focused on the showcases. If there are many students who would like to showcase, we add weeks for them to do so.
Participants created over 60 unique custom GPT assistants. Three assistants have expanded to department-wide use.
Many common use cases staff use AI for include summarizing documents, rewriting documents (for tone, not content), translating short phrases/sentences, and finding information. Some particularly memorable use cases include a grant writing assistant that helped the City secure $12 million in grant funding and an assistant that summarized customer requests for City services.
Teaching the Limitations of AI
LLMs are still relatively new compared to existing tools, and their impact on the workplace has not been fully studied. However, there are budding studies around LLMs, and their results should be considered during the course.
First, as recent blog posts on the Coalition’s Substack have shared, there are strong environmental considerations that should be taken into account when considering the use of AI tools. AI is very thirsty and power hungry, sometimes to the extent of depriving residents sufficient access to water and electricity in drought-prone areas. In the City’s AI Upskilling course, part of the course is devoted to outlining the environmental costs and ways to mitigate these concerns – namely, use effective prompting techniques to reduce the amount of time and effort needed to get the results desired. Additionally, participants learned about AI's limitations to understand what AI can and cannot do well, helping them avoid inefficient use of resources.
Second, AI contains inherent biases. Even explicitly unbiased LLMs can form biased associations. Additionally, LLMs excel at sounding confident while hallucinating, so there are reminders to staff to always think through the sources of data AI is using and to use their critical thinking skills with AI – always check its work and field test AI.
Finally, overuse of AI negatively impacts critical thinking skills, especially when AI is used as the first resort. In a recent study from MIT Media Lab, researchers found that using ChatGPT to write your essays on the first pass reduces critical thinking skills since users are less likely to critically engage with the material. No such effect was seen in users who used Google Search or did not use ChatGPT in the first pass of essay writing and used ChatGPT in the second pass.
These concerns support the view that AI is great for second passes of documents but not first passes. AI can be a thought partner to bounce ideas off of, to get feedback from, and to challenge ideas, but AI is not recommended as a resource of first resort. For example, write the policy argument. Ask AI to critique your argument, but don’t have AI come up with that argument.
Conclusion
Knowing how to use AI safely, effectively, and responsibly is a skill that’s here to stay. Upskilling staff with AI is essential for the government workforce. With managerial support, hands-on demonstration of LLMs, and clear examples of how AI can help staff workload, you too can create an AI Upskilling program for success. Teaching staff the limitations of LLMs helps staff quickly identify which use cases LLMs can help them the most with. It sets clear expectations for what LLMs can do, encouraging staff uptake.
AI expertise, like all skills, is developed with consistent practice and experimentation. Knowing how to use AI and when to use AI will be one of the key factors for successful organizational AI adoption. Human judgement is incredibly important. While AI should not do the bulk of the thinking or the decision-making, it can make the load a little lighter. Residents can get their needs met and staff can go home at 5 p.m.
Because AI is still so new, anyone can become an expert in a relatively short amount of time. In five years, your organization may need to outsource AI expertise if it isn’t developed in-house. This is especially important because governments cannot serve residents well if AI biases proliferate in internal systems without human expertise and oversight. I encourage you to be a steward of responsible AI use in your organization, actively shaping and guiding AI adoption, to ensure that AI implementation is carefully thought through with the human judgement it needs.
About the author
Angela Zhao is a former Privacy and AI Analyst at the City of San José.
Note: The opinions expressed in these articles are solely those of the author(s) and do not necessarily reflect the views, positions, or policies of the GovAI Coalition or the authors’ affiliated professional organization(s).
—
This is the blog for the GovAI Coalition, a coalition of local, state, and federal agencies united in their mission to promote responsible and purposeful AI in the public sector.
