Artificial intelligence could soon play a bigger role in how Harris County government operates, potentially speeding up services, reducing paperwork and giving employees more time to help residents with complicated cases.
But as the technology advances, experts say local governments must be careful about how they spend taxpayer money, protect personal information and make sure people can still reach a human when something goes wrong.
Harris County commissioners were expected to discuss an updated artificial intelligence policy Thursday, Nov. 12, but the item was pulled from the agenda. It is unclear when commissioners will revisit the proposal.
The county already has an AI policy approved in November 2024. Just two years later, officials are considering a more comprehensive set of rules as the technology continues to evolve.
Dr. Dorothy Jiang, an associate professor of information systems and co-director of the Human-Centered AI Institute at the University of Houston’s Bauer College of Business, says AI has the potential to make government more efficient.
“I think AI is everywhere and a lot of government work is paperwork, routine work,” Jiang said. “AI can definitely help with that type of work, reviewing forms, answering those same questions over and over, checking documents. AI can do much of that heavy lifting. So county employees can spend more time even on the cases that really need a person.”
That could mean faster responses, shorter wait times and more assistance in different languages for residents across Harris County.
However, Jiang cautions that those benefits come with risks.
Harris County first introduced generative AI guidelines in October 2023 before approving a countywide responsible-use policy on Nov. 12, 2024.
That policy established expectations for employees, contractors and others using AI on behalf of the county. It addressed responsible use, security, privacy, human oversight and the purchase and operation of AI systems.
In January 2026, commissioners directed the county’s Universal Services Department to develop a more comprehensive policy addressing emerging requirements and the growing use of artificial intelligence across county departments.
The proposed update would replace the 2024 policy and strengthen countywide standards for cybersecurity, data privacy, responsible AI use and oversight.
County officials have recommended May 2027 as an implementation target to give departments time for employee training and guidance.
The proposal has not been approved.
RaSheem Barnett, CEO of Emagine Reality, a consulting firm that helps organizations improve operations through AI and automation, says the technology is changing faster than many organizations can keep up with.
“AI has changed significantly in those last two years,” Barnett said. “In fact, the last six months, every day, there’s new advancements within AI, nothing like anything we’ve ever seen with the technology over our lifetimes.”
He says governments must find a balance between adopting useful technology and understanding the risks that come with it.
One of the biggest concerns is what happens when a government employee enters a resident’s personal information into an artificial intelligence program.
Jiang identified three major risks: privacy, bias and inaccurate information.
“Someone could put residents’ personal information into a tool that the county doesn’t even control. The county hasn’t even approved the use of that tool,” Jiang said.
She also warned that AI systems can repeat biases found in the data used to train them and produce incorrect answers that sound convincing.
To reduce those risks, Jiang recommends several safeguards.
“In my view, one of the biggest risks is … a busy employee who just pastes someone’s information into a free AI tool to save time,” Jiang said. “So the rule should be simple. County data only goes into county-approved tools. And each tool should only see the information it needs to do its job, nothing more.”
She says county governments should also establish contracts that prevent outside technology companies from using residents’ information to train their own AI models.
“When the county buys an AI tool, the contract should say the data belongs to the county. The vendor cannot use it to train their own products, their own AI models, and the data is deleted when the contract ends,” Jiang said.
She also recommends independent audits to make sure those safeguards are actually being followed.
“The county shouldn’t just grade its own homework,” Jiang said.
Barnett agrees that protecting residents’ information must be a priority.
“We also need to limit the information collected, right? Use only what’s needed for the task. If a system doesn’t need a Social Security number or other sensitive data, don’t enter it,” Barnett said.
He says access to that information should be limited to authorized employees, and contracts should clearly explain where the information goes, who can see it and whether it can be used to train AI systems.
These are recommendations from the experts, not necessarily requirements contained in the county’s proposed policy.
Implementing artificial intelligence can involve purchasing software, developing systems, maintaining technology, protecting data and training employees.
In March 2026, Harris County commissioners approved $425,000 in supplemental funding for contractor resources supporting AI compliance, governance and implementation.
That amount does not represent the county’s total spending on artificial intelligence.
Barnett says governments should evaluate the full cost of a system before deciding whether it will actually save taxpayer money.
“At the end of the day, the government also has to run their business like a business, right? And so we need to make sure that we’re making the right financial choices,” Barnett said.
He explained that simply making a task faster does not automatically mean an organization is saving money. Governments also need to determine how employees will use the time they save and whether the technology provides a measurable benefit.
Barnett cautions against purchasing expensive AI systems without first identifying the problem they are supposed to solve.
“AI is not a magic wand, a magic stick as it’s been promoted,” Barnett said. “What do you need your people to do? What is the process in which you need them to execute that? And then and only then find the right technology to either buy or build that enforces the process that you’ve created.”
He says organizations can waste money when they purchase technology first and then try to change their operations to fit the software.
Both experts emphasized that artificial intelligence should help government employees do their jobs, not replace human accountability.
“If a county department uses an AI tool, the department is responsible for the outcome. The same as if an employee had made the decision,” Jiang said.
She believes residents should be told when AI is involved in decisions that affect them and should have a straightforward way to challenge incorrect information.
“It should be very simple because many of us, including me, have been stuck in a chat box loop typing on, ‘I want to talk to a person’ again, again, again. But we’re not even getting anywhere,” Jiang said.
She also warned that employees can become too trusting of AI-generated information, particularly when systems appear confident in their recommendations.
Barnett says governments need to make sure someone has the authority to review and override those recommendations.
“AI can inform a decision, but a person should be responsible for consequential decisions and able to answer or override a system’s recommendation,” Barnett said.
He believes residents should always have a clear path to a county employee who can help resolve a problem.
“At the end of the day, AI should be another doorway into government, not a locked door between residents and the people there to help them,” Barnett said.
While commissioners consider updating the county’s AI policy, experts encourage residents to stay informed and ask questions about how the technology will be used.
Residents can attend Commissioners Court meetings, participate in public comment opportunities and ask county leaders which departments are using AI, how personal information is protected and who is responsible when mistakes happen.
They can also advocate for clear ways to contact a county employee, report errors and request human review of decisions involving AI.
Jiang says transparency and accountability will be essential as governments continue adopting the technology.
“My advice for residents just is to stay informed and ask questions,” Jiang said. “And then my advice for the county is to be open and keep people in the driver’s seat and earn trust step by step.”
She says the ultimate goal should not simply be making government faster.
“AI should make government work better for people, just not just faster.”
Residents can find upcoming meetings, agendas and public comment information through the Harris County Commissioners Court website.
For now, the county’s existing AI policy remains in place while officials determine when to bring the proposed update back for consideration.
