How Kansas City’s AI agents are transforming people management

“I asked Hannah, and she told me I need to talk to Daryl,” LaToya Black, Talent Acquisition Manager at Kansas City, recalls hearing from team members at City Hall discussing an employment issue.

Nothing unusual there except Hannah and Daryl are not humans.

They are artificial intelligence (AI) agents and Black is enthused by how well employees have adapted to their use.

Hannah is described as “the helper”, while Daryl assists managers to prepare supervisory documentation. Carla, another AI assistant, provides employee coaching guidance.

“We kind of made it a family,” Black says. “That way, it’s easier for employees to identify with.”

Support around the clock

Using AI agents is a response to the practical challenge that the city operates 24 hours a day. Managers work across different shifts and may face workforce issues when HR professionals are unavailable.

Kansas City also has four generations in its workforce, creating a range of management and workforce development needs.

“Some of them have never had to do interviews, never document interviews, never had to look at a candidate pool and figure out who’s the best one, not just the person they like,” Black says.

Black began exploring how AI could give managers access to recruitment and workforce guidance when her team was unavailable. At around the same time, Dominic Schuman, Senior Human Resources Specialist at Kansas City, was considering how agents could reinforce the city’s leadership training.

The work on AI agents followed a request from the city manager’s office to explore how AI could make municipal operations more effective and efficient. Andrew Ngui, Chief Digital Officer at the city, showed Schuman how to build an agent. Schuman then sought further advice from a colleague in the city’s IT department.

The experimentation led Schuman to question whether conventional training was enough. The city had invested considerable time in teaching managers about coaching, counselling, supervisory documentation and leadership development, but employees could forget much of the material if they did not apply it.

“What we can’t control is what people do with the information, but what we can control is the information that we give them and the resources that we give them to support that information,” Schuman says.

Rather than expecting employees to retain everything taught in a leadership course, Kansas City is placing guidance within their daily workflow. Managers learn the underlying skills through formal training and can then consult an AI agent when they encounter a real workplace situation, including outside normal HR operating hours.

The aim is to close the gap between knowing what to do and putting that knowledge into practice. The city describes this as behavioural change by default, with AI supporting employees at the point when a decision or action is required.

From learning to action

The agents supplement training rather than replacing it. Schuman compares the approach to teaching someone long division in mathematics before then showing them how to use a calculator.

Managers first learn the relevant principles and skills in a leadership development course. Carla can then propose an approach, meeting agenda and responses to possible employee reactions. Kim, the counselling agent, provides similar support, while Daryl prompts managers for discussion points and formulates a note for the supervisory file.

The agents can also help managers determine what kind of intervention is required. Someone may be unsure whether a situation calls for coaching, counselling or referral to the city’s equal employment opportunity or labour and employee relations specialists.

“The idea here is that we are supporting that learning curve long term,” Schuman says. “Here’s a resource that you can use to make sure, even if you forgot, that you can in real time recall that information and be walked through it as you need it.”

This point-of-need support is central to Kansas City’s approach to behavioural change. Instead of relying on a course by itself, the city gives managers a resource to consult when they need to make a decision.

Ngui describes the objective as reducing the “intention-action gap” between theory and practice.

“Successful and sustainable organisational change really starts with one person,” he says.

Building the agent family

Black has also developed Alex, an application assistant, and Iris, an interview guidance and compliance agent. Iris reviews proposed questions and flags those that could breach equal employment opportunity or anti-discrimination laws.

A manager might, for example, propose asking whether an applicant has children because the position requires them to work a particular shift.

“Iris would say absolutely not,” Black says. “It would say: ‘What is it that you want to get to? What is the task that you want to complete?’”

Once the business requirement is established, the agent suggests an appropriate alternative.

“The hours of this position are Tuesday through Saturday from 7pm to 3am. Can you work those hours?” Black says. “That is the business-related question.”

Managers remain responsible for the final decision. They can upload a job description and ask Iris to suggest interview questions, but they decide which questions to use and whether to amend them. The agent provides guidance without removing managerial accountability.

The family has expanded beyond management and recruitment. Harry, the general HR helper, directs employees to the correct resource, while Benji, the benefits agent, provides information to support benefits selection.

The city has also introduced Sam, a standard operating procedure writer, and Janice, a job aid creator. A compensation agent is in development.

The agents form part of Pathways Plus, an extension of the city’s leadership development programme. Kansas City began with three agents, expanded to five and then incorporated Black’s work, bringing the number to eight. The ecosystem has continued to grow.

Policies and guardrails

The city began developing the agents around what Schuman calls “prevention as protection”. Each agent is given the city policies, procedures, administrative directives, laws, training materials and recommended practices relevant to its role.

Schuman describes the foundational instructions given to Carla:

“Never go outside of the manual. Never make anything up. If you do not know the answer, explicitly say that you do not know the answer and redirect the end user to a human resource to be able to field that question.”

The city documents any changes to the instructions underpinning each agent and provides a SharePoint channel for reporting responses that appear inconsistent with policy or procedure.

HR staff and frequent users tested the agents, including through negative testing intended to reveal how they responded to inappropriate questions or language.

An audit process examined privacy, reliability, security, bias, fairness and governance. It also reinforced the need to update affected agents when a policy or procedure changes.

Privacy remains an ongoing concern. Managers are instructed not to enter personally identifiable information when using the tools, and guidance on safe use is included in agent workshops and leadership development courses.

Black’s advice to other cities is direct:

“Always consult. Always keep law in the process along with you,” she says. “You don’t want to create something huge and you just love it and then it gets nixed because maybe you didn’t do something along the way.”

Encouraging adoption

Kansas City developed the agents using the Microsoft 365 Copilot capability included in its existing licence. Ngui says obtaining the paid agent version could widen access and expand the initiative.

The city does not yet have the licence needed to analyse usage in detail, but Black says feedback from managers has been positive. Workshops combine demonstrations with hands-on experience, with managers bringing their laptops and trying the agents themselves.

The city is not requiring employees to use the agents. It is making them available as an additional resource while building confidence among staff who may be apprehensive about or unfamiliar with AI. For Ngui, leaders cannot simply hand AI tools to employees and expect adoption. They need enough practical understanding to guide their staff.

Leadership teams have attended workshops on using Sam and Janice, while employees are beginning to approach Schuman’s team after hearing that an agent exists for a particular task.

Black says curiosity and experimentation have been critical to the work.

“There has to be a level of inquisition,” she says. “How do we make more with less? What is a way that we can do this?”

The city is also considering how to help more employees develop ideas safely. Schuman framed this as a combination of education and organisational safeguards.

“We are going to teach you, and we’re going to build some guardrails, and we’re going to make sure that you know the best practices and build in some safe stops as well,” he says.

A systemic approach

Employees have already suggested agents for internal teams and services for residents. Black says useful knowledge and ideas may come from staff whose formal roles are unrelated to AI.

Kansas City expects it may eventually establish a group or committee to review proposals for new agents.

“Gain the input from the organisation on whatever it is you want to create your agent on,” Black says. “Find out what’s needed.”

The approach treats AI deployment as part of workforce and leadership development, rather than as a standalone technology project. Employees learn management skills, use agents to apply them and contribute ideas for further support.

A manager can seek help from Hannah, use Carla to prepare for an employee coaching conversation and turn to Daryl to produce the supervisory documentation afterwards. As more needs are identified, the city can add specialists while retaining a familiar way for employees to navigate the support available.

Black says of the AI agents: “They have their own little personalities and their own little looks. It’s easier for the family to grow versus just having this team of agents. Our family is growing.”

Main image: Far Chinberdiev on Unsplash