AI Agent

P R O J E C T T Y P E

SaaS

D U R A T I O N

2 months

M Y R O L E

UI & UX Designer

T O O L S

Figma, Teams, Jira, Slack

About Project

The SaaS application project for analysing sales offer alerts was created to improve the analysis and resolution of internal issues within the company and its representatives. My task was to create a functional interface that responded to the real needs of the company, according to business guidelines and the technical limitations of AI Agents.

My role

In this project, I played a comprehensive role as a UX/UI designer, responsible for the entire design process – from requirements analysis to the final digital product. My tasks included research, conceptual and visual aspects, as well as coordination:

1. Gathering business and technical requirements

I participated in discussions with stakeholders to understand the project's objectives, end-user expectations, and technical and product limitations.

2. Development of information architecture

I designed a logical application structure, taking into account user paths, content hierarchy, and flows between screens.

3. Creating personas, user journeys, and usage scenarios

I developed models of user behaviour and needs, which formed the basis for further design work.

4. Prototyping and wireframing

I prepared low-fi and high-fidelity mock-ups, iterating them based on team feedback and preliminary usability tests.

5. UI design and styling system

I created the visual direction of the project, selecting typography, colours, iconography and components. I also built a consistent mini design system tailored to the specifics of the product.

6. Cooperation with the development team

I consulted solutions on an ongoing basis, verified implementations and ensured that implementations were consistent with the project.

7. Preliminary usability tests

I conducted test sessions and gathered feedback from users and the product team, which allowed us to quickly identify potential errors or ambiguities.

8. Consulting and strategic support

I supported the product team in making design decisions, also advising on the scalability of the solution in further stages of development.

Process of designing

I follow a comprehensive end-to-end workflow to ensure product success. By balancing user needs with technical feasibility, I transform complex business requirements into functional, high-fidelity digital solutions ready for development.

P A R T 1

Work organisation

We began the project by gathering information about the alert management process. As part of each sprint, I conducted regular design review sessions attended by the Product Owner and two representative end users. During the meetings, I presented the current designs and discussed every aspect of the interface, encouraging discussion about the functionality and intuitiveness of the solutions.

P A R T 2

Wireframing

At the same time, I conducted interviews with users to better understand their needs, expectations, and problems in their daily work. I used the information gathered to make decisions about functional priorities together with the team. I then translated the findings into a UI design, ensuring that the interface supported real-life usage scenarios.

P A R T 3

Implementation supervision

After preparing the wireframes, I organised consultations with the development team to verify the compatibility of the proposed features with the technical capabilities. This made it possible to detect potential limitations early on and optimise solutions before the implementation stage.

This cycle of iterative testing, consultation, and validation ensured that the final product not only met user needs but was also technically feasible and business-effective.

Solution

S T A T U S E S

Visualizing the Agent’s Reasoning

I designed a system of dynamic statuses that map out the agent’s order analysis process. By exposing the AI's "thought process," I eliminated the "black box" effect, building user trust and providing clear visibility into the system's current state.

F E E D B A C K

Human-in-the-Loop = Expert-Led Training Interface

I implemented a feedback loop for domain experts to verify and refine the agent’s accuracy. Through a simple rating system (thumbs up/down) and a correction module, specialists can negate errors and provide the ground truth to continuously train the model.

E M A I L C O M M U N I C A T I O N

Contextual UI & Seamless Workflow

To meet users in their existing workflow, I integrated the agent’s output directly into their inbox. I designed interactive emails with built-in control buttons (Approve/Reject), allowing users to manage AI results without switching platforms.

Visite my Behanace

This and other projects are also published on Bahence.com. This is the place where I publish my work on a regular basis. I encourage you to take a look and leave a comment.

Go to Behance

Page updated

Google Sites

Report abuse