Stanislav Dovidenko
UX/UI Design Product Case AI Systems

AI Software Management Ecosystem — Sber UserSoft

Role

Product Designer

Scope of Work

UI/UX for AI Agent, Software Card Redesign, Tone of Voice (ToV), Q&A Hub Design, and Request Flow Redesign.

Context

Process optimization for 300K+ employees during the migration to SberOS and MacOS M-series.

Interaction and key animations showcase of the updated User Soft v1.0

🏆 Results & Impact

Problem & Research

The internal software catalog was perceived as a "black box". Complex forms with an abundance of technical fields caused cognitive overload, and the lack of prompt support forced employees to seek answers from colleagues. When replacing a device, restoring the workspace took up to 2 business days.

Customer Journey Map: Analysis of barriers and user drop-off points

Fig 1. Customer Journey Map: Analysis of barriers and user drop-off points in the legacy system.

* Note: This illustration was generated using Gemini AI to convey the artifact's general structure. The text on the image is a mockup and contains generation artifacts. All actual data and insights are described in the case text below.

Journey Stage User Actions Thoughts & Emotions Barriers (Pain Points)
1. Search & Request Searches for software. Fills out the form, reaches step 3. "What should I write in the justification? Which workstation type do I pick?" Drop-off due to complex form. Confusion over technical fields (distribution, alternatives).
2. Approval (Black Box) Waits for Working Group and InfoSec approval. "Where is my request? Why was it rejected?" Lack of transparency. Unclear request statuses. Rejections due to bureaucracy.
3. Installation & Tracking Attempts to install or update the software. "Where is the install button? Another error..." Update drop-off. Fear of breaking the environment due to complex algorithms.
4. Access & Support Encounters an error, looks for a solution. "I'll message Alex, IT support takes too long." Shadow IT support. IT service overload, seeking help in personal chats.

Quotes from the Field: What happened at the "dreaded Step 3" of the form

Project ManagerField: Justification

"I don't understand what to write in 'Business Justification'. InfoSec constantly rejects my request because they don't like the wording. Just give me an example!"

System AnalystBank Alternative

"How am I supposed to know if there is a 'Bank Alternative'? I just want to install the program for work. Why can't the system check this itself against the database?"

Frontend DeveloperWorkstation Types

"I'm stuck on selecting 'Virtual/Standard Workstation'. I have no idea if I need a virtual layer. If I pick the wrong one, will my request be rejected a week later?"

QA EngineerDistribution Link

"They rejected the publication because of the distribution link. I provided a link to our internal drive, but apparently, they need the official website. That wasn't written anywhere!"

Shipped Solutions

1. Software Card Redesign & Information Architecture

I reworked the card's composition, focusing on modularity and data density. My goal was to group technical parameters and statuses so users could instantly find what they needed.

Software Card Redesign: Mockup focusing on technical specs and compatibility status

Fig 2. Modular software card design: reducing cognitive load on the user.

2. Q&A Community Hub (Discussions)

Introduced a public Q&A space within the software card. This created a self-learning knowledge base where answers from software owners are accessible to everyone, radically reducing the volume of routine support tickets.

Discussions interface in the software card: user questions and official expert answers

Fig 3. Q&A Section: crowdsourcing knowledge and reducing support load.

3. AI Agent & Tone of Voice

Designed the Conversational UX for a GigaChat-based assistant. I developed the Tone of Voice (system persona) and confirmation logic so that automated software installation felt like a reliable, user-controlled process.

Agent conversation flow: from installation request to success confirmation

Fig 4. AI Agent: a human-centric approach to complex automation.

4. Request Redesign (Bureaucracy as a Service)

I redesigned the interfaces for requesting and publishing software. Forms became cleaner, with smart validation for workstation types and contextual hints helping users avoid errors at the "dreaded step 3."

Example of an updated software request form with logical hints and alternative checks

Fig 5. Form optimization: transforming a complex process into a clear sequence of steps.

5. "Magic" Restore (Profile Migration)

UX support for the one-click software restore feature. Thanks to backend integration (Persona system), the interface automatically prompts users to restore their familiar software stack on a new device.

Screen showing the list of software for migration and restoration progress indicator

Fig 6. Seamless migration: a magical experience for workspace restoration.

Why It Works (Design Logic)

Hierarchy

The new software card architecture eliminated paralysis caused by data overload.

Reliability

The Agent's Tone of Voice and transparent request statuses built trust in the system.

Automation

Shifting from 'informing' to 'executing' (via the AI agent and migration tool) saved thousands of human hours.

Reflection & The Revolut Mindset

In this project, my primary challenge was to tame backend complexity. I acted as a bridge between data architecture and the user who just needs simplicity. This experience of building scalable, efficient interfaces perfectly resonates with the Revolut culture, which values velocity and the radical automation of routine tasks.