Users don't know what the agent can do, so they type "hi" and leave.
We design the first screen: starter prompts based on real tasks, example requests, and limits stated in plain words. A visitor understands within seconds what to bring to it.
AI agent interfaces people actually use and trust
We design the interface for agents (AI that takes actions on its own) and copilots (AI that works beside a person). Users see what the agent is doing, what happens when it fails, and how to reach a human.
Quote and schedule after your brief
Who it's for
- SaaS and service companies adding an AI chat or agent to their product or customer portal
- Businesses that launched an AI bot that few people use or trust
- Startups that need a convincing AI product prototype for investors or first customers
- Teams whose internal copilot for sales or support sits unused
Problems we solve
The agent works in the background and nobody knows what it did. People hesitate to click "approve".
We map the agent's states: thinking, calling a tool, waiting for approval, done, failed. Irreversible actions get a confirmation step with a short summary and an undo.
The model makes things up, and customers notice.
We add sources to answers, a visible uncertainty signal and a fast way to report a bad reply. The error is made visible and fixable rather than hidden.
When the bot can't cope, the customer gets stuck and leaves.
We design the handoff to a human: what the operator sees, how history carries over, what the customer reads at that moment. This ties the agent into your CRM and workflows.
What's included
- User task and scenario review: what the agent does alone and what it only suggests
- Agent state map: waiting, streaming (text arriving in parts), error, partial answer, refusal
- Chat screen design plus non-chat patterns: side panels, inline actions inside your existing UI
- Confirmation rules for actions involving money, data or messages to customers
- Human handoff and operator notification screens
- Interface copy: prompts, error messages, refusal wording
- Clickable Figma prototype; a working prototype with real model calls where needed
- Testing with 5–8 users and a prioritized fix list
- Figma component library with state and behavior notes for developers
- Handover of files and access, all in your accounts
Choose the scope
Use a package to describe what you need. We agree on the quote and schedule after reviewing your brief and requirements.
01
Audit
You already have an AI chat or agent, but usage is low or users don't trust it
- Review of real conversations (you supply logs without personal data)
- Prioritized problem list
- Rewritten key copy and starter prompts
- Mockups of 2–3 fixed screens
02
Agent Interface
You are launching a first agent or copilot and need a prototype to show customers and developers
- Scenarios and agent state map
- Figma mockups of core screens
- Clickable prototype
- Test with 5 users
- Interface copy in English or Ukrainian
03
AI Product System
Your product has several agents or a copilot across multiple areas and needs consistent logic
- Everything in Agent Interface
- Component and state library in Figma
- Rules for confirmations, permissions and human handoff
- Working prototype with real model calls
- Developer support during implementation
How we work
Brief and requirements
You describe the task and current workflow. We clarify requirements, choose a first stage and prepare a quote.
Scenarios and agent limits
We go through your real conversations or tickets. We write down what the agent does alone, what it only suggests, and what always goes to a person.
Mockups and states
We design the screens and every agent state, including errors and refusals. Copy is written at the same time, not after the visuals.
Prototype and testing
We build a clickable prototype and walk 5–8 users through it. We note where they stall and fix it.
Handover
You get Figma files, components and behavior notes. All files and access sit in your accounts under the agreed terms. Payment is in stages.
Scope and limits
- We don't train the model or build the agent backend. That is a separate service; here we design what the user sees.
- We don't promise the AI will never be wrong. The interface makes mistakes visible and recoverable, but answer accuracy depends on the model and your data.
- We don't guarantee sales growth or specific numbers. We offer a pilot with a measurable target agreed in advance and an honest report on the result.
FAQ
How is UX for AI different from regular UX?
AI replies are unpredictable, slow at times and sometimes wrong. Besides screens, we design waiting, errors, action confirmations and handoff to a human. Most of that doesn't exist in a conventional interface.
What should I prepare before we start?
A description of the agent's job, examples of real conversations or tickets without personal data, a list of actions the agent can take, and API docs if you have them. If you have none of this, we start with the project brief and build the scenarios together.
Do you work with any model?
The interface isn't tied to one model. For working prototypes we use common model provider APIs. You can swap the model without redesigning screens as long as it supports what you rely on, such as streaming responses.
Do you build the interface in code?
The design packages end with mockups and components for your team. A working prototype with real model calls is included in the top package. Full product development can be ordered separately from our studio.
How do you check that the interface works?
We run the prototype with 5–8 people from your target audience and watch where they get lost. This is qualitative testing, not a statistical study. A measurable result needs a pilot with real users.
How do you estimate the cost?
We prepare a quote after reviewing your brief and requirements. A first stage can cover: review of real conversations (you supply logs without personal data); Prioritized problem list. Scope, payment and schedule are agreed before work starts.
Start with your task
Describe your business, the outcome you need and your current workflow. We will clarify the requirements and prepare a quote after your brief.