Mira
An AI agent that shops for balance transfer offers, applies for you and moves your credit card debt, while you stay in charge of every decision.

Overview
Mira is Upgrade's AI refinancing agent. It shops top credit card companies for balance transfer offers with no impact on your credit score, works out what you could save and, only if you say yes, applies and submits the balance transfer for you. I designed the end-to-end mobile flow: onboarding, the quiz that gathers what Mira needs, the live search, receiving and accepting an offer, and the balance transfer itself.
The real problem
Refinancing credit card debt takes a lot of legwork: comparing offers across issuers, typing the same details into application after application, then setting up a balance transfer on another bank's site. Many people give up before they save anything.
An agent can do that work, but letting an AI apply for credit on your behalf is a big ask. The design had to make the agent's work visible and keep every decision with the customer.

What good looked like
- Mira does the legwork, the customer makes the decisions
- Show what Mira is doing at every step, live
- Never leave anyone guessing when something goes wrong
How I solved it
Consent up front
The intro explains in three steps what Mira will do, and that it only applies if you choose to continue. Agreeing authorizes Mira as an AI agent, with every agreement one tap away.
A short quiz, mostly prefilled
Mira confirms what Upgrade already knows, then asks only for what issuers need, such as housing payment and income. Connecting card accounts through Spinwheel does not affect the credit score.
Progress you can watch
While Mira works through each issuer, a progress list shows where it is. Watch opens a live view of Mira filling in the issuer's form.
Hand-offs, not dead ends
When an issuer asks for something Mira can't answer, Mira pauses and asks. The customer can take control, skip that card or carry on. If an issuer is down, the screen says nothing was submitted and offers to try again or notify them later.
From offer to transfer
Offers are ranked by estimated savings, with a top pick. After approval, Mira finds the cards to pay off, suggests the amount and moves one or several balances to the new card.
What I designed
Live search across issuers
A step-by-step progress list for each issuer, with a live view on demand.
Ranked offers
Recommendations with estimated savings, transfer fees and APR details, and a top pick.
Recoverable errors
Clear states for stuck, slow and failed steps, each with a next action.
Balance transfer, one card or many
Pick the cards, confirm the amounts, authorize, then follow the transfer from the Mira dashboard.
Key decisions
- Say what was not submitted. Every error state opens by telling the customer whether anything was sent, because that is the first thing they worry about.
- Watching is optional. Most people only need the progress list, and the live view is there for anyone who wants to see the work.
- Ask before every step that can’t be undone: applying, accepting an offer and authorizing a transfer each need an explicit tap.
Outcome
- An end-to-end flow from onboarding to balance transfer, including transfers from several cards at once
- A full set of states for when Mira is stuck, an issuer fails or a step takes a while
- A clickable prototype of the whole journey
What I learned
People trust an agent more when they can see its work and step in at any moment.
Credits
- Role
- Product designer
- Company
- Upgrade, Inc.
- Year
- 2026
- Platform
- Mobile app
- Built with
- Figma
Got a project?
let's talk
Bringing AI into your team's workflow, or designing a product people love? Let's find the real problem first.
Get in touch