Business owners asked their bank for AI help, so I designed an assistant that tells them in plain words what needs attention today.

This is a concept, not a shipped product. In 2024 I led design on a future-facing mobile banking prototype for a global bank's small business customers, and this concept adds an AI assistant to that prototype, built on my research into how owners get paid and protect their money from fraud.

  • Client

    A global bank

    I changed names and details to respect my nondisclosure agreements.

  • My role

    Product research lead and mobile design lead

  • Focus

    AI help for business owners in mobile banking

  • Result

    The 2024 prototype shaped multi‑year roadmaps

Where it pinched

Small business owners run their own banking, invoicing, and fraud checks in the minutes between everything else.

In 2023 and 2024 I led product research for small business payments at a global bank. Two products filled most of my days. The first was a hub where business customers get paid by card, invoice, and ACH transfer. The second was a payment protection service that holds checks and ACH debits the owner has not approved until the owner decides.

Many of the customers I studied had fewer than 10 employees and less than $1 million in yearly revenue. They kept their businesses running and also kept up with payment technology that changed faster than they could. When I asked what would make their day-to-day banking easier, many asked for AI that could help them manage their business banking. In the same conversations, they asked for a real person whenever money was moving.

Fraud raised the stakes. Owners told me their worry was about money more than technology, because a payment that goes out the door is hard to get back. Any assistant I designed had to earn that trust before it could save anyone time.

What I built first

The 2024 mobile prototype gave the assistant a home.

After the get-paid hub launched to a strong response, I led design for a future-facing mobile business banking prototype with partners from marketing, engineering, product, and banking. We started with whiteboard brainstorms drawn from what we learned before and after the launch, then worked through several versions of wireframes until the direction was clear. Click or tap a card to turn it over.

What owners kept telling me

Owners wanted help that speaks their language and leaves the decisions with them.

The assistant concept grew straight out of my research notes. I did not start from a model and look for a use. I started from the jobs that already frustrated owners and asked where an AI could take work off their plate without taking control away from them.

Three jobs came up again and again: knowing whether cash will cover the bills, getting clients to pay, and deciding fast on a payment that looks wrong. Those three jobs became the assistant's whole scope.

  • Owners asked for AI to help them manage their business banking.
  • Owners wanted everyday words instead of banking terms.
  • Many owners wanted to talk with a person before anything involving sending or receiving money.
  • Owners wanted to see where each incoming payment came from, line by line, like a ledger.
  • Owners often discovered fraud because a payment broke their usual pattern.
  • Relationship managers said customers wanted phone alerts for urgent decisions, such as a blocked payment.
  • Some owners wanted to set things up once and stop thinking about them, while others wanted to review every payment themselves.
How I kept owners in charge

Six rules kept the assistant useful and honest.

Every screen in the concept follows the same rules, so an owner can predict what the assistant will and will not do. Click or tap a card to turn it over.

Try it yourself

Spend a morning at Your Co, LLC with the assistant.

I rebuilt my 2024 mobile screens as a working prototype under the sample name Smith Bank and added the assistant concept on top, with AI writing the code from my designs. Your Co, LLC and every client, payee, and amount are sample data, and the assistant's answers are written ahead of time.

Things to try

  1. Open the flagged check in the briefing, then pay or return the check.
  2. Open the rent item to see the forecast and how sure the assistant is.
  3. Draft a reminder to the late client and send the reminder.
  4. Tap the chat icon and ask the assistant a question.
  5. Open More and change how much the assistant brings to you.
The kit behind it

One confident blue, serif numbers, and fine rules keep the money front and center.

I restyled the prototype as a business bank app with the calm of a fashion magazine: a Smith Bank wordmark instead of a logo, large serif headings and balances, small spaced capitals for labels, and thin rules in place of boxes. Blue is saved for actions and for everything the assistant says, and every text color passes current contrast standards.

Color

  • Smith blue#005EB8Primary buttons, links, and chart lines
  • Deep blue#004A93Button hover states and the assistant's labels
  • Ink#14212EHeadings, balances, rules, the wordmark, and the selected tab
  • Slate#5A6878Labels, secondary text, and inactive tabs
  • Mist#F3F5F8Hover states behind secondary buttons
  • Assistant tint#EAF3FDThe background of everything the assistant writes
  • Alert red#C0283DOverdue money, flagged payments, and shortfalls
  • Done green#127A3EConfirmed decisions and sent reminders

Type

  • Wordmark

    Smith Bank

    Playfair Display

  • Greeting and screen titles

    Good morning

    Playfair Display

  • Balances and amounts

    $12,345.67

    Playfair Display

  • Labels

    Decide by 4:00 PM

    Open Sans

  • Body

    Rent clears if your two unpaid invoices arrive.

    Open Sans

Components

  • Daily briefing

    Up to three numbered items, sorted by deadline and by dollars at stake, answer the question of what needs the owner today.

  • Why these disclosure

    A short expandable note says what the assistant read and repeats that the assistant never moves money.

  • Forecast with confidence

    Two lines show the balance if invoices arrive and if they do not, with a plain note on how sure the forecast is and what it leaves out.

  • Flag review

    A flagged payment lists its reasons in plain words, says that a pattern is not proof, and offers Pay, Return, or a person.

  • Confirm before money moves

    Every decision opens a dialog that repeats the payee, the amount, and what happens next before anything changes.

  • Linked chat answers

    Each chat answer links to the screen behind the answer and always offers a way to talk to a person.

Principles

  • Put the amount and the deadline in the first line.
  • Let the assistant suggest and leave every money decision with the owner.
  • Say how sure the assistant is and what the assistant cannot see.
  • Keep a real person one tap away from every decision.
What it led to

18–74%

lift in payment product enrollment within one to two quarters of the get-paid hub launch

What changed for people

The research behind this concept changed a shipped product, and the 2024 prototype shaped long-range plans.

The assistant is concept work, so it has no launch numbers. The work it grows from does. Click or tap a card to turn it over.

Where this goes

I bring the same rules to teams putting AI into money and safety tools.

As a principal designer offering AI-native product research and design, I help teams start from the tasks that already trip people up, decide what an assistant may and may not do, and test the trust moments before launch. If your team is adding AI to a banking, payments, or fraud flow, I would be glad to review that flow with you.

Still curious?

There is plenty more where this came from.

Names, brands, and data in my case studies are changed to protect my clients.

Connect

Tell me about your project.

Tell me a little about it on my contact form, and a real person will get back to you soon.