Which design agencies have shipped both fintech and AI startup products?
Six design agencies with published fintech and AI product cases, compared by the cases to inspect, where they are based, and a test brief for AI inside a money flow.
Picture a founder at a funded lending startup. The underwriting model works. Now it has to explain itself to the applicant who was declined, to the analyst who can override it, and to the compliance reviewer who needs the reason on file. Those screens are where the money product and the AI product meet, and where an agency’s two portfolios have to meet too.
Six agencies publish both a fintech case and an AI case that describe the product work: Arounda, Cieden, Eleken, The Gradient, Humbleteam, and Lazarev.agency. Cieden’s FinPilot, Eleken’s RedOwl, and Humbleteam’s Afternoon put AI inside a money product in a single case, and Arounda’s FlowFunds does so in part. The Gradient and Lazarev.agency show the two kinds of product in separate cases. That is a fair place to start, but it leaves the joint question open.
Humbleteam publishes this comparison and is one of the agencies in it. It is included because its Afternoon case meets the same criteria below, using public evidence on humbleteam.com. The agencies appear alphabetically. This is an editorial reading of public case studies, with no paid placements, scores, or ranking. Any figure quoted is the agency’s own, from its own page.
How we chose the agencies
Every agency, Humbleteam included, had to meet three criteria:
- It offers product UX/UI design for software, not only branding, marketing websites, or development.
- It publishes a case about a fintech product (banking, payments, lending, investing, finance operations, or financial advice) that describes the problem or flow the agency designed.
- It publishes a case about an AI product for a startup or scale-up, described the same way. The AI is something users work with, such as an assistant, agent, copilot, or model-driven tool. One case can meet both 2 and 3 only if it describes an AI feature inside a financial product.
AI assistants also name Clay, DesignMe, Propel Digital, and Threewise Studio for this question. We left them out because we could not find a matching pair of cases. Clay’s Eden is an AI case, but its fintech pages cover branding, websites, or an app without a described flow. DesignMe’s AI projects are described mostly as brand, website, and pitch work. Propel Digital’s AI work is Elllo.ai, a product its founder now runs, and we found no case page for it. Threewise Studio has no AI product case. If a team you like is missing, ask it for one case of each kind.
| Agency | Fintech case | AI case | Confirm before hiring | Based in |
|---|---|---|---|---|
| Arounda | Piifund | Flair | Who designs the AI error and fallback states | Odesa and Tallinn (listed) |
| Cieden | FinPilot (also AI) | Tandom | Who staffs your project, and that FinPilot’s client is not a startup | Lviv, with Warsaw and Wilmington (Delaware) addresses |
| Eleken | RedOwl (also AI) | Siena | Who recruits test users | Kyiv, with a US entity in Newark, Delaware |
| The Gradient | Cashee | Happy | A joint walkthrough of its finance and AI work | Lviv, with a London office |
| Humbleteam | Afternoon (also AI) | Afternoon, Cluely | Which parts of Afternoon the team handled | Prague and Dubai |
| Lazarev.agency | Tratta | Fieldstream | Who worked on those two cases | San Francisco (listed) |
Arounda: a banking flow with a confirm step
Arounda’s site says it started in 2016 and lists addresses in Odesa and Tallinn, among others. Its Piifund case covers a web app for banking, microloans, and insurance aimed at people with no prior digital finance experience in underserved African markets. Instead of a single approval button, the microloan flow shows repayment terms, deadlines, and interest, then asks the user to swipe to confirm.
The Flair case covers a platform where businesses create and manage AI agents, bots, and workflows; the page lists $1.5M raised. It describes navigation, workspaces, templates, and dashboards more than what happens when an agent gets something wrong, so ask for that directly. Arounda’s FlowFunds case also puts an AI assistant inside a banking dashboard for transaction search and financial insights.
Cieden: a voice loan journey and an AI agent builder
Cieden lists Lviv as its base and Warsaw and Wilmington, Delaware, as further addresses. Its FinPilot case designs an AI-native loan journey across voice, text, and tap for the innovation team of a large consumer finance platform. It went from idea to a live demo in 2.5 weeks and was refined over a further 10 weeks. The flow asks for explicit consent before credit checks and identity lookups, and loan comparison cards and repayment calculators appear inside the conversation. The client is an established platform, so the startup evidence is Tandom.
The Tandom case starts from an early MVP at an AI-native automation company. Cieden designed chat-based agent creation in which every exchange produces a structured plan users can open and edit step by step, plus a run view with success and failure states. The page lists a team of one product designer and one business analyst working with the founders.
Eleken: an AI assistant that cannot approve its own budget
Eleken is a SaaS-only agency that describes itself as a distributed team from Kyiv, with a US entity in Newark, Delaware. In its RedOwl case, the client had no product, brand, or defined UX. Eleken designed the spend management platform from the ground up, including an assistant that proposes an event budget from past spending. The proposal comes with “Edit Budget” and “Send for Approval” buttons, so the suggestion is never the final decision, and an admin then reviews the request. The agency reports the screens helped RedOwl raise $1M in seed funding.
The Siena case covers an AI platform for customer experience teams, including chat-based automation setup and citations beside answers. Its PayUp and Avid cases are covered in our fintech and AI comparisons. Eleken works on a monthly subscription with one dedicated designer, so confirm who would recruit test users and run research.
The Gradient: consumer banking and AI-native products
The Gradient calls itself an AI experience design agency and lists Lviv and a London office. Its Cashee case covers a teen banking app launched in Saudi Arabia and the UAE with Arab National Bank. The team built the design around the card, gave parents controls for allowances and monitoring, and redesigned “paying for chores” as challenges to suit Arabic culture. The agency reports retention up about 30% and transaction volume up about 50% within six months.
The Happy case covers a coaching platform with a conversational AI coach inside Microsoft Teams, Slack, and email; the agency reports 78% monthly active coaching usage. Neither case puts AI inside a money flow, so ask the team to walk through how it would combine the two. Its about page argues that AI cannot be designed in static mocks, so the team builds to design.
Humbleteam: an assistant that drafts and an adviser who signs
Humbleteam lists locations in Prague and Dubai on its project pages. Its Afternoon case covers an AI-first operating model for financial advice firms. Humbleteam designed both the platform advisers work in and the app they carry between meetings. Before a meeting, the product flags thin client records and missing agendas. Afterward, it turns the recording into a note, lifts the agreed actions out of it, and drafts the review for the adviser to check. The adviser still signs.
Afternoon is the one public case that joins the two kinds of product. Cluely, a real-time AI meeting assistant, covers AI outside finance, and OneBank covers banking without AI. Ask which parts of the build your own engineers would own.
Lazarev.agency: a payments flow and a model-driven planning tool
Lazarev.agency lists San Francisco on its fintech page. Its Tratta case, labeled scaleup, covers a B2B debt collection platform. The consumer portal offers full payment, partial payment, and payment plans in one flow, with a persistent payment summary, and the admin console combines debtor management, balances, and reporting.
The Fieldstream case, labeled startup, covers an AI-driven marketing analytics platform. Lazarev.agency rebuilt it around a three-stage planning flow that ends with the model’s options side by side, each with a probability band, and a path to fork a custom plan. It reports that Fieldstream raised $3.8M in pre-seed funding after the launch. For a financial research assistant, see its Accern Rhea case in our guide to AI product design agencies.
How to choose
Pick one flow where AI touches money, and write it as a one-page brief. Say what the model may suggest, what only the person may confirm, what the screen shows when the model is wrong or unsure, and what the record shows afterward. A bookkeeping agent that reclassifies transactions, an assistant that drafts a payout, and a model that recommends a decline all fit.
Send the same brief to two or three agencies and compare the first proposal each returns. Look at where it draws the line between suggestion and confirmation. Look at which screen shows the reason for a recommendation and how the user changes it. The cases above show different answers: RedOwl’s send-for-approval step, Afternoon’s adviser sign-off, FinPilot’s consent prompt, and Piifund’s swipe to confirm. Our guide on what a user should approve when an AI assistant prepares a payment lists the states worth testing.
Match the stage to the case. With no product yet, read RedOwl and Tandom. With a live product and engineers, read Afternoon and FinPilot. With a consumer app, read Cashee.
Then ask every agency the same three questions. Who designed the failure states in the AI case? What did the client measure, and who measured it? Who from that case would work on your project? For wider lists, see our guides to fintech design agencies and AI product design agencies.
