Senpi: Profitability-first UX for an AI trading product

Overview

Senpi is a Base-native wallet and AI trading platform. Traders follow high-performing signals, automate execution through rule-based strategies, and act on contextual alerts without leaving the app.

Core surfaces: Feed and Alerts, Discover, Strategies, Groups, Orders, and Wallet.

The problem

Onchain markets move fast and traders often face a tradeoff: either manage too many tools at once or miss timing windows. Many products overload traders with data, while others require repetitive manual steps at the moment of action. The product had technical capability but lacked a coherent user-facing logic. The question was not what to build next but what problem the product was actually solving, and for whom.

Discovery confirmed this framing. Traders were not primarily optimizing for speed. They wanted to know a trade happened for a reason they could verify, and to be able to intervene if it did not.

Discovery and framing

I led competitive benchmarking independently and ran discovery interviews with traders alongside the founding team. JTBD analysis and positioning were my own work.

I began by benchmarking Senpi against two categories: wallets and trading platforms. Wallets reach over 90% of crypto users but stop at balances and price changes; ROI and outcome visibility are absent. Trading platforms drive most volume but overload users with data, shift risk onto the trader through disclaimers, and issue alerts without rationale. Senpi sat at the intersection: the reach model of a wallet, the depth of a trading tool. That positioning shaped every design decision that followed.

Competitive landscape: wallets prioritize reach and safety; trading platforms prioritize speed and volume. Senpi was designed to combine both.

Interviews reinforced this: traders were not motivated by automation for its own sake. They wanted confidence that automated decisions would not destroy their position while they were not watching. Trust and inspectability were the actual jobs to be done. Applying the JTBD framework shifted the roadmap: profitability visibility and risk controls moved earlier; social discovery features moved later.

JTBD prioritization: onboarding, wallet functions, limit and stop-loss orders, and strategies scored highest. These shipped in Phase 1.

Roadmap sequencing

I translated JTBD scores into phase sequencing and proposed the delivery order. Priorities were reviewed and agreed with the founders before committing to the roadmap.

The JTBD scores translated into a four-phase roadmap. Phase 1 established core execution: wallet, orders, strategies, and agent-based trading. Phase 2 added profitability visibility: portfolio PnL, discovery, and leaderboards. Phase 3 introduced engagement features: notifications, home feed, and mobile rollout. Phase 4 planned AI expansion with cross-chain access and strategy optimization.

Sequencing trust-building features before engagement features was a deliberate choice. In a product where funds are at stake, credibility cannot be deferred.

2025 roadmap: four phases from core execution and trust to profitability, engagement, and AI-powered expansion.

Design principles

I defined these principles independently to create a shared decision-making framework with engineering and product.

Four principles guided decisions across all surfaces:

Key surfaces

I owned end-to-end design of all surfaces, from problem framing through specs and QA, working async with engineers throughout. Usability tests were run by me; sessions were observed by the team.

Onboarding

Onboarding: follow groups, favorite traders, enable alerts, land in Feed.

The flow was sequenced to deliver immediate value: follow groups and traders, enable Smart Alerts, then land in the Feed with relevant context already populated.

Home Feed

Home Feed: alert context, trade of the day, trending tokens, and top groups in a single scroll.

The Feed combines a personal snapshot, contextual alerts, and lightweight discovery widgets that route directly into actions. Every alert card shows actor, action, and outcome.

Smart Alerts

Smart Alerts: each notification opens a thread with context, quick actions, and AI-assisted replies.

Alerts were generating noise: redundant messages, no clear next action, no rationale. I redesigned the alert logic using context rules and priority tiers, reducing redundant alerts by 30%. Each alert now opens a thread: what happened, why it matters, what to do next.

Discover

Discover: compare traders, view profiles, copy-trade or add to a group without leaving context.

Surfaces traders and groups ranked by outcome metrics: PnL, ROI, win rate, scam rate. Quick actions and profile views are decision-ready without losing list context.

Groups

Groups: follow curated signal sources, track performance, and configure auto-trade rules.

Traders can follow curated groups of signal sources, track group performance by PnL and win rate, and configure auto-trade rules against group signals. The group view surfaces outcome data upfront so traders can evaluate signal quality before committing.

Strategies

Strategy setup: choose a signal source, set buy and sell conditions, review, and activate.

Rules are scannable, states are explicit (active, paused), and controls are predictable so traders can intervene fast.

Orders

Orders: open and closed views, order detail with trigger context, and filter controls.

A state-based view separating open exposure from closed outcomes. The open/closed split was added after usability testing revealed that a unified view caused confusion and increased support volume.

Wallet

Wallet: view assets, inspect token details, swap tokens, and review transaction history.

Supports custody without breaking momentum: balance-first views, token details with market data, swaps, and receipt-level confirmations.

AI Assistant

Chat assistant with contextual skills and templates for common trading actions.

Embedded into alerts and chat. Explains what happened, surfaces quick actions, and executes commands without leaving the notification context.

Iteration

Issues were identified through usability tests I ran and post-launch usage patterns surfaced by the team. I designed each response; engineers shipped.

Senpi evolved rapidly after launch. The following iterations show how specific pain points translated into design changes with measurable outcomes.

Reflection

Traders keep using a product when it reduces interpretation work. The most useful changes were the ones that made execution and automation readable in seconds: separating open versus closed orders, showing strategy status as active or paused, and structuring alerts so each update points to the exact screen where the trader can intervene.

Working in a trust-sensitive environment early in a product life also clarified something about prioritization: the features that build trust need to ship before the features that extend capability.

When funds are at stake, credibility is the product.