← Selected work

Bespoke

bespke.co
Role
Product studio + transformation
Timeline
2025 · ongoing
Stack
AWS · LiteLLM · Temporal · pgvector · Langfuse

Overview

The intelligent execution layer for modern investment management.

Bespoke is an AI-native platform for asset and wealth managers that sits on top of a firm's custodians, order systems, and CRMs.

We work within Bespoke in two capacities: as an embedded product team, owning product strategy, technical architecture, and end-to-end product delivery; and as an internal transformation partner, applying the same agentic approach to workflows across marketing and engineering.

The problem

Investment firms are capacity-constrained.

Advisors lose hours to admin, meeting preparation, portfolio-drift monitoring, reporting, and compliance. The data they need is scattered across custodians, order systems, CRMs, and market feeds.

The question is simple: how would you scale if capacity wasn't the limit?

The product

A governed agent stack, built for regulated investment management.

In a regulated business, “the model decided” is not an answer. Every agent action is auditable, proposing is separated from executing, and sensitive actions wait on a person.

Task-specific agents handle research, portfolio monitoring, reporting, and compliance, drawing on live portfolio, market, CRM, trade, and proprietary knowledge.

A shared tool layer connects the agents to the firm's systems, while page-level citations trace every claim back to its source. Deterministic workflows and golden-fixture tests keep the system reliable as it evolves.

AI-native operations

Taking the agentic approach inside the company.

We built a content engine that turns research into content, distribution, and demand. AI helps turn articles and insights into social content, with humans reviewing and shaping what gets published.

For high-intent prospects, proprietary research feeds personalized market briefs generated through a DAG, giving each prospect relevant insights and creating a path from interest to conversation.

We are also introducing an agentic development process for engineering, where agents work from machine-checkable requirements and route ambiguity back to people.

We're helping Bespoke build an AI-native way of working, transforming processes across the company and letting the team do more.

Let's start building.

Talk to ushello@strangeloop.works
next case — 003
Pairvi