Case N° 01 / 03

Pier & Barrow · Private Banking · 2025

A relationship intelligence platform for the next generation of wealth.

We rebuilt the way Pier & Barrow's relationship managers prepare for, conduct, and follow up on every client conversation — and rebuilt the data fabric beneath it.

Plate — 012025

London · Geneva · SingaporePier & Barrow

Client

Pier & Barrow

Sector

Private Banking

Year

2025

Region

London · Geneva · Singapore

Duration

11 months · two phases

Team

Two product engineers, one ML engineer, one designer, one strategist

● Brief

Pier & Barrow had grown to £14.6 billion in assets under advisement on the back of personal trust, long memories, and an army of analysts producing bespoke decks for every client meeting. That model was breaking. Notes lived in eleven different places. Younger principals expected the polish of a consumer product. The bank needed a way to make every relationship manager sound like the firm's most senior partner — without losing the discretion the brand was built on.

● Challenge

The brief looked like a CRM project on paper. It wasn't. The data was scattered across a Bloomberg feed, two custodian platforms, an internal portfolio engine written in 2009, four different note-taking habits, and a Symphony chat archive. Compliance would not allow client material to leave the bank's network. The relationship managers — average tenure seventeen years — had no patience for a tool that demanded behaviour change. We had a hard ceiling on adoption: if usage at week six was below 70 per cent, the project would be shelved.

● Approach

How we went about it.

N° 01

Six weeks inside the meetings

We embedded one engineer and one strategist with three relationship teams in London for the first six weeks. No code, no slides — we sat through 41 client briefings, watched 28 portfolio reviews, and read four years of meeting notes from one principal who agreed to share. The actual workflow looked nothing like the bank's own diagrams. We threw out our initial scoping document and rewrote it.

N° 02

A retrieval layer, not a model

The temptation with private wealth data is to fine-tune. We did the opposite. We built a retrieval-first architecture — an indexed, permissioned vector store sitting on top of the bank's existing systems — and used Claude 3.5 Sonnet behind an internal gateway for synthesis. The model never sees raw client data outside an audited request envelope, every retrieval is logged, and the bank's compliance team can replay any answer the system produced down to the sentence.

N° 03

An interface designed to be ignored

The product disappears into Outlook, Teams, and the existing portfolio terminal. There is no separate app to open, no dashboard to learn. Briefings appear in the calendar invite the morning of a meeting. Follow-ups draft themselves into a relationship manager's outbox. The win was making the technology invisible enough that seventeen-year veterans stopped noticing it was new.

N° 04

Two months of shadow running

Before the system wrote a single client-facing word, it ran in shadow mode for nine weeks. Every output was reviewed by a senior analyst against the deck the team would have produced manually. We tuned, we rejected, we tuned again. By week eight the analysts were arguing with the system rather than rewriting it — which was the moment we knew it was ready.

● In their words

“It is the first piece of software I have used in twenty years that disappears. I forget it is there until I notice I have my afternoon back.”

— Senior Principal, Pier & Barrow

● Outcomes

What it delivered.

4.2x

advisor productivity per quarter

96%

active weekly use, month 9

−86%

briefing prep time

data systems unified

compliance escalations

N° 01

Adoption the bank had never seen

By week six, 89 per cent of the in-scope relationship managers were using the system in at least four meetings a week — comfortably above the 70 per cent gate. By month nine that figure was 96 per cent and the firm had stopped tracking it as a metric.

N° 02

Advisor productivity, measured carefully

We did not count keystrokes. We counted client-facing hours per relationship manager per quarter — the only number the partners cared about. That number is now 4.2 times higher than the comparable quarter the year before. Briefing preparation has dropped from an average of 3.4 hours per meeting to 41 minutes.

N° 03

A second-order effect on hiring

The firm has begun recruiting principals five to seven years more junior than its historical bench, because the platform compresses the experience gap. That is a structural advantage; it was not on the original brief.

● Selected stack

Claude 3.5 SonnetpgvectorPostgresNext.jsMicrosoft GraphAzure Private LinkTerraform

● A similar problem in your business

If the shape of Pier & Barrow’s problem rhymes with one of yours, the most useful conversation is rarely an email exchange. We will sit with two of your operators for an hour and tell you whether we can help, whether someone else can help better, or whether the problem is not yet ready to be solved. That conversation is on us.

Open the conversation All ways to reach us

Previous case

Helia Diagnostics — Healthcare

Next case

Aldoria Logistics — Freight & Supply