01
The data is still spread everywhere
Portfolio data in one system. Client notes in another. Private-market information in statements, PDFs and inboxes. Someone still assembles the complete picture by hand.
For UK asset managers, family offices and wealth firms
You have the data.
You have the use case.
You still don't have the workflow.
The prototype works. Then it meets PDFs, portfolio systems, approval rules and the people who still have to check every output.
We build that last mile. One controlled workflow, on the platforms you already own, without hiring a permanent AI engineering team.
01 One workflow
02 Your existing stack
03 Human approval stays
04 A working process, not a deck
Where the work stalls
The model is rarely the hard part.
The work breaks in the gap between a promising result and something the firm can run every week.
02
Every output still needs checking
A draft saves nothing if an analyst has to rebuild the numbers, trace every source and rewrite the report before anyone senior will use it.
03
The use case is waiting on engineers
Central technology has a longer queue. Your data team has a full roadmap. Hiring a permanent AI platform team for one business workflow makes no sense.
04
You cannot add another platform
The firm already owns Microsoft, Snowflake, Salesforce, a portfolio system or all four. The answer cannot be another layer nobody wants to maintain.
The market, in its own words
“As a smaller business, RAM will never have a team of engineers to build out a big AI platform.”
Stephen Wood, COO, Rathbones Asset Management Source ↗
92%of UK wealth firms outsource part of their business.
45%use or are considering AI within the next year.
FCA Wealth Management Survey, August 2026. Source ↗
The offer
Controlled Workflow Build
A fixed-scope engagement to take one manual, data-heavy workflow from source systems to reviewed output.
01
Week 1
Define the exact job
Map the inputs, decisions, exceptions and sign-off. Agree what good looks like before anyone writes code.
02
Weeks 2–4
Build inside your stack
Connect the required data. Encode the repeatable rules. Add the model only where it earns its place.
03
Weeks 4–5
Test the awkward cases
Run real work through it. Measure errors, route exceptions and keep a person at the points that need judgement.
04
Week 6
Put it into operation
Ship the working process with ownership, runbooks and a record of what happened at every step.
What leaves the building
A process your team can run.
No strategy deck dressed up as delivery. No open-ended “transformation programme.”
- One live workflow using your real data
- Connections to the systems it actually needs
- Clear human approval and exception points
- Tests for the errors that matter to the business
- Traceable sources, decisions and outputs
- Runbook, ownership and handover
Good first workflows
Start where expensive people are still assembling the answer.
Private markets
Capital calls and private-asset reporting
Extract information from notices and statements. Reconcile it against internal records. Route exceptions to the right person before the report moves.
Keep the review. Remove the assembly work.
Investment operations
Investment and product reporting
Bring governed portfolio, risk and performance data into a repeatable report with source evidence and approval.
Wealth operations
Client meeting preparation
Assemble the portfolio view, recent activity, CRM history and open actions before the adviser starts work.
Regulated documents
Document processing with a decision trail
Classify, extract and draft from client documents while preserving the sources, reviewer and final decision.
Who this is for
Complex enough to need it.
Too focused to build a platform team.
The best fit is a UK investment business with valuable proprietary data, an existing technology estate and a workflow that already has an owner.
You are not looking for ideas. You need the thing to run.
A strong fit
- Boutique asset manager
- Multi-family office
- Complex wealth or fund-services firm
- COO, operations or data-owned use case
- Existing platform and available data
- A clear manual cost or growth constraint
Not the engagement
- A general AI strategy
- A new enterprise platform
- Autonomous financial advice
- A chatbot with no workflow behind it
- Staff augmentation with an AI label
- A prototype with no business owner
Why a specialist team
The job crosses too many boundaries for a model vendor.
Business process before architecture
The workflow, exceptions and approval rules decide the technical shape. Not the other way round.
Your systems remain the systems of record
We work with the data platform, portfolio system and Microsoft estate already approved inside the firm.
Deterministic where money is at risk
Calculations and hard rules stay in software you can test. Models handle the work that benefits from interpretation.
The evidence trail is part of the build
Sources, exceptions, approvals and outputs are captured from the start. They are not added after compliance asks.
Who you are working with
Built by someone who has worked inside the machinery.
I'm Peter Idah. My background spans production systems, regulated industries, enterprise delivery and commercial strategy.
This work sits between the business owner who knows the process and the technical teams who know the estate. That translation matters. A technically impressive system that does not match the operating reality becomes another thing to maintain.
I take on a small number of builds at a time. You work directly with me from the first workflow map to the handover.
Before we talk
Questions a COO will ask.
No. The engagement starts from the systems already approved inside the firm. If the workflow requires a new product, you will see the reason, cost and ownership before a decision is made.
No. The build defines where software can proceed, where an exception stops the process and where a named person approves the output. Financial judgement stays with the firm.
A specific workflow, a business owner and access to the people who run it. The source data should exist, even if it is fragmented. You do not need a finished AI strategy.
The technical and control owners join the work early. Data access, architecture, security, review and evidence requirements shape the build before it reaches production.
Your team receives the working process, source, tests, operating runbook and ownership map. Ongoing support is available when the firm wants it, but the build does not depend on a permanent consultancy retainer.
Start with the stuck workflow
What is your team still assembling by hand?
Send two or three sentences. Name the workflow, the systems it touches and where it stops today.
Email Peter about the workflow ↗peter@theagenticfounder.com