Cordrey Consulting
Practical AI for regulated service firms

Run a calmer firm.

We build practical AI and automation into the systems your firm already runs — so the admin, the retyping and the chasing get handled by software, in-house, not by another set of people.

Your time, your control, your best work — back.

Built on tools you already pay for No data team, no transformation programme Every step logged
The real product

The point isn't the hours. It's what they're for.

Plenty of firms will promise you time back. The question is what your week looks like afterwards — and who's in charge of it.

Less of this

  • Retyping the same client details into three systems
  • Chasing documents, signatures and replies
  • Formatting reports nobody enjoys writing
  • Hunting through old files for "how did we handle this last time?"
  • Evenings spent on admin that never made anyone a penny

More of this

  • Client conversations — the work you actually trained for
  • Judgement calls made with the full file in front of you
  • Capacity to take the referrals you used to wave away
  • New services you couldn't staff before
  • A firm you run, rather than a firm that runs you

Every automation is approved by you, logged end-to-end, and reversible. Control isn't the price of the time — it's part of the product.

What we actually build

Three kinds of leverage. Most stop at the first.

Workflow automation — Zapier, Make, Power Automate — is a decade old and you've been sold it before. It's the floor, not the ceiling. The real shift is what comes after.

2 of 3 · Augment

The easy stuff — handled, and finally reliable.

Enquiries acknowledged, records created, documents chased, filings logged. The workflows you've heard about for ten years — built on the tools you already pay for, and actually maintained.

This is the floor. If a consultant's pitch ends here, you're being sold 2015.

Routine enquiry, start to filed0.0s
  1. Read the email & extract the details
  2. Draft the reply — for your approval
  3. Create the record in your CRM
  4. File, log, notify the right person
By hand: ~20 min. Here:

A second pair of eyes on every file.

Before an annual review, the whole client file — CRM, inbox, documents — is read, cross-checked against your own checklist, and turned into flags with reasons and a draft agenda. You walk in prepared; nothing slips.

Zapier can't do this. This is judgement work, augmented — and it's where the last two years of AI actually changed the game.

Pre-meeting file review0.0s
  1. Read 14 documents across CRM, inbox & drive
  2. Cross-check against your review checklist
  3. Flag what needs attention — with reasons
  4. Draft the meeting agenda for approval
Risk profile is 26 months old — outside your 24-month policy
Fixed rate ends 14 March — no review booked
£14,200 of this year's ISA allowance unused — raise at review
By hand: an afternoon. Here:

Your expertise, working at scale.

The knowledge locked in your firm's heads can become integrated systems that sit alongside your existing infrastructure — client portals, self-serve fact-finds, calculators — freeing your people to spend their time on the work they're best at, and serving clients you couldn't previously afford to take on.

Not hypothetical: this is Keystone — live today, replacing a 40-tab Excel fact-find at an international advisory firm. It's built around your fact-find, not a generic one: your questions, your sections, your wording. And the answers don't stop there — they fill in your own application forms and provider paperwork straight from your templates, so nothing is ever retyped.

Advisers advise; the workbook is gone.

demo.keystone.cordrey.co
KEYSTONE
Client fact-find — without the spreadsheet
1 · Personal detailsCOMPLETE ✓
2 · Employment & incomeCOMPLETE ✓
3 · Pensions & investmentsIN PROGRESS
● live — validated as the client types; nothing retyped, nothing lost
After hours

And it keeps working after you've gone home.

Chasing, filing, reconciling, drafting — running quietly overnight, with every step logged and waiting for your approval in the morning.

Governance

Built for regulated environments.

Ten years running operations in financial advice and banking taught me what defensible looks like. AI in a regulated firm is a governance question before it's a technology question — and every system here runs under the same written rules.

  • A person approves anything that leaves the building. Client emails, filings, publications — drafted by the system, sent by your firm.
  • Every automated step is logged and reversible. An audit trail you can hand to a compliance officer.
  • Your data is never used to train models. Processing runs in your own accounts; nothing is pooled or resold.
  • Least-privilege access, UK & EU GDPR by default. Scoped before we build, written down, revocable by you.

You get this as a one-page governance note with every project — and the full working detail lives in the AI governance series in the knowledge base, mapped to your regulator's actual hooks.

Your systems

CRM · inbox · document store — your data stays where it lives

read via your own accounts, scoped to the job

Processing

runs under your credentials, UK / EU region

nothing leaves without sign-off

Human approval

someone at your firm — the last step is always a person

Never

model training on your data · third-party sharing · shadow copies

Free template

The AI governance pack, ready to fill in.

The whole programme above as an 8-item checklist and a fill-in PDF template — the register, the policy outline, the DPIA prompts, the incident playbook. Built for a firm with no risk department.

Get the governance pack
The means

Three ways in.

The services exist to get you the week described above — start at whichever step fits where your firm is now.

Works alongside what you already run — not instead of it Microsoft 365 Google Workspace Xero Zoho HubSpot your CRM Zapier · Make · n8n ChatGPT · Claude · Gemini …and the spreadsheet you secretly still use
Built in the open

Judge the work, not the promises.

Most consultancies show you slides. Everything I recommend to clients, I practise on my own estate first — and the proof is public.

7 tools live in Labs — calculators, portals, full working systems you can try right now
0 days from discovery call to a first working system — one process, fixed price
100% of automated steps logged and reversible — control is part of the product

Start with the Labs — Keystone, the Spain mortgage calculator, the editorial desk — or read what one month of the software factory actually looks like.

Thinking in public

From the Journal & Knowledge

One conversation. No programme, no pitch deck.

A free 30-minute call finds the smallest honest first project — usually one process, fixed price, results inside a month. If there isn't one, I'll tell you.

Book a discovery call