Quality and control
Can a reviewer trace claims to approved sources? Record unsupported answers, permission failures and corrections. A quick draft that takes longer to repair is not an improvement.
How CURN works
From a useful question to a change your team can run.
Discover, experiment, design, implement and help the team learn. The method adapts to the engagement: advice may resolve the question; implementation and adoption can need sustained work.
Delivery method
Make each next step answerable.
These are activities to use where they help, not five compulsory packages. Scope, timing and price are agreed before work begins.
01 /
Discover
Understand the decisions, people, information and constraints. Identify what is worth improving and record how the work performs now. The useful output may be a clearer decision about where to focus.
02 /
Experiment
Test options against representative material and a defined quality standard. Compare existing software, configuration and a selective build. Keep an explicit record of errors, review effort and where a human must intervene.
03 /
Design
Set out how the work should run: inputs, outputs, permissions, approval and exceptions. Agree responsibilities and what evidence would justify moving into use.
04 /
Implement
Configure or build the agreed change, connect only approved systems and test the complete process. Give the people using it a way to inspect, correct and stop it when needed.
05 /
Help the team learn
Practise on real work within the agreed data boundaries. Make ownership and maintenance clear, record useful methods and review whether people can run the work without relying on the original implementer.
Evaluation
Decide whether the change is worth keeping.
Time and effort
Compare complete tasks, including preparation, review and rework. Keep the task difficulty and acceptance standard comparable. Track recurring software and maintenance costs alongside time saved.
Independent use
Can another team member use the method on a new case? Check who maintains sources, resolves exceptions and updates the process. Agree when to continue, change or stop.
For a practical starting point, read what AI adoption means in practice.
Illustrative situations
What the method could change.
The examples below describe possible approaches to investment and advisory work. They are illustrations, not claims of delivered client projects.
Where it lands
The work, specifically.
Most of this you can buy this afternoon. We work on the part that does not come in a box — and we say which is which.
One loop, six rooms
-
Meetings and calls
Everyone
Notes get typed during the call, then live in someone’s OneNote, or nowhere anyone else can find.
Just buy it
Off-the-shelf notetakers are good and cheap. Start here.
Then change the work
The note has to reach the file, not just sit in a transcript.
Where it breaks
Compliance deletes recordings the same day, so the summary is all you keep.
-
Long question sets
Fund managers and investor relations
Each DDQ arrives in the investor’s own format, gets split across compliance, operations and IT, and someone hunts old answers through past submissions.
Just buy it
Answer-library tools exist, and some investors pick the tool for you.
Then change the work
One approved library, every edit leaving a compliance trail.
Where it breaks
Answers have to match what compliance signed off last quarter, not just sound right.
-
Proposals and pitches
Advisory and corporate finance teams
Bid writers rebuild credentials, case pages and CVs from the last deck and the SharePoint graveyard.
Just buy it
Proposal and CV libraries are mature. Buy one.
Then change the work
Someone owns the library, or it rots inside a quarter.
Where it breaks
No tool fixes an ownerless library of stale credentials.
-
Research and synthesis
Investment and research teams
Expert calls, transcripts, broker notes and models get synthesised once, filed by personal habit, then redone when the analyst leaves.
Just buy it
Search and summarising across public sources is solved. Buy it.
Then change the work
Capture becomes part of doing the work, not filing afterwards.
Where it breaks
When the answer has to cite the firm’s own prior work, including why you passed.
-
Reading document sets
Deal teams
The index gets mapped against a request list in a spreadsheet, and juniors read for red flags while the Q&A log fills.
Just buy it
Your data room vendor already ships AI. Use theirs before buying another.
Then change the work
The first pass runs overnight, so people arrive at judgement instead of page one.
Where it breaks
Every claim needs a citation, and the room cannot be exported.
-
Decision and review packs
Investment committees, wealth and advice teams
Committee packs get stitched from models, notes and reports on a Sunday night; client review packs mean chasing valuations and re-keying between systems.
Just buy it
Nothing off the shelf assembles your pack, from your systems, in your house style.
Then change the work
The pack assembles from its sources with every figure pinned to one.
Where it breaks
Prior decisions live in email, so the next pack starts from memory.
Owned, not rented
Some of this can be yours outright. When the data cannot leave, we build the local version: a private answer library, meeting notes that never leave the building, search across the firm's own files. Built selectively, owned by you, stewarded with us.
Your version of this work will differ. That is what the first call is for.
Software and services
Software follows the work.
CURN advises, implements and builds selectively today.
Explore engagement options or read the published customer accounts.
Start with a conversation
Start with the problem.
You’ll speak directly with the people who’d do the work, about the problem and whether CURN is the right fit.
Any questions? Email info@curn.io to find out more.