What we do
One decision at a time. Measured, designed and shown.
An acquisition, a credit approval, a valuation. We take one kind of decision your people commit the firm to. We measure it on your own past deals, loans or valuations, design how your experts and AI work together on it, and show what a better decision is worth on your own figures.
How we work
Four steps, and what you end up with.
The method is hybrid intelligence: experts and AI on the same decision, each doing what it does best, by design.
-
Measure the decision
We take past deals, loans or valuations of one kind. Your senior people agree what a sound decision looked like before anything is run. Your experts then work the cases with AI, and without. We look at where the AI was right, where it was wrong, and where your people followed it either way.Did the decisions improve?
You end up withA task-by-task view, for that kind of decision, of where the AI can be relied on and where it cannot, and of where your people's trust runs ahead of it or behind it.
-
Design the way of working
We sort the tasks: what the AI does, what expert and AI do together, and what the expert decides. We settle what must be opened and checked, and what is recorded when someone commits the firm: the paper that goes to committee, the credit approval, the signed valuation. We set out the options and what each costs in your experts' time. Your senior people choose. Then the same past decisions are worked again, the new way.Where can we rely on it, and where must we look?
You end up withA way of working written down step by step, a record of what each decision rested on, and the before and after on your own past decisions, judged by your own senior people.
-
Show what it is worth
A better decision shows in money, not in how well the paper reads. So we work the measured difference through on your own figures: in price for an investor, in loan to value and loss for a lender, in the figure reported and the claim it could bring for a valuer. Axrea promises no return and no saving. We set out the arithmetic and every assumption in it, and you judge it.What is a better decision worth to us?
You end up withA short paper that sets the measured difference beside your own figures, in price, loan to value, loss or claim, ready for an investment committee, a credit committee or a board.
-
Re-test
The tools change. So do the habits of the people who use them. A way of working that was right when it was designed can drift without anyone deciding that it should. The same past decisions, run again, show when it has.Who tells me when it drifts?
You end up withYour own set of past deals, loans or valuations to re-test on, to run yourselves or have us run, and a re-check whenever the tool, the data or the work changes.
The split, in each decision
Know what to delegate. Know where to think.
For each kind of decision there is a right split between the AI and the expert. Three illustrative splits, one for each decision.
An acquisition paper for an investment committee. The paper reads as settled. The exit yield inside it was never tested.
The AI does
- Reads the leases, the reports and the data room
- Builds the tenancy schedule and reconciles it to the rent roll
- Drafts the paper, once the case is decided
Together
- List every assumption the price rests on
- Set each assumption beside the evidence for it and against it
- Mark every document read and every one not
The expert decides
- The exit yield and the rental growth
- What a break, a void or an incentive does to the price
- The price to offer, and what would change it
A credit paper for a loan secured on commercial property. The summary of the valuation was accurate. What the valuer was never shown was not in it.
The AI does
- Reads the valuation report, the leases and the borrower's information
- Extracts the income, the terms and the covenants
- Works out the loan to value and the interest cover on the figures given
Together
- Set what the valuer was shown beside what the lender holds
- Test the income against breaks, rent-free periods and expiries during the term
- Work the loan through at a lower value
The expert decides
- Whether the income will be there for the term of the loan
- Whether to lend, how much, and on what terms
- What goes to credit committee, and what is flagged in it
A valuation for secured lending. The lease summary was accurate. The break option was in the side letter.
The AI does
- Reads the leases and side letters
- Extracts the terms
- Assembles the comparable evidence
Together
- Mark what was found and what was not
- Reconcile the tenancy schedule to the documents
- Set competing evidence side by side
The expert decides
- The weight each comparable carries
- What a break or an incentive does to the value
- The figure, and the basis it is reported on
What we design and build
Built around your experts.
Where the work needs it, we design and build what helps your experts decide. In plain words:
Every statement tied to its source
Every figure in a paper or a report linked to the lease, the letter or the evidence it came from, and the gaps shown before anything is read.
Your own past decisions, ready to compare
Similar deals, loans or valuations from your own record: what was assumed, what was decided, and what then happened.
A clear view of where the AI can be relied on
Task by task, so your people know where to rely and where to look.
A way to re-test when anything changes
Your own past decisions, run again whenever the tool, the data or the work changes.
What we build for you is yours.
Three engagements
Begin with the first. The second is where the work is.
A team that already knows which decision it wants to start with can go straight to Design and Evidence.
Where to begin
Adoption Review
One kind of decision, looked at end to end: where AI belongs in it, where it does not, and what to do first.
You receive
- Who does what today, and where judgement must stay
- A first measure on a few recent deals, loans or valuations
- The opportunities, ranked
- A plan for the first change
The core of the work
Design and Evidence
We design how your experts and AI work together on that decision, and show the difference on your own past deals, loans or valuations.
You receive
- The measure: what your people decide on your own past decisions, with AI and without
- The design: the way of working your experts chose, and what is recorded when someone commits the firm
- What it is worth: the difference worked through in price, loan to value, loss or claim, on your own figures
- The evidence: before and after, judged by your own senior people
As AI changes
Adoption Partner
The tools and your people's habits keep moving. We re-test as they do, so you know when the way of working has stopped being right.
You receive
- A regular review of how the work is really being done
- Your past decisions run again, by us or by you, when the tool, the data or the work changes
- A short note each time: what moved, and whether the design and what it is worth still hold
- Advice you can call on
What we need from you
What it takes from your side.
The measure needs past decisions to work on. What counts as one differs by team, and it is better said now than found out later.
Valuing
Past valuations, like for like.
A valuation team has many: the same kind of property, on the same basis, some since sold or valued again. Each one is a case to work again.
Lending
Past loans, like for like.
A lender has a book of them: the credit paper, the valuation relied on, and how the loan has performed since. Each one is a case to work again.
Investing
A handful of deals is not a sample.
An investor doing a handful of deals a year does not have many decisions of one kind. So the measure works on the assumptions: the rents, yields, voids and costs in your past investment papers, set beside what then happened.
01
One kind of decision
An acquisition, a credit approval, a valuation. And a sponsor who wants it decided well.
02
A few of your experts
Some to work the past decisions, and one or two senior people to agree what a sound decision looked like.
03
Time with whoever runs the tools
To see the AI as your people have it. We need nothing from the vendor.
Your past deals, loans and valuations stay in your environment throughout. An Adoption Review needs only a few recent examples and a session with the tools.
For innovation, data and risk teams
You test the model. We test the decision a person commits the firm to.
Both are needed, and they can disagree. Every field in a schedule can be right and the decision built on it still wrong.
You test the model
- The unit is a field, a prediction or a run
- The figures are fields correct, drift and cost a run
- It tells you how the AI performs
We test the decision, with your experts
- The unit is the price offered, the loan approved or the valuation signed
- The figures are how far the decision sat from a sound one, where the expert followed the AI, and what the gap is worth
- It tells you whether your experts and the AI decide well together
- You keep the past decisions, the scoring rules and the tools
- You can run them again on every new version
- Every step is written down for your team to check
How it starts
Three steps to a first engagement.
The pack for your team
Ten pages, for investing, lending or valuing. Ask for the pack for your team.
Thirty minutes with our founder
Bring one kind of decision, with no client details. We sketch how it would be measured.
A proposal
The scope, the fee and what the work can show, all agreed before you commit.
Tied to no vendor. We resell no one's product. Fixed fees, agreed before we start.
Start here
Get the pack for your team.
Ten pages. Pages 2 to 4 take five minutes.
Thirty minutes with our founder, not a salesperson. Bring one kind of decision, with no client details, and we will sketch how it would be measured.
Arrange the 30 minutes