Find where your margin is won and lost.

Then know if the next deal makes money before you commit.

Start with the data you already have: savings ranked across your suppliers, pricing gaps against your customers, and money owed to you on both sides. From there, every purchase order, bid and contract can get a verdict before signature. Used by enterprises with long customer contracts and by private equity teams across whole portfolios.

Start a four-week sprint The full film · 7 min

What the platform does

Four teams of AI agents, working on both sides of every deal.

Each team does the work on what you pay suppliers and what you charge customers, and your people approve what matters. Open a card to see what a team does, then any capability to see it at work. You can start with any team.

Always on

Across everything

AI agents that work across every team, on both sides of every deal.

Agents in this team

Chat agent (Cassian)
Visualisation agent
Data agent

6 capabilities

Start here

Identify Team

Agents that find where margin is won and lost, on both sides of every deal.

Agents in this team

Data agent
Market agent
Pricing agent

19 capabilities

Or start here

Recover Team

Agents that track every saving to EBITDA and get the money back.

Agents in this team

Verdict agent
Data agent
Visualisation agent

4 capabilities

Includes Identify

Protect Team

Agents that check the next deal before anyone signs it.

Agents in this team

Simulation agent
Pricing agent
Verdict agent

7 capabilities

Copper10,179 $/t▲ 2.4%

Aluminium2,702 $/t▲ 2.7%

Steel HRC724 $/t▲ 0.8%

Diesel105.5 p/l▲ 2.0%

UK power84.4 £/MWh▲ 3.1%

Freight1,949 index▲ 0.9%

Brent75.9 $/bbl▲ 3.1%

Labour4.148 % y/y▲ 2.3%

EUR/GBP0.888▲ 1.8%

Polymers1,206 €/t▲ 3.1%

Copper10,179 $/t▲ 2.4%

Aluminium2,702 $/t▲ 2.7%

Steel HRC724 $/t▲ 0.8%

Diesel105.5 p/l▲ 2.0%

UK power84.4 £/MWh▲ 3.1%

Freight1,949 index▲ 0.9%

Brent75.9 $/bbl▲ 3.1%

Labour4.148 % y/y▲ 2.3%

EUR/GBP0.888▲ 1.8%

Polymers1,206 €/t▲ 3.1%

Finance, procurement and commercial teams at these companies already work this way.

Why it works on real data

The agents follow a playbook our analysts wrote by hand, one spend line at a time, since 2020.

01Six years of checked answers

Every category rule and every contract check comes from work our analysts did by hand for clients. Anyone can connect a language model to a spreadsheet. Knowing what a correct answer looks like is what takes the time.

02The model asks. Tested code calculates.

The language model decides what to ask of your data and explains the result. Every number comes from tested, deterministic tools, and anything that cannot be traced to your own data is removed before you see it.

03Built by a product team, tested by clients

A full-time product and engineering team builds the tools, and live clients put them through real contracts and invoices every day.

How the AI works and how your data is kept safe →

One contract, 31 months

Most contracts that lose money looked fine the day they were signed.

MONTH 0Signed at 11% margin. Everyone is pleased.

MONTH 7Copper rises 18%. The supplier invoice changes the next month. The customer price does not.

MONTH 19Margin is down to 4%. Nobody can say which contract did it.

MONTH 31It loses money on every job, with 29 months still to run.

DAY 0With Spendkey, the verdict comes first. One copper clause, and the margin holds for the whole term: +£890k of EBITDA kept.

NORTHLINE RAIL · 5-YEAR MAINTENANCEMONTH 0

The film

Watch the platform follow that contract, end to end.

Your data, the market, margin and buffer pricing, the invoice check, your rules, a verdict on every purchase order, Cassian and EBITDA, in seven and a half minutes.

Customers

The product comes to life in what our customers did with it.

Traffic lights at a wet city junction at night, representing the traffic systems sector

Infrastructure

€2.03bn

Every supplier, every invoice, agreed to the euro

9,919 supplier names cleaned to 8,046 parent groups, with every dashboard total agreed to the euro.

Yunex TrafficRead the story →

Rooftop heating and cooling plant at dusk, representing the building services sector

Health and fitness

€10m–€15m

Savings found in HVAC and cleaning, club by club

€131m in 13 categories, every supplier invoice line classified, and 12,162 supplier records rolled up to 79 groups.

Basic-FitRead the story →

Rows of servers in a data centre hall, representing the data centres sector

Retail

€4.8m

Technology costs cut, and banked

€46m a year of technology spend renegotiated across 26 suppliers, and the savings landed in the budget.

Global Fashion GroupRead the story →

Pharma

$25m

44 ERP instances, one clean view

200,000 transactions a year from 44 ERP instances in 20 languages. Reporting went from four weeks to instant.

MundipharmaRead the story →

Industries

Built for companies whose contracts run for years while costs move every month.

Electrical switchgear being assembled in a factory, representing advanced and contract manufacturing

Advanced and contract manufacturingParts priced on metal, resin and energy, sold on fixed supply agreements

A train in a rail maintenance depot, representing the infrastructure sector

InfrastructureMaintenance and frameworks priced once, delivered for years

A fleet of service vehicles parked at a depot at dusk, representing industrial services

Industrial servicesRate cards on one clock, fuel and labour on another

Drums of copper cable in a warehouse, representing every other industry

And every other industryThe supplier side works wherever you buy from more than 200 suppliers

The platform, live

Everything a deal touches comes in. Every result comes out in EBITDA.

Many agents, one clean dataset. They classify every line, read every contract, and link supplier and customer contracts to the invoices, bids, purchase orders and transactions behind them. They watch the markets, write the verdicts, and tested code runs every sum.

Supplier contractsrates, index clauses, rebates1,840

Customer contractsprices, reviews, pass-through963

Bids and quotesbefore they go out384

Purchase ordersbefore they are raised2,140

Invoicessupplier and customer38,120

Transactionsevery ledger line482,100

Market pricesLIVE

Copper10,086 $/t▲1.3%

Steel HRC730 $/t▲1.9%

Diesel105.4 p/l▲1.8%

UK power83.6 £/MWh▲1.5%

THE MARGIN CONTROL PLATFORM

Data agentMarket agentPricing agentSimulation agentVerdict agentVisualisation agentChat agent

One clean dataset · every figure traced

Verdicts before signature11of 214 orders this week need the CFO

Money brought back£612kovercharges and unbilled revenue

Questions answered1,240by Cassian this month

EBITDA, this year+£4.60magainst plan

Real-time connectors, several ERPs at once

Xero

Ariba

Google BigQuery

Azure Synapseand 50 more

Who it is for

Everyone around a deal gets the same verdict, on their own screen.

A chief financial officer reviewing figures on a laptop

##INLINE1##

Sign off eleven, not two hundred.

##INLINE2##need your decision

11 reach you, with the fix

Two hundred purchase orders a week, and only the eleven that break a rule reach you, each with the money and the fix.

A procurement lead checking stock on a tablet in a warehouse

##INLINE3##

Purchase orders at the contract terms.

£4.2mbanked in EBITDA

##INLINE4####INLINE15##£7.4m

Agreed##INLINE16##£5.6m

Banked##INLINE17##£4.2m

A verdict on every purchase order and supplier contract before approval, and leakage stopped before it is paid.

A commercial director meeting with colleagues around a table

##INLINE7##

Quotes at the right price.

3.5%price buffer, copper

##INLINE8## ##INLINE9##

Every customer contract checked against the costs underneath it, the buffer each bid needs, and every price rise the contract allows billed.

A private equity operating partner looking out over the city from an office

##INLINE10##

One view across the portfolio.

£38mEBITDA opportunity

6 companies × 4 levers, refreshed weekly

Every company's spend, contracts and margin rolled into one portfolio view, and the companies ahead of plan named.

Security and trust

Enterprise security, and a number you can check.

Your data reaches us through one encrypted transfer and stays in a region you choose, on Microsoft Azure. Cassian reads it through read-only tools and cannot change a record.

Figures traced to your own data100%

Unit tests before every releaseMore than 10,000

Transactions live with clientsMore than 100 million

Sign-inSingle sign-on, role checks on every request