Update, : Added a demo with fictional data that shows the three parts on a made-up club.
Ask a football club’s finance director a simple question: what is our cash position today, and what will it be at the end of January?
The first half is usually easy. The bank balance is known. The second half is harder than it looks.
Money arrives from many directions on many schedules. League distributions, broadcast payments, UEFA prize money, sponsorship instalments, ticketing, hospitality, transfer fees owed by other clubs. Money goes out on its own schedule too: wages, bonuses, transfer instalments owed to other clubs, agent fees, suppliers.
Much of it is governed by contracts with unusual clauses. A transfer fee might be paid in four instalments, with add-ons triggered by appearances, a sell-on percentage and a different amount if the buying club is relegated. A sponsorship deal might include bonuses for reaching the Champions League. Thousands of invoices sit on top of that.
An AI-generated illustration, made with ChatGPT. The charts in it are decorative, not real data.
#Why it is getting harder
Football has quietly become a credit market. Transfer fees are rarely paid upfront. They are spread over several years, which means clubs owe each other large sums on fixed schedules.
One analysis by Paul Quinn for The Esk put the figure, at 30 June 2025, at 3.50 billion pounds of transfer payables for the twenty Premier League clubs, against 1.25 billion pounds of receivables. The same analysis estimated that Premier League clubs committed a record 3.48 billion pounds in gross fees in the summer 2026 window alone.
On top of transfers, the other big income streams arrive on their own calendars. League distributions follow the league’s payment dates. UEFA prize money depends on results and arrives in stages. Sponsorship and commercial contracts have their own instalments, often with performance bonuses. Matchday income follows the fixture list.
Some of that money is also borrowed against before it arrives. The same analysis describes clubs borrowing against transfer receivables, including from alternative lenders. That adds another layer: a receivable that has been financed is no longer cash the club can count on in full.
And the rules have tightened. UEFA’s squad cost rule and domestic spending rules mean a club needs to know not only whether it can pay, but how a deal will look against its limits at the next reporting date.
#What a late payment costs
Take a simplified example. A club sells a player for 20 million euros, paid in four equal instalments of 5 million euros over two years. It plans its January window on the assumption that the second instalment arrives in December.
The buying club pays six weeks late. Nothing dramatic happens legally. There may be a reminder, perhaps a claim later. But in the meantime, the selling club has 5 million euros less than it expected during the one month when it most wants to act.
It can wait, borrow or change its plans. Waiting may mean losing a target. Borrowing has a cost. Changing plans may mean selling a player it wanted to keep.
Now multiply that across a real club: several transfer receivables, each with its own schedule and add-ons, sponsorship bonuses that depend on the league table, and thousands of supplier invoices with their own terms. The finance team usually knows each piece. In the clubs we have talked to, bringing them together into a reliable picture of the next six months often depends on spreadsheets maintained by one or two people.
This is a classic ERP problem. The ledger records what happened. The harder question is what is likely to happen next, and when.
#What we are building
We are building a layer on top of a club’s finance systems that answers one question well: what is our financial position now, and what is it likely to be on a given date?
The first part is contract terms. Transfer agreements, sponsorship deals and commercial contracts contain the clauses that decide when money moves: instalment dates, add-ons, sell-on percentages, performance bonuses, relegation clauses. The system extracts those terms from the documents and turns them into expected payments. A person checks each extracted term before it is used. We extract the terms with Claude Sonnet 5.5, run in October 2026.
The second part is the cash forecast. Confirmed terms, open invoices, wage commitments and expected distributions are brought into one schedule. The calculation is designed to be deterministic: to follow the contract rules and the figures supplied, so that every number can be traced back to the clause or invoice behind it.
A finance director can then ask:
“What is our expected cash position on 31 January if we reach the Champions League quarter-finals, and which receivables are we relying on to get there?”
The third part is payment delay prediction. Every club has a history of who paid on time and who did not. The aim is to use the club’s own payment records to estimate the likelihood that a given receivable will arrive late, and by roughly how much. We make these predictions with Claude Opus 5.5, run in October 2026. That turns a single forecast into a range: the expected date, and a more cautious one.
This is the only part where a model makes an estimate rather than following fixed rules, and it is labelled that way.
The demo in this note shows the three parts on a made-up club, Northbridge FC. Everything in it is fictional: the club, the ten contracts, the payment history and the numbers. The extracted terms and the delay estimates were written for the demo, not produced by the models. On 31 January, compare the expected and cautious figures and see which late receivables push the cautious one below the board minimum. Then move the date to 30 April and switch the Champions League scenario from quarter-finals to round of 16. Open Contract terms, pick a contract from the list and hover a term to see the clause it came from.
#What still needs people
Contracts in football are full of judgement. Whether an add-on is likely to be triggered depends on whether the player will actually play, which depends on the coach. Whether a sponsorship bonus is realistic depends on where the team finishes. A finance team can set those assumptions. A system should not guess them silently.
Relationships matter too. A club that pays late every year may still be a reliable partner on other deals. A model that reads the payment history will flag it. Whether to chase, renegotiate or accept the delay is a decision for the finance director.
And the board decides how much risk the club is willing to carry. A forecast can show the range of outcomes. It cannot decide which end of the range the club should plan for.
#What it will not solve
It cannot make a club solvent. If the spending plan does not work, a clearer forecast will show that sooner. That is useful, but it is not a solution.
It depends on clean data. If invoices are not recorded properly, or side letters to a contract sit in a drawer, the forecast will be wrong in ways that are hard to spot.
The payment delay predictions have their own limits. They are based on the club’s history, so they say less about a counterparty the club has never dealt with. A buyer in financial trouble may pay late for reasons no historical pattern would predict. The estimate should be treated as a warning, not a fact.
There is also the familiar risk that a neat chart looks more certain than the numbers behind it. People need to see which figures are confirmed, which are assumptions and which are predictions.
#What we are still testing
So far we have run all three parts. We ran the extraction on 30 real contracts shared by partner clubs, with some details anonymised for confidentiality. We are not naming the clubs. After our team checked each extracted term against the contract, 27 of the 153 extracted terms (18%) needed correcting. A correction meant a wrong amount, a wrong date, a missing term or a misread clause. Most of the reconciliation work came from the contracts representing amounts in different ways.
We ran the cash forecast and the delay predictions on test data, not on a club’s records. They have not yet produced a result we can check against what actually happened.
We are testing how accurately the system extracts payment terms from real contracts, especially the unusual ones: conditional add-ons, sell-on clauses with exceptions, bonuses tied to several conditions at once.
We are also testing whether the delay predictions are useful in practice. A model can be accurate on average and still unhelpful if it is wrong about the one payment that matters in January.
And we want to understand the checking work. If the finance team spends longer reviewing extracted terms than it used to spend reading the contract, the tool has not helped.
We do not yet have evidence that this improves financial decisions. For now, the practical question is whether a finance director can get a reliable answer to “where will we be on this date?” faster than today.
We still want to learn how other club finance teams approach this problem.
#Beyond football
Any organisation whose income arrives in large, irregular, contract-driven amounts has the same problem. Construction firms paid by project milestones. Biotech companies funded by grants and licensing deals. Schools and universities waiting on public funding cycles.
The contracts and rules differ. The question does not: given everything we have signed and everything we are owed, where will we actually be on the date that matters?
#My take
Football finance gets attention for the headline numbers: record fees, record wages, record broadcast deals. The day-to-day difficulty is less visible. It is knowing, with some confidence, how much money will actually be available at a specific moment.
I think AI is useful here for two narrow reasons. It can read contracts and invoices faster than people can, and it can find patterns in a club’s own payment history. Neither replaces the finance team. Both can give it a clearer starting point.
I would start with one season of transfer receivables and payables, compare the forecast with what actually happened and let the finance team pick it apart.