Skip to main content
Fleur Lamont Logo Fleur Lamont

The North East AI plan and your late payments

AI
AI Small Business North East Late Payments Cash Flow

A stack of paper invoices and a worn ledger on a wooden desk at dusk, with a pen, a mug of tea and a phone, and a blurred shopfront through the window behind

The North East's AI plan contains two numbers, and they are not about the same thing.

"The Government has already committed to investing an initial £5 million to boost AI skills, adoption and innovation across the region."

"Private sector partners are planning to invest £30 billion in the physical assets and wider ecosystem over the next decade."

Both from page 23 of the North East AI Growth Zone Prospectus, a draft for consultation from the North East Combined Authority, May 2026.

Six thousand to one. That is the ratio, and for a firm of twenty people it is the whole story.

What the thirty billion buys

The thirty billion buys concrete, power and chips. £10 billion of it was committed by QTS and Blackstone for a data centre campus at Blyth, due to open in 2028 (Department for Science, Innovation and Technology, 12 May 2026). The second site is Cobalt Park at Wallsend. More than 5,000 new jobs came with the designation (Department for Science, Innovation and Technology, 17 September 2025).

The five million is spread across skills, adoption and innovation, and a good part of the skills money is aimed at classrooms. On 12 May 2026 the Mayor put £750,000 into the government's TechFirst programme, on top of £1.5 million already invested by government. That brings AI and digital skills to 30,000 primary school children and backs 1,000 teachers (Department for Science, Innovation and Technology, 12 May 2026).

None of that is wasted. It is aimed at schools rather than at overdue invoices.

Neither headline is about a firm of twenty people. You are not a data centre operator and you are not a school. The plan reaches you in one place.

What the plan says about small firms

Page 12 of the prospectus, under the heading for AI adoption:

"For our small businesses, this means tools that will help to tackle the challenges that hold them back, like chasing late payments, managing cash flow, and tracking inventory."

Late payments first. Before cash flow, before stock, before anything else.

That sentence was written by somebody who has watched a firm lose sleep over an unpaid invoice.

On 21 October 2026 the North East Combined Authority published a notice on Find a Tender, the government's public contracts site, reference 102557-2026. It says the programme "is moving into its delivery phase", and it sets out three forthcoming work packages: support for high-potential digital and AI companies, an ecosystem profile and coordination, and one called "AI Adoption in Key Growth Sectors", to help "businesses in key sectors of the North East economy to move from early experimentation towards the operational adoption of AI technologies".

The document is a preliminary market engagement notice under the Procurement Act 2023. Its subject is a webinar for prospective suppliers, an early look at three work packages not yet tendered, and a questionnaire to gather market insight ahead of the formal procurement process. The buying has started. Nothing has been bought.

The notice also puts a figure on the delivery: an estimated total value of £5,000,000 including VAT, running from 16 November 2026 to 15 November 2030. That is the same five million. On 22 September 2026 the Combined Authority's Cabinet approved the programme and agreed £5 million of funding for it.

Keep those two things apart. The work packages are for businesses in key sectors, not for small firms. The small-firm sentence is one line in a May 2026 consultation draft. The notice does list small and medium-sized enterprises as suitable participants, but that is about who can bid, not who the work is for.

The region has committed money to adoption. Whether any of it lands on a firm of twenty is not what the notice says.

What a chasing tool knows

Here is the part that is easy to miss. It is not about software.

A tool that chases late payments works on the average case. It knows the invoice is overdue, because the invoice system holds the amount and the date. It sends the reminder, then the firmer one, then it tells somebody who to ring on a Tuesday morning. That is genuinely useful, and for most late invoices it is the whole answer.

What it does not know is why the customer has not paid. It is not going to ask. Sometimes the invoice went to a person who left in September. Sometimes the customer is holding it because the work was not right, and the fix is not a reminder. Sometimes their own customers have not paid them and they are two weeks from a problem of their own.

And this is where the tool quietly makes things worse. The reminder goes out, and the customer who was going to raise the fault is annoyed instead. The average case gets better. The hard case gets a firmer letter. A system that knows the amount and the date will treat a problem that would have been settled on a phone call as one to settle by email.

So the honest question, the one the programme does not answer: of the money that is late to come in, how much is late because nobody chased it, and how much is late because the work was not right?

That is answerable. A firm of five to fifty people can get it out of its own ledger in an afternoon, and decide whether a chasing tool is the right spend.

The split

Do this before you spend anything.

Start with the unpaid invoice list, which your accounting software can usually produce. You want a number, not a list, so start with the total and then the shape of it: what is overdue by more than thirty days, more than sixty, more than ninety. Firms running this for the first time usually find most of the money is old.

Then sort the list into three piles.

Nobody chased it. The test is simple. Search your email for the customer's name, look for anything you sent about the invoice. If there is nothing there, and you did not even know to chase it, it is in this pile. Nobody in the business has asked for this money.

Under dispute. You know these ones. You have a memory attached. The job was not right, or the customer thinks the amount is wrong, or they are holding the money until you send a revised quote. If you were to ring them today, they would tell you why they have not paid.

They cannot pay yet. The work was fine, there is no complaint, and the money is not there. Their own customers have not paid them. They are waiting on a payment run. You cannot tell this pile from the first without asking, which is why it is worth asking.

Put a value on each pile rather than a count. Not how many invoices, how many pounds. The number that matters is the value in the first pile, because that money is late for a reason entirely inside your business.

The whole exercise takes an afternoon. Most of the work is sorting rather than analysis, and you already know roughly which way it goes. Roughly is not a number, and a number is what tells you what to do.

What each answer means

Most ledgers are mixed. The point is not that every invoice lands in one pile. It is that you find out which pile is carrying the money, because the three piles have three different answers.

If most of it is in the first pile, the cause is inside your business. Nobody chased it, so the fix is chasing more or chasing sooner. Software that chases sooner earns its place, because it does the part you keep not doing. If nobody owns collections, then the spend is a change of process. A tool will chase for a firm that has decided to chase. It will not chase for a firm that has not.

If most of it is in the second pile, a chasing tool is the wrong spend, and you have saved the cost of one. The money is not late because you were silent. It is late because something about the work is not right, and that is a conversation, a concession or a piece of rework. A firmer letter to a customer who is right about a fault does not make the fault go away.

If most of it is in the third pile, the question is not AI at all. It is credit control: terms, deposits, limits on what one customer can owe you, and whether you would rather have a smaller amount on time than a larger amount late. A firm that has never set that limit has a policy problem, and no software writes it.

Where this sits

This is the same discipline as the European Technology Risk Assessment: read the exposure honestly, name what is actually at stake, and decide the response before the event rather than during it. The ETRA looks at that question across a region. This looks at one ledger.

If you run a business of between five and fifty people, the seven checks are on one printable page, free, at fleurlamont.com/guide.html. The European Technology Risk Assessment is free to download at fleurlamont.com/ebook.html. The point is not to buy the tool. It is to know whether the tool is the answer.

Sources

This site does not offer financial advice. It offers a way to look at the week in front of you.