AI May Democratise Legal Drafting. Can the Courts Survive the Volume?

The central tension

AI could widen access to legal information and help people articulate cases they could not otherwise afford to present.

But every new document still needs to be received, understood, tested and decided.

AI can scale legal production much faster than we can scale judges.

Five things to understand first

1. There is no established evidence that AI is causing the current UK caseload increase.
Recent official statistics show higher volumes in several jurisdictions, but causation should not be invented.

2. The capacity problem nevertheless exists already.
The Crown Court, County Court and private Family Court are all operating under substantial pressure.

3. AI may make legal participation cheaper.
That is potentially a major access-to-justice benefit.

4. AI may also make overproduction cheaper.
The same technology that helps someone formulate a legitimate case can generate 100 unnecessary pages.

5. Courts will need technology too — but under very different constraints.
Judicial AI must preserve independence, confidentiality, fairness and human responsibility.

AI could remove a genuine barrier to justice

Producing legal work has historically required one of two scarce resources:

money or time.

A represented person buys professional time.

A litigant in person supplies their own.

Either route can be expensive.

Generative AI changes that equation.

It can help a person:

• understand unfamiliar terminology;
• identify procedural questions;
• organise evidence;
• build a chronology;
• structure a document;
• reduce repetition;
• identify information gaps;
• and prepare more effectively for professional advice.

That has enormous potential.

The UK Government’s 2026 Legal Services Advisory AI Growth Lab explicitly identifies faster and more affordable legal services and improved access to justice as objectives of responsible AI innovation.

We should take that opportunity seriously.

For many people, the alternative to AI-assisted legal understanding is not a fully funded solicitor.

It is no assistance at all.

But access for one person creates processing for another

A legal document does not disappear once it has been generated.

If it is filed, somebody must process it.

If it contains allegations, somebody may need to answer them.

If it cites authorities, those authorities may need to be checked.

If it raises issues, a judge may need to determine whether those issues are relevant.

AI therefore changes the economics on both sides.

The cost of producing material falls.

The cost of consuming it may remain.

That is the system-level version of the “AI slop” problem.

One litigant’s ability to generate another twenty pages may create twenty pages of work elsewhere.

The UK court system is already carrying substantial volume

The latest official figures are important.

County Court

Between April and June 2026, County Court claims increased by 11% compared with the same quarter in 2025, reaching approximately 571,000 claims.

Defended claims increased by 15%.

Private Family Court proceedings

During the same quarter:

• 14,833 private-law Children Act cases started — an increase of 11%;
• 41,637 private-law cases were open at the end of June — 14% higher than a year earlier;
• 1,672 had been open for 100 weeks or longer;
• and 51% of private-law disposals involved neither party having legal representation.

Criminal courts

At the end of June 2026:

• the magistrates’ court open caseload stood at 380,230 — a series peak;
• the Crown Court open caseload stood at 80,829 — also a series peak;
• and 23,706 Crown Court cases had been open for at least a year.

These statistics have different causes in different jurisdictions.

They should not be collapsed into one story.

But together they make one point unavoidable:

judicial capacity is already scarce.

There is no evidence here that AI caused those increases

This distinction is critical.

It would be tempting to place the summit prediction beside the current statistics and declare:

“AI is already flooding the courts.”

We cannot responsibly say that.

The official County Court statistics, for example, identify growth in money claims and indicate that much of the increase is likely connected to private parking debt recovery.

Family-court volumes have their own drivers.

Crown Court pressures have their own.

Walters himself was careful to describe the link between generative AI and increased filing as conjecture rather than established causation.

That is the correct evidential position.

The more useful question is prospective:

If AI continues lowering the cost of producing legal material, what should courts do now to prepare for the possibility of substantially greater volume?

The technology asymmetry problem

Imagine the following system.

The claimant has powerful AI.

The defendant’s lawyers have powerful AI.

Both can produce substantially more material.

The judge has the same number of hours in the working week.

That is not technological equality.

It is an amplification machine feeding a fixed-capacity decision-maker.

Something has to change.

Either:

• parties generate less;
• procedural rules become stricter;
• courts gain tools capable of processing more;
• or delay increases.

Probably all four will occur to some degree.

The judiciary is not standing still

The idea that parties will have AI while judges remain entirely analogue is already becoming outdated.

The judiciary has issued specific AI guidance for judicial office holders.

The October 2025 guidance addresses:

• hallucination;
• bias;
• confidentiality;
• private information;
• and personal judicial responsibility for material produced in a judge’s name.

In June 2026, the Ministry of Justice announced work on:

• AI legal assistants;
• AI-supported case analysis;
• and an AI tool intended to assist Crown Court listing.

The Government’s wider court-reform programme has continued to identify AI-assisted listing as part of its response to delay.

So the policy question is no longer whether courts will encounter AI.

It is how institutional AI can be deployed without weakening the safeguards which distinguish adjudication from automated administration.

Access to justice and AI slop are not opposites

We need to be capable of holding two propositions at the same time.

AI can improve access to justice.

And:

AI can create enormous amounts of low-value court material.

The answer is not to decide that one proposition cancels the other.

It is to design procedure which encourages the first and discourages the second.

That means focusing on:

• page limits;
• defined issues;
• evidence source;
• proportionality;
• verified authorities;
• structured documents;
• and clear consequences for persistent procedural non-compliance.

Read the related JSH Law analysis: AI Slop, Costs and Litigants in Person: Should Courts Sanction Bad AI Use?

Private Family Court proceedings may be where this tension becomes most visible

The latest figures show that in more than half of private-law disposals neither party had legal representation.

That makes private children proceedings fertile ground for responsible AI assistance.

A parent may use AI because they need help understanding:

• an order;
• a Cafcass report;
• a statement;
• a chronology;
• PD12J;
• or what information a judge actually needs.

Used properly, that could improve participation.

But family litigation also creates unusually high incentives to overproduce.

The parties often possess:

• years of messages;
• emails;
• photos;
• school information;
• medical material;
• audio recordings;
• screenshots;
• and competing narratives about relationships.

AI can search all of it.

That does not mean all of it belongs before the court.

The ability to find 300 possible examples does not mean the judge needs 300 examples.

The real value may be using AI to identify the ten which best illustrate the relevant issue — and then verifying them properly.

Family procedure already contains the right principle

FPR Part 1 requires the Family Court to deal with cases justly, expeditiously and fairly, proportionately to the nature, importance and complexity of the issues.

It also requires appropriate allocation of finite court resources.

Active case management includes:

• identifying issues early;
• deciding which issues require full investigation;
• considering whether the likely benefit of a step justifies its cost;
• using technology;
• and ensuring that proceedings progress efficiently.

AI does not make those principles obsolete.

It makes them more important.

AI can help you organise a case. It should not make the court’s job harder.

JSH Law supports litigants in person with evidence organisation, chronologies, statements, schedules, Cafcass material, appeals and hearing preparation.

If AI has helped produce a large amount of material, the next task may be reducing it to what actually assists the court.

The objective is not to show the judge everything AI found. It is to make the relevant evidence easier to decide.

Book a consultation

Could AI help courts cope with the volume?

Potentially, yes.

There are obvious administrative uses.

AI could help:

• identify missing documents;
• detect scheduling conflicts;
• extract deadlines;
• categorise incoming material;
• identify duplication;
• support legal research;
• and surface information for human review.

That could release valuable court time.

But the closer the system moves towards prioritisation or substantive evaluation, the harder the questions become.

Automated triage is not a neutral administrative function

Suppose an AI system helps identify which cases require urgent judicial attention.

What factors does it use?

Length of delay?

Domestic-abuse allegations?

Child age?

Risk terminology?

Whether a document has been labelled “urgent”?

Historical outcomes?

Every criterion distributes scarce judicial attention.

That means the system requires:

• transparent purpose;
• appropriate data;
• bias testing;
• human override;
• auditability;
• security;
• and accountability.

AI can assist administration.

The more it influences whose case receives attention, the more clearly that influence should be governed.

What would a sustainable AI-enabled justice system look like?

I think it requires four layers.

The purpose of technology at each layer should be different.

Confusing those purposes creates risk.

The answer is not to make access difficult again

If AI enables somebody with a legitimate claim to understand their rights, that is not a problem to be solved.

If it helps a parent organise evidence properly instead of arriving at court with an incoherent bundle, that is useful.

If it helps a small business pursue a debt which would previously have cost more to recover than the debt itself, that may be genuine access to justice.

The objective should not be to restore scarcity simply because the court system was designed around scarcity.

The challenge is to prevent cheap generation from overwhelming expensive judgment.

The JSH Law Court-Ready AI test

Before filing AI-assisted material, ask:

The aim of good AI use should be to reduce the court’s cognitive burden.

Not increase it.

The real test of AI and access to justice

We should want a justice system in which ordinary people can understand and enforce their rights.

Technology can help build that.

But access to drafting is not the same thing as access to justice.

Justice still requires:

• independent decision-makers;
• procedural fairness;
• time to understand evidence;
• effective participation;
• reliable law;
• and sufficient institutional capacity to reach reasoned decisions.

AI can multiply words.

It can multiply claims.

It can multiply responses.

It cannot multiply judicial attention without limits.

The next access-to-justice problem may not be whether people can produce legal material. It may be whether the justice system has enough capacity to hear them once they can.

That is why the AI transition cannot be designed only around lawyers and litigants.

The courts have to be part of the architecture too.

Related JSH Law analysis

• AI Slop, Costs and Litigants in Person
• Delay Is Not Administrative
• The Jury Trial U-Turn Is Not the End of Court Reform
• AI Can Write 300 Pages. That Does Not Make It Advocacy.
• The JSH Law Six-Question Check

Sources and further reading

• Clio EMEA AI Summit 2026 — event presentation
• Ministry of Justice — Civil Justice Statistics Quarterly, April to June 2026.
• Ministry of Justice — Family Court Statistics Quarterly, April to June 2026.
• Ministry of Justice — Criminal Court Statistics Quarterly, April to June 2026.
• Courts and Tribunals Judiciary — Artificial Intelligence Judicial Guidance.
• Ministry of Justice — AI Technology Ambition for the Justice System, June 2026.
• UK Government — Legal Services Advisory AI Growth Lab.