HMCTS Is Testing an AI Case Readiness Assistant: What Could This Mean for Family Court Delays?

1. What HMCTS has actually announced — and what it has not

On 2 October 2026, HMCTS Deputy Director for Digital Delivery and AI Amar Morjaria described a pilot to begin in October at Inner London Crown Court. Its target is a familiar administrative difficulty: a case reaches the hearing stage, but a necessary task, piece of information or procedural step is still outstanding. The issue then has to be managed late, sometimes with avoidable consequences for court time.

The proposed assistant uses AI to review case information and identify and track possible outstanding actions. The information is then considered by court staff. This is an operational support system, not an automated judge.

• Location and jurisdiction: Inner London Crown Court; criminal proceedings, not private- or public-law family cases.
• Purpose: highlight tasks that may remain incomplete before a hearing.
• Decision-making: staff review any flags and decide the administrative action, if any; the tool does not make or recommend judicial decisions.
• Listing: HMCTS expressly says listing remains a judicial function.
• Status: a limited pilot; HMCTS says wider adoption depends on evidence of value and safe use.

HMCTS has not published a quantified reduction in adjournments attributable to this pilot. Nor does its announcement identify an approved Family Court version, a public interface that litigants can use, or any new duty to use AI. Claims of a proven reduction in delay would be premature.

The official blog also places the pilot within wider experiments concerning transcription, summarisation, anonymisation and search capabilities in case-management systems. The Ministry of Justice’s AI Action Plan for Justice: One Year On provides broader policy context, but an ambition to modernise is not the same thing as a completed or validated deployment.

2. Why Family Court readiness is not merely an administrative question

Family cases require far more than a neat list of completed forms. A case may be technically ‘ready’ on an administrative dashboard while a parent cannot participate safely, an allegation of coercive control has been misunderstood, a report is missing its underlying sources, or a child’s welfare is affected by months of delay.

A missing sealed order is different from a missing safeguarding assessment. A statement served late is different from a child being exposed to a serious and unresolved risk. A system that places all those things on the same simple ‘complete / incomplete’ scale may produce an appearance of progress without doing the work justice requires.

Where the family-justice consequences arise

• Private-law children proceedings: safeguarding checks, Cafcass information, disputed domestic abuse allegations, participation arrangements, and clear directions may determine whether the next hearing can usefully progress.
• Public-law children proceedings: delay can affect the child’s timescale; disclosure, evidence, expert assessments and participation obligations may need active judicial management.
• Financial remedy proceedings: incomplete disclosure, valuations, pensions material and unanswered questionnaires can prevent informed negotiation or a fair hearing.
• Litigants in person: barriers may include inaccessible digital systems, a lack of clear instructions, disability, language, financial pressure, trauma or uncertainty about who is responsible for a task.

An efficient alert system might flag that a directed report has not been filed. It cannot determine the proper evidential weight of that report, whether it contains an untested allegation, or whether its recommendations adequately address the relevant welfare factors.

This is also why the distinction between process readiness and substantive readiness matters. The first asks whether required tasks appear to be complete. The second asks whether the decision-maker has the material and participation safeguards needed to determine the actual legal questions fairly.

3. The legal framework any future Family Court tool would have to respect

The overriding objective

The Family Procedure Rules 2010, Part 1 require courts to deal with cases justly, having regard to welfare issues. That includes expedition, fairness, proportionality, equal footing and appropriate use of resources. Rule 1.4 concerns active case management, including early identification of issues, timetables and control of proceedings.

An assistant which helps a human identify an administrative omission may support those objectives. It cannot substitute a generic computer-generated definition of what matters for the court’s actual exercise of its case-management powers.

Children’s welfare

In the relevant Children Act 1989 proceedings, section 1 of the Children Act 1989 makes the child’s welfare the court’s paramount consideration when it determines the specified questions concerning upbringing. The statutory welfare framework must be applied to the individual child, rather than to a generic prediction or a machine-scored notion of case completion.

Domestic abuse and the need to avoid false equivalence

Practice Direction 12J requires the court to address domestic abuse issues in applicable child-arrangements proceedings, including identifying factual and welfare issues and considering what directions and protective arrangements are necessary. It is not enough to establish that both parties have uploaded their documents. Allegations, admissions, professional reports and judicial findings remain different categories of information.

Particularly in cases involving coercive control, the absence of an obvious ‘incident’ in a short summary is not proof that no safeguarding issue exists. Patterns of behaviour, context and the effect on the child or parent may be decisive. A narrowly trained administrative tool must not erase that distinction.

Participation, vulnerability and confidential information

The Family Procedure Rules Part 3A provide a framework for participation directions involving vulnerable persons. A case should not be marked as effectively ready without considering required interpretation, reasonable adjustments, special measures and other participation matters relevant to the individual proceedings.

Information from children proceedings may also be subject to court restrictions and the rules reflected in Practice Direction 12G. A court technology procurement or pilot needs to account for access permissions, authorised purposes, retention, security and the handling of sensitive family records.

These are existing legal and procedural considerations. The HMCTS criminal pilot has not amended them.

4. What an AI readiness assistant could flag — and what it might miss

The following comparison is JSH Law’s illustrative analysis of a possible family-law adaptation. It does not describe functions confirmed for the Inner London Crown Court pilot.

The danger of a reassuring green tick

A ‘ready’ status could give false assurance if the underlying record is incomplete. Conversely, an erroneous ‘not ready’ flag could cause unnecessary chasing, expense or pressure on a vulnerable party. Both errors matter. Accuracy must be assessed not merely by whether the tool detects missing form fields, but by whether those flags lead to better, fairer preparation without new forms of harm.

A system could even encourage the wrong incentive if success were defined solely as more cases marked ready. Staff might be pushed to clear warnings rather than investigate them. Good governance must allow an unresolved concern to remain visible, with clear ownership and an explanation.

5. Seven safeguards I would want before any family-justice rollout

1. Clear statutory and procedural scope: state precisely which administrative tasks the tool may support and which issues remain exclusively for judicial or professional judgment.
2. Source-linked flags: each alert should identify the document, order, entry and date from which it arose, allowing a responsible person to verify the original.
3. Real human challenge: staff and parties need a practical route to correct mistaken records rather than treating the machine’s output as authoritative.
4. Safeguarding-aware design: distinguish disputed allegations, recorded disclosures, findings and untested reports, and test for harms caused by omissions or misplaced confidence.
5. Equality and accessibility testing: assess whether unrepresented parties, disabled users, survivors of abuse and people with limited digital access experience worse outcomes.
6. Secure information controls: verify who may access confidential papers, where processing occurs, supplier obligations, audit logs and appropriate retention.
7. Independent impact assessment: publish meaningful evaluations of missed problems, false alerts, adjournment effects, delay, user experience and adverse incidents before wider deployment.

The judiciary’s 9 October 2026 position statement on judicial AI expresses a cautious but positive approach, emphasising that AI should support rather than replace judicial decision-making. That is the right starting point; the harder challenge is proving that oversight remains effective in busy real-world settings.

The Law Society has likewise warned in its 5 October 2026 statement on agentic AI that AI systems capable of taking actions may arrive faster than justice institutions can agree workable responsibilities and safeguards. The case-readiness pilot is a much narrower tool than autonomous legal action, but the need to establish accountability early is common to both.

6. What litigants in person should do before their next hearing

You do not have to wait for a court AI system to conduct a basic readiness audit of your own case. Nor should you assume that any court is doing it for you. The safest approach is a short, structured review using the sealed order, the material actually filed or served, and the hearing notice.

1. Start with the latest sealed order. Record every direction, the person responsible, the date and whether completion has been verified. A draft or discussion is not the same as a sealed direction.
2. Separate filing from service. For each required document, record whether it has been filed at court, served on the correct people and acknowledged where possible. Those are different actions.
3. Check what is genuinely outstanding. If a report, disclosure response or order is missing, distinguish information you have requested from information you have received and information the court has directed another person to produce.
4. Identify the purpose of the hearing. What questions has the court said it will decide? What material is needed for those questions? Avoid overwhelming the court with irrelevant material.
5. Check participation and safety. Note any protected-address concern, interpreter need, access requirement, domestic abuse participation issue or urgent welfare development, and raise it through an appropriate safe route.
6. Prepare a concise outstanding-issues schedule. For each entry, state the direction or issue, source, responsible person, deadline, status, and precise action you ask the court to consider.
7. Act before the deadline if something cannot be done. Check the correct application or correspondence route. Do not assume that writing to the court or using an AI assistant extends a direction or permits late filing.

A simple readiness register

This is an illustrative working table, not a prescribed HMCTS form or a direction to put protected information into a public spreadsheet. Store it securely and adapt it to the case.

If you are concerned about late evidence, see JSH Law’s practical guide to missed Family Court deadlines. If you are at the start of a private-law case, the C100-to-final-order roadmap explains the main procedural stages.

7. What legal teams and court administrators should consider

For legal teams, a readiness check can be a useful quality-control step: record the directions, identify who is responsible, maintain a controlled index and compare the record with what the other participants and the court actually hold. No AI purchase is necessary to introduce that discipline.

For organisations considering automated checking, start with process mapping and error analysis, not a software demonstration. Identify which administrative omissions currently produce the most avoidable delay, what evidence supports that finding, and whether the proposed solution improves the position compared with a reliable rules-based checklist.

• Governance: appoint an accountable owner for the workflow and a procedure for correcting erroneous alerts.
• Testing: include difficult examples with conflicting dates, withdrawn directions, amended orders, confidential attachments and different levels of party representation.
• Information security: avoid assuming that every document in an electronic case file may be made accessible to a new external tool.
• Staff training: ensure users can recognise when an apparently simple missing-field problem is a judicial issue or a safeguarding concern.
• Audit trail: log which document generated an alert, what action was taken and why, rather than merely recording that a warning was cleared.
• Evaluation: compare actual hearing readiness, fairness and user burden before and after introduction; record adverse effects as well as efficiency gains.

The legal profession’s responsibility for the accuracy of submissions and proper management of client information does not transfer to an AI vendor. Human checking must mean verifying the actual primary records, not simply approving a confident summary.

8. Questions the pilot still needs to answer

The public announcement establishes what the assistant is intended to do. It does not yet tell us how accurately it performs in live use, what proportion of alerts prove useful, how often significant missing information goes undetected, or whether it will reduce avoidable adjournments.

Those unanswered questions do not make the pilot improper. They are precisely why a limited, carefully evaluated pilot is preferable to pretending that a promising demonstration is proof of operational success.

A family-justice perspective

The larger policy question is not whether an AI assistant can produce an attractive checklist. It is whether the justice system can use technology to prevent predictable procedural problems without allowing an incomplete digital record to define the issues that matter.

I support intelligent administrative support where it is tested, transparent, proportionate and genuinely helpful. I do not support treating a green status, a machine summary or an apparently complete set of form fields as proof that the court has the evidence it needs to protect a child or determine a disputed allegation fairly.

A child’s welfare cannot be reduced to a missing-document alert. But equally, there is little virtue in allowing a hearing to fail because a known, fixable administrative task was never chased. Good legal technology should help us distinguish those problems, not confuse them.

9. Official sources and related JSH Law guides

Official sources reviewed

• HMCTS (2 October 2026): Testing how AI could help improve case readiness in the Crown Court
• Ministry of Justice (24 September 2026): AI Action Plan for Justice — One Year On
• Judiciary (9 October 2026): Position Statement on the Judicial Use of AI
• Family Procedure Rules Part 1: Overriding Objective
• Practice Direction 12J: Domestic Abuse and Harm
• Family Procedure Rules Part 3A: Vulnerable Persons
• Children Act 1989, section 1
• Practice Direction 12G: Communication of Information

Related JSH Law analysis

• Evidence Readiness in Family Court — the existing practical guide
• Your Family Court Case Needs a System: CRM, AI and Evidence
• AI Is Already in the Courts: Why Family Justice Needs Hard Red Lines
• Private Children Proceedings: The Roadmap from C100 to Final Order