Tech Corner: Artificial Intelligence in Financial Remedy Practice: Opportunity, Risk and Professional Responsibility

Artificial intelligence (AI) has now arrived in family law in a way that feels both gradual and sudden. Gradual, because technology has been entering practice for years through case management systems, digital bundles, online portals and increasingly capable research platforms. Sudden, because generative systems have altered the texture of daily work almost overnight. A practitioner can type a prompt and receive a coherent narrative, a neatly structured argument or a plausible sounding summary within seconds. That combination is powerful. In the hands of a busy lawyer, it can feel like relief. It can also be a trap.

Financial remedy practice sits directly in the path of this change. It is intensely human in subject matter but relentlessly structured in its mechanics. The work is driven by disclosure and deadlines: questionnaires, schedules, replies, chronologies, narrative statements, budgets, offers and draft orders. The figures must be right and the document trail must be complete. There is constant professional pressure, shaped not only by client expectation, but also by a strict regulatory and ethical framework governing accuracy, disclosure and conduct, alongside client emotion in equal measure. AI appears well suited to this environment because it can process information quickly and present it neatly. Yet the risks matter more here too, because an error is rarely just an error. It becomes part of a case theory, a negotiation position or a document placed before the court.

The judicial warning

The clearest recent judicial warning about the misuse of generative AI in legal work in England and Wales comes from the Divisional Court in Ayinde v The London Borough of Haringey; Al-Haroun v Qatar National Bank QPSC [2025] EWHC 1383 (Admin). The court did not suggest that the use of AI is inherently improper. It recognised that such tools are already being used and will continue to be used. What it did do, in unusually plain language, was explain why generative systems create a particular professional risk.

The risk is not simply that AI can be wrong. Lawyers are accustomed to checking sources. The distinctive problem is that generative systems can be coherently wrong. They can produce content that reads as though it has been researched and verified. They can make confident assertions that are entirely untrue. They can cite sources that do not exist. They can purport to quote passages from real authorities that do not appear in the actual text. This combination of plausibility and falsity is what makes the technology different from earlier forms of legal assistance.

In Ayinde, the court dealt with the inclusion of fictitious authorities in submissions. In doing so, it reinforced a fundamental principle: responsibility is personal. The duty to ensure that material placed before the court is accurate rests with the lawyer who drafts, settles and signs the document. That duty cannot be delegated to a tool. Whether the error originates in a generative system, an online summary, hurried copying and pasting or inadequate supervision, the obligation to verify remains exactly where it has always been.

The companion case, Al Haroun, demonstrates more directly how AI can contaminate court material if not properly checked. The judgment records admissions about the use of publicly available AI tools alongside conventional research platforms, and a failure to verify the resulting citations and quotations. AI can produce something that looks finished. But if that apparent saving displaces verification, it is not a saving at all. It is a professional hazard.

Why this matters in financial remedy work

At first glance, financial remedy practice might appear less exposed than commercial litigation to fabricated authorities. We do not routinely file extended skeleton arguments in interim applications. Much of our work is fact-heavy rather than law-heavy. That is precisely why the risk is subtler.

Financial remedy work depends on accurate factual matrices. Asset schedules, business valuations, pension sharing calculations and housing needs assessments are built on detailed numerical information. An incorrect date, an omitted liability or a misplaced decimal point can materially affect advice and outcomes. If generative tools are used to summarise disclosure or draft schedules without careful checking, errors can become embedded in negotiation positions.

There is also a second dimension. Generative systems do not merely fabricate authorities. They can import invented propositions into otherwise routine drafting. A draft position statement that contains a confident but unsupported assertion about the court’s approach to needs, conduct or non-matrimonial property may influence negotiation tone even if it never reaches a judge. Plausible language is not a substitute for legal accuracy.

The fabricated authorities problem

It is worth addressing directly the issue of fabricated authorities because it has now moved from anecdote to reported judicial concern.

Generative systems are capable of inventing cases that do not exist. They can provide a case name, a neutral citation, a summary of facts and a ratio that reads entirely credibly. In some instances, they will cite real cases but attribute to them propositions that cannot be found in the judgment. The result is not a typographical error. It is a fictional legal landscape.

The Divisional Court in Ayinde and Al Haroun made clear that the inclusion of non-existent authorities is a serious matter.¹ The court emphasised that counsel and solicitors are under a duty to ensure that authorities relied upon exist and stand for the propositions advanced. The fact that material may have been generated by AI does not diminish that duty.

For family practitioners, the lesson is straightforward, AI must never be treated as an authority. If a proposition matters, it must be checked against the judgment itself or an authoritative source. There is no shortcut around that obligation.

Artificial intelligence as evidence factory

If the research risk is obvious, the evidential risk may prove more disruptive in family practice.

Financial remedy cases frequently involve digital material. Screenshots of messages, social media posts, emails, downloaded transaction histories and audio recordings are routinely relied upon to support allegations about lifestyle, dissipation, cohabitation or hidden assets. Even without deepfakes, authenticity disputes are not uncommon. As generative tools improve, practitioners should assume that some parties will test the limits.

Publicly available tools can now generate convincing audio and visual material. The technology required to manipulate a recording is no longer confined to specialists. That creates practical questions for family lawyers. How confident are we about the provenance of digital material provided to us? Do we routinely request original files? Are we alive to the evidential significance of metadata? At what point should we consider forensic input?

None of this is new in principle. The Family Court has long dealt with contested authenticity. What may change is frequency. What was once exceptional may become routine. Financial remedy practitioners should expect greater scrutiny of digital evidence and be prepared to respond.

Professional responsibility does not shift

Against this background, it would be unrealistic to treat AI as something that can simply be excluded from practice. It is already part of the professional environment. It influences how clients behave, how work is carried out and what risks arise. The more constructive question is how to use it in a way that strengthens practice rather than undermines it.

The starting point must be that professional responsibility does not shift. The Solicitors Regulation Authority’s Standards and Regulations 2019 require solicitors to act with integrity, to uphold public trust and to provide a competent service. Those duties apply regardless of the tools used. AI does not dilute them. If anything, it heightens the need for active supervision and verification.

Firms should consider clear policies on the use of generative systems. Practitioners should be trained in their known failure modes, particularly the risk of hallucination and fabricated citation. Where AI is used to assist with drafting, its outputs should be treated as a first draft requiring careful review, not as a finished product.

Confidentiality and data protection also require active thought. Placing client information into publicly available systems without understanding how that information is stored or used creates obvious risk, not least the inadvertent waiving of legal advice privilege or litigation privilege. In financial remedy cases, where the information involved is often commercially sensitive and deeply personal, casual use is unlikely to be defensible.

A distinction worth drawing

It is helpful to distinguish between general generative tools and technology designed around legal workflows.

General systems are trained to produce fluent language across a vast range of topics. They are not trained to be correct in any specific jurisdiction. They do not understand professional duties. They optimise for plausibility rather than accuracy.

Technology designed specifically for legal workflows can be built differently. It can be constrained to defined processes. It can remain anchored to client supplied information and source documents. It can be transparent about what it has done. It can be designed to avoid the invention of legal authority and to support lawyers in meeting professional obligations rather than tempting them into shortcuts.

That distinction is not simply technical. It is about how technology fits within professional responsibility.

It is in that context that Nova sits.

Nova is a legal services technology platform developed in collaboration with practising family lawyers and academic partners. Nova Engage[[1]] is an AI-assisted client onboarding system designed for family law professionals. It is not intended to generate legal argument or replace professional judgement. Its focus is on improving the quality of information available at the earliest stage of a matter.

That focus is significant in financial remedy practice. The first meeting often determines the trajectory of the case. If the client arrives with only a partial understanding of their financial position, the meeting becomes an exercise in reconstruction. If key documents are missing, advice is heavily caveated. If important facts emerge later, strategy shifts and risk increases.

Structured onboarding can assist clients to assemble relevant information before the first consultation. Nova Engage guides them through the disclosure landscape in a systematic way. It can create a clear digital record of what has been provided. For the solicitor, using Nova Engage allows the first meeting to focus on analysis, options and strategy rather than basic information gathering.

There is also a risk management dimension. Early structured information can help identify jurisdictional issues, safeguarding concerns or complex asset structures requiring specialist input. It can assist with supervision and file review. It can support more accurate scoping of work and more realistic fee estimates.

Importantly, systems such as Nova Engage are designed around the information provided by the client. They are not large language models generating free form legal propositions. They do not hallucinate authorities. They operate within defined pathways intended to support, not replace, the professional.

Client expectations in the age of artificial intelligence

There is another practical reality. Many clients now consult generative tools before speaking to a solicitor. Some arrive with a confident narrative about what they are entitled to that has no foundation in law. Others rely on explanations that sound plausible but are inaccurate.

The judicial warning about plausible but incorrect content therefore applies not only to lawyers, but also to the information environment surrounding every case. Practitioners may increasingly find themselves correcting AI-generated misunderstandings at the outset of a retainer.

Structured digital engagement can help to reset expectations. By guiding clients through a process focused on their actual circumstances and documentation, it shifts the emphasis from generic answers to bespoke advice. It reinforces the message that accurate advice depends on accurate information. Used properly, it can improve the quality of the first meeting and reduce misunderstandings later.

Authenticity, supervision and governance

As AI becomes embedded in legal culture, authenticity disputes are likely to increase where digital material is central. Practitioners should consider routine steps to preserve provenance, including requesting original files and retaining metadata where relevant. Where authenticity is seriously in issue, early consideration of expert forensic input may be appropriate.

At firm level, governance matters. Supervision cannot be passive. If junior lawyers or paralegals are using generative tools, supervisors must understand how and when. Signing off work that contains AI-generated content without verification exposes both individual and firm to risk. The Divisional Court’s message was clear: the duty to check is not optional.

Opportunity alongside risk

It would be easy to end this discussion with a list of cautions. That would miss the opportunity. AI, used thoughtfully, can improve preparation, reduce administrative burden and allow lawyers to focus more fully on the human aspects of financial remedy work.

If structured systems can ensure that initial information is more complete, meetings can become more analytical and less interrogative. If routine drafting is assisted but carefully checked, time can be redeployed to negotiation and strategy. If digital engagement improves client understanding at the outset, mismatched expectations may reduce.

The profession should not be complacent. The risks identified by the Divisional Court are real. Fabricated authorities are not an abstract possibility. Synthetic evidence is not science fiction. But neither should practitioners assume that the only safe response is abstention.

Financial remedy practice is built on trust, accuracy and professional judgement. AI does not alter those foundations. It does, however, require us to be deliberate about how we build on them.

Used without verification, it will undermine practice. Used within clear professional boundaries, it can strengthen it. The responsibility for choosing which path we take rests, as it always has, with us.

[[1]]: Nova Law Limited, product information relating to Nova Engage is available at https://novaforlawyers.novalaw.co.uk/