Criminal law is a hard environment in which to fake utility.
The work is too urgent, the material too uneven, and the consequences too serious. A criminal lawyer is rarely dealing with a clean research question in isolation. More often, they are moving between police facts, criminal histories, witness statements, hearing transcripts, medical records, client instructions, sentencing material, practice notes, bench books, subpoena schedules and the immediate pressure of the next court date.
That is why the recent uptake of Habeas by criminal lawyers has been particularly interesting. The use cases are not confined to research memos or polished drafting exercises. They sit inside the ordinary machinery of criminal practice: preparing for hearings, testing the prosecution case, reviewing transcript evidence, working through sentencing material, narrowing subpoenas, checking procedural changes and turning large reference documents into usable working tools.
One criminal lawyer put it very simply:
“This thing is immensely powerful. We literally use it for almost everything we can.”
Most criminal matters are shaped long before the advocate stands up in court.
They are shaped in the reading of the brief, the comparison of one account against another, the small inconsistency in a timeline, the unsupported assumption in the facts, the line of cross-examination that should have been put, the medical report that cuts both ways, the prior version of a practice note, the sentencing material that needs to be organised into something coherent.
A lawyer may have the skill to narrow down the key issues, but not the luxury of unlimited time to search every source, compare every version, or re-read every transcript with the same level of attention. Criminal practice often demands a standard of preparation that the economics of the matter do not comfortably support.
Habeas is being used actively in that space everyday: as a way to bring more of the material into view, earlier, and in a more structured way.
A practitioner can upload the brief, as well as any key external resources, then interrogate the documents as a working set. Where are the contradictions? What has changed between the statement and the oral evidence? What was not put in cross-examination? Which parts of the prosecution case are supported by evidence, and which parts are assertion? What further instructions are needed?
Those are practical questions. They are the questions lawyers actually ask while preparing a matter.
One of the strongest criminal law use cases is cross-examination preparation.
A hearing transcript can be useful and difficult at the same time. It contains the detail, but the detail is not always easy to work with. The relevant contradiction may be buried across a statement, an answer in examination-in-chief, a concession under cross-examination and a factual assumption in the crown's material.
Habeas allows a lawyer to query those materials together. It can identify inconsistencies between a witness statement and transcript evidence, draw attention to possible gaps in questioning, summarise what a witness did and did not accept, and help the practitioner organise the material around the forensic issues in dispute.
Additionally, cross-examination is rarely improved by volume alone. The useful work is selection. Which contradiction matters? Which issue is peripheral? Which point supports the defence theory? Which question should be left alone? Which answer requires further instructions?
The lawyer remains responsible for those decisions. Habeas improves the working surface on which those decisions are made and enhances visibility across the brief.
Criminal lawyers are also using Habeas to test the structure of the prosecution case.
That can involve mapping the alleged facts against the elements of the offence, identifying where the brief supports each element, and locating weak points in the evidentiary chain. It can also involve testing credibility issues, inconsistencies between witnesses, gaps in police facts, or factual propositions that appear to have been assumed rather than proved.
This is a different task from summarising the brief. A summary may tell the lawyer what the material says. Case testing asks whether the material does the work the prosecution needs it to do.
That distinction is critical in criminal practice. A brief may look substantial because it is large. It may contain many documents, multiple statements and a detailed set of police facts. But the size of the brief does not answer the forensic question. The issue is whether the evidence proves what must be proved, and whether the case has weaknesses that should be pressed.
Habeas is useful in this regard, because it helps the practitioner move from volume to structure.
Sentencing preparation is another example use case. A sentencing matter often requires the lawyer to work across competing kinds of material: the objective seriousness of the offending, the client’s history, prior record, plea timing, rehabilitation, treatment, medical evidence, psychological material, references, comparable cases, legislative provisions and sentencing statistics or guidance.
The work is both technical and human. The lawyer has to understand the applicable sentencing framework while also presenting the person before the court with accuracy and care.
Habeas can help organise that material and apply external material from databases like the JIRS or the Law Codes, and apply that to the facts of the matter. It can summarise medical or psychological records, identify key themes, flag internal inconsistencies, extract relevant chronology, and assist with structuring submissions around the matters the court is likely to consider.
Usually, a practitioner might export a PDF version of a particular benchbook (e.g. from JIRS), upload that to Habeas, and use it as a searchable resource for a very particular legal query and fact scenario. The work is powerful and on-point immediately, especially with an effective prompt from the lawyer that cuts to the heart of the matter.
This is especially valuable where the material is dense. The first pass through the material becomes faster, more systematic and easier to check.
Criminal lawyers also work with large reference materials that are indispensable but not always convenient to use. Bench books, procedural guides, sentencing materials, practice notes and court resources often run to hundreds of pages. They are written for accuracy and completeness, not always for rapid use in a busy practice. A lawyer may know the answer is somewhere in the material but still lose valuable time finding the relevant passage.
Habeas allows these documents to become working references. A practitioner can upload a large bench book or practice note and ask direct questions of it through the search engine, or the agentic chat assistants. They can search for a direction, compare procedural changes between versions, locate the relevant sentencing discussion, or pull together the current position on a recurring issue.
Overall, we are witnessing a significant shift in day-to-day utility and the very way lawyers interact with legal resources in an AI-native search era. The document stops being a static PDF sitting in a folder and becomes something the lawyer can interrogate. It becomes a living, searchable resource.
For criminal practice, that matters across bail, sentencing, Local Court procedure, trial directions, domestic violence proceedings, subpoena issues, mental health applications and other recurring areas where the answer is often contained in a large body of guidance but difficult to extract quickly.
Some of the most valuable use cases are also the least glamorous. Subpoena schedules are a good example. A proposed subpoena can be too broad, poorly targeted or vulnerable to objection as a fishing expedition. The drafting task requires more than clean language. It requires the request to be tied to a proper forensic purpose.
Habeas can review a proposed schedule, identify categories that appear too wide, suggest narrower drafting, and help the lawyer articulate the connection between the documents sought and the issues in the matter.
In practical terms, a better subpoena schedule can save time, reduce objections and sharpen the evidentiary purpose of the request. A poor one can create delay and distract from the substance of the case.
The same applies to hearing summaries, file notes, chronologies and internal preparation documents. These documents are part of the architecture of good criminal practice. When they are produced with better structure, the whole matter becomes easier to manage.
However, we do note that for chronologies and especially for the kinds of matters that involve 10s of 1000s of files, there are specialist tools catered to this one category of work (from providers like Automatise, Mary Technology and Relativity). For lawyers only interested in chronologies, additional software is worth review.
The legally aided context is particularly important and an interesting reflection of how tools like Habeas in the custom legal AI space are changing the realities of practice. Many criminal lawyers are doing serious, skilled work within tight fee structures. The preparation required for a matter may be substantial, but the funding available may not reflect the actual time needed to do that work comfortably. As we all know, the fee structure for legal aid matters is substantially different compared to other matters, and this leads to key choices in what type of work is taken on and time allocated.
That is not a minor administrative inconvenience. It affects what is realistically possible in practice. A lawyer may want to review the transcript more deeply, compare the brief more carefully, prepare a more detailed chronology, interrogate the medical material more closely, or test an issue against the authorities. The limiting factor is often not ability. It is time.
Legal AI has a meaningful role here because it can reduce the mechanical burden of preparation. Searching, summarising, comparing, structuring and first-pass issue spotting can all be made faster. That gives the lawyer more room to exercise judgment where it matters.
For everyday people who depend on legally aided representation, that is not an abstract productivity gain. It can affect the quality of preparation that is practically achievable within the constraints of the system. We have also heard directly that lawyers are more incentivised to take on work in this domain, and we're seeing this across all sorts of practice areas. For example, litigators have noted they're more likely to take on matters in small claims court, to help out clients they may have previously not taken on before the adoption of Habeas.
Long-term, the hope is that this can have a substantial impact on access to justice in the Australian legal ecosystem as more lawyers figure out and begin to adopt Habeas in their respective practices.
Criminal law has a particular rhythm. Bail, sentencing, hearings, cross-examination, police facts, medical material, practice notes, subpoenas, bench books and urgent instructions all sit close together. The work does not divide neatly into “research”, “drafting” and “document review”.
That has obvious consequences for product design. A legal AI platform used in criminal practice has to handle research and documents together. It has to support fast movement between authorities, facts and draft outputs. It has to let the lawyer check the source material. It has to be useful inside a matter, not merely impressive in a demonstration.
This is one reason criminal lawyers have become such an important user segment for us to learn from.
The long-term product direction of Habeas continues to be shaped by that reality: better matter-based workflows, stronger document analysis, more useful treatment of large reference materials, and drafting support that reflects how lawyers actually prepare cases.
There is a temptation in legal technology to describe AI in grand terms. Criminal practice is a useful corrective. The work brings the conversation back to discipline, judgment and responsibility.
While Habeas is incredibly useful for ideation on strategy, a criminal lawyer cannot outsource the final declarative judgement calls and there is always a strong ethical dimension to any decision made. They cannot outsource the decision to put a proposition to a witness, concede a point, press an objection, run an application or make a submission affecting another person’s liberty.
