Pragmatic AI consulting for NZ legal practices, from boutique firms to mid-sized regional and metro practices. We focus on document-heavy workflows where AI assistance plus lawyer signoff genuinely moves the needle: contract review, due diligence, research, and chronology drafting.
The NZ legal services market is in the middle of a quieter version of the same structural shift hitting accounting. Volume pressure is up, the obvious tasks are getting more commoditised, clients are price-sensitive on the commodity work, and a small number of firms are visibly using AI to win on turnaround. Five specific pressures we see in practice.
For most firms with a commercial or property group, contract review is bread-and-butter work. The volume of standard SME contracts (supplier agreements, service contracts, lease reviews, shareholder agreements) is up year-on-year for most practices we speak to. Senior partners are absorbing too many "quick reviews" that aren't actually quick, while junior associates are taking too long on agreement types they haven't seen often.
On M&A and significant property transactions, due diligence is one of the most time-pressured, document-heavy parts of the practice. AI assistance on first-pass document review, redline comparison, and risk flagging can compress timelines meaningfully, when the workflow is designed correctly with appropriate confidentiality handling.
Hours of unbillable or under-billable research time, particularly on questions that are well-documented in NZ case law or statute, can be compressed substantially with AI research summarisation paired with the lawyer's own follow-up reading. The lawyer still does the analysis. The AI accelerates the "what's already on the record" step.
For practices with a litigation function, electronic discovery volumes have grown steadily. AI-assisted document review on discovery (privilege screening, relevance ranking, chronology drafting) is now mainstream in larger metro firms and is increasingly viable for mid-sized regional practices using off-the-shelf tooling.
The Law Society has signalled increasing attention to professional supervision when AI is involved. The substance hasn't changed (the lawyer is responsible for the work product, full stop), but the documentation expectations are tightening. Firms with a written AI policy, a documented workflow, and a clear audit trail are in a much better position than firms relying on individual lawyers' judgement about what's appropriate.
Most legal practice engagements start with a Workflow Integration on a single high-value task (typically contract review or document discovery), because the impact is fast and measurable. Broader audits often follow once the practice has seen what a well-designed AI workflow looks like.
Two-week structured diagnostic, scoped to legal practice patterns. ROI mapping for contract review, due diligence, research, chronology, and client communication.
Structured evaluation of legal-AI vendors and horizontal alternatives. We've tested most of the products pitched to NZ practices. Live pilot, TCO modelling, vendor contract review.
Embedded delivery on one or two document-heavy workflows. Most common first target: standard SME contract review. Lawyer signoff designed into every output step.
Privacy Act 2020, Law Society conduct rules, copyright on training data, PI insurance considerations. Acceptable use policy drafted, PI insurer briefed.
AI use in a legal practice sits across overlapping frameworks. Most are addressed by good workflow design rather than by avoiding AI altogether.
Client information is both personal information under the Privacy Act and subject to the lawyer's professional confidentiality obligation. Two practical implications: vendor data residency (NZ or AU strongly preferred), and contractual terms that prohibit any use of client data for training future models. Public AI chat interfaces are categorically not suitable for client-identifying information without anonymisation. We help firms pick tools that handle both cleanly.
The Law Society's conduct rules require lawyers to act with professional competence, confidentiality, and care, and to supervise work product. None of those rules prohibit AI assistance. They do require that the responsible lawyer maintain judgment over the output, that supervision is documented, and that any junior staff using AI-assisted workflows are appropriately supervised. Our workflow designs put the lawyer's signoff at every output step, with the AI draft archived as supporting material rather than as the deliverable.
There are two distinct copyright issues, and they should be addressed separately. First, whether the underlying model was trained on copyright material without licence (a vendor-level question, currently being litigated in several jurisdictions). Second, whether client work product sent into the tool might be retained and used to train future models, which is contractually addressable and we strongly recommend prohibiting in your vendor terms. The second question is much more directly under the firm's control.
Increasingly best practice and we recommend it. The clause we typically draft for engagement letters covers: categories of AI tools used by the firm, protection of client data, the fact that lawyer professional judgment remains the lawyer's responsibility, and a contact point for client questions. Most clients accept the clause without comment. The handful who ask follow-up questions tend to be reassured by the documented approach.
Most NZ PI insurers are now actively asking about AI use as part of renewal. Carrier expectations vary, but the common themes are: documented workflow, named-lawyer signoff, audit trail, written acceptable use policy, and disclosure to clients. A Risk + Compliance Review delivers exactly the documentation packet PI insurers prefer to see. We've briefed practices' PI insurers directly as part of engagement closeout.
For an AI-assisted contract review, a Law Society-aligned audit trail looks like: client matter file with the source contract, AI-generated first-pass review archived alongside, named lawyer's reviewed and edited markup as the final deliverable, with a brief sign-off note recording the lawyer's review and any judgment calls. None of this is structurally different from how a typical firm already documents the human-only equivalent.
An 18-lawyer mid-sized NZ practice was drowning in contract review work. We rebuilt the workflow around AI-assisted first-pass review, with a lawyer signoff and certification step. Nine lawyers were trained and certified, ready for production use in 7 weeks.
Per-contract average time dropped from 42 minutes to 8 (81% reduction). Error rate fell 67%. Reviewer weekly capacity rose 340%. The firm chose to absorb the gains into higher-value advisory work and better partner work-life balance, not into headcount reduction.
Read the full case study →Illustrative engagement · real client outcomes published with consent only
Yes, when properly designed. The Law Society's conduct rules require lawyers to act with professional competence, confidentiality, and care, and to supervise work product. None of those rules prohibit AI assistance. They do require that the responsible lawyer maintain judgment over the output, and that supervision is documented. Our workflow designs put the lawyer's signoff at every output step, with the AI draft archived as supporting material rather than as the deliverable.
Client confidentiality is the floor. We help firms pick tools with NZ/AU data residency, contractual processing terms appropriate for legal work, and per-matter prompt isolation to prevent cross-contamination. Public AI chat interfaces are categorically not suitable for client-identifying information without anonymisation. The Privacy Act 2020 sets the baseline; the lawyer's professional obligation sits above it.
Three workflows tend to be the highest ROI: standard contract review (the most consistent uplift), due diligence document review on transactions, and legal research summarisation. Each is document-heavy, partially repeatable, and benefits from AI first-pass with senior signoff. Litigation work and unique-per-matter advisory are typically lower ROI on first pass, though litigation discovery review is a strong second-wave target.
Two distinct copyright issues. First, whether the underlying model was trained on copyright material without licence, which is a vendor-level question currently being litigated in several jurisdictions. Second, whether client work product sent into a tool might be retained and used to train future models, which is contractually addressable and we strongly recommend prohibiting in your vendor terms. The second question is much more directly under the firm's control.
Disclosure is increasingly best practice and we recommend it. A brief clause in engagement letters covering AI use categories, data protection, and the fact that lawyer professional judgment remains the lawyer's responsibility. Disclosure tends to reduce client concern, not increase it. Several PI insurers now expect or prefer disclosure as part of risk management.
It works for both, with different patterns. Transactional practice tends to have higher concentrations of repeatable document work, so the ROI is faster. Litigation tends to benefit more in document review on discovery, research summarisation, and chronology drafting, but the gains are more dispersed across matters. Both are worth doing. We recommend transactional first because the payoff is faster and the use case is easier to scope.
Most NZ PI insurers are now actively asking about AI use in practice as part of renewal. Carrier expectations vary, but the common themes are: documented workflow, lawyer signoff, audit trail, written policy, and disclosure to clients. A Risk + Compliance Review delivers exactly the documentation packet PI insurers prefer to see. We've briefed practices' PI insurers directly as part of engagement closeout.
Book a 30-min discovery call. No deck, no pitch. We'll tell you, honestly, whether AI is the right intervention for your highest-volume task, or whether the answer is a better template library first.