Dazychain

AI Centrality

How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.

Artificial intelligence is present, useful and peripheral, and the vendor's own design decisions say so. The product is sold first as matter management, described in its own footer strapline as capture, triage, manage, resolve, and the pricing table shows a complete platform at every tier with AI as one section among matter management, documents, spend, collaboration, workflow, reporting and integrations. Three published facts place it here. The AI is capped by usage on the entry tier and unlimited above it, so it is priced as an add-on rather than as the mechanism. It runs only when a user engages a function by clicking a button, so nothing happens by default. And the FAQ states that AI can be completely turned off for an account, leaving a working matter management system behind. What the AI does is summarise, search, answer questions and compare a contract to a playbook, all of which sit on top of the record rather than constituting it. Checked 4 September 2026.

Citation Accuracy and Hallucination Disclosure

Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.

Accuracy is asserted with an unusually candid caveat and nothing is measured. The FAQ states that AI extractions are designed to be highly accurate but, like any automated system, may require review, and that users can validate and edit extractions directly in the platform before anything is finalised or shared. A limitations entry adds that the AI works only within the matter or document it is engaged on and that clear, well-structured and complete documents lead to better results, which is a real statement of input sensitivity that most vendors omit. Grounding scope is therefore described. What is absent is measurement of any kind: no accuracy figure, no test set, no evaluation, no error rate, and no description of how an extraction traces to the passage it came from. The one hallucination reference is not the vendor's own: the security section attributes minimisation of model hallucinations to Amazon Bedrock, which is a claim about the infrastructure rather than a published position on the product's output.

Autonomy and Oversight Model

What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.

The system does nothing on its own, and that is published as a design commitment rather than implied. The FAQ states that AI accesses data only when a user actively engages an AI function such as clicking a button, so privileged and confidential matters remain untouched until assistance is explicitly requested; that the AI works only within the matter or document it is engaged on and does not reach data outside that context; and that extractions may require review, with users validating and editing them in the platform before anything is finalised or shared. AI can also be turned off entirely for an account. That is a genuine control structure with a real review point. What holds it below the top band is that the constraints are binary rather than graduated: no confidence signal is surfaced against an individual extraction, no threshold or abstention state is described, and the FAQ question asking whether AI can be restricted by role or matter type is answered only as to switching it off for the whole account, leaving the finer-grained half of its own question unanswered.

Operational and Outcome Evidence

Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.

Named customers with named people in stated roles, and no figures. Two named legal officers are quoted with linked case studies: Helen Carr, Deputy General Counsel at MYOB, describing using Dazychain's matter data to demonstrate an increase in the volume and complexity of matters and to advocate successfully for additional headcount, and Mercia Chapman, Senior Legal Counsel at Equity Trustees, on email capture and retrieval. A Client Stories section carries further interviews and was not opened in this pass. The MYOB account is the stronger of the two because it describes a specific operational outcome that the platform's reporting made possible rather than a satisfaction sentiment. What keeps this at B is measurement: nothing is dated, no quantity is attached to either customer, and the corporate claims run to unquantified language about efficiency. This is materially better deployment evidence than the seed list predicted, which flagged the name as directory-level only.

Privilege and Confidentiality Posture

How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.

Substantive commitments across most of the ground, published on a security page rather than in an agreement. Segregation is stated plainly: data segregation is fully supported and tenant aware, and no user company can access another's information unless permission has been given, as in a collaboration. Encryption is AES-256 at rest in a MongoDB store with HTTPS and TLS 1.3 in transit, documents sit encrypted in Amazon S3, role-based access applies at every tier, and every access and every modification to documents, matters and deliverables is logged and preserved. The AI position is the strongest part and is unusual: the model reaches nothing until a user clicks, it is confined to the matter or document engaged, Amazon Bedrock does not store inputs or outputs, each model runs in a dedicated AWS account, and customer information is not used to train external models. Two things hold it below the top band. No privilege or work product treatment is named anywhere, which the top band requires as its own limb. And there is no customer agreement of any kind on the site, so every commitment here is policy rather than contract.

UPL and Professional Responsibility Posture

Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point.

Nothing published addresses the advice line, on a product that drafts documents and answers questions about them. The site inventory was taken from the navigation and footer across four pages on 4 September 2026 and contains no terms of service, no end user agreement and no disclaimer page; the only legal document linked anywhere is a privacy policy. No statement was located that outputs are not legal advice, that a lawyer must review generated content before it is relied on, or that self-service document generation by a business stakeholder carries any limitation. The gap is pointed rather than formal because of one published feature: business clients and non-lawyers complete an intake form which automatically generates a tailored document, such as a non-disclosure agreement, from a pre-approved template and can send it back without manual intervention. That is a document produced for a non-lawyer with no described legal checkpoint, and nothing published sets a boundary around it. No rule of professional conduct, bar authority or jurisdiction limit appears on any surface.

AI Governance and Bias Disclosure

Published governance over model behaviour: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.

No governance position was located. There is no responsible AI page, no principles statement, no named owner accountable for model behaviour, no pre-release evaluation regime, no testing results and nothing at all on bias. The AI FAQ is substantial and entirely operational, covering what the AI can do, where it runs, what data it uses, whether it can be turned off and what its limitations are, which is security and scope rather than governance. The security page is detailed and is likewise about information security, which the axis definition treats as a separate subject and which is graded on the stewardship and certification rows rather than counted twice here. The one adjacent artifact is an Information Security Management System with documented policies, annual staff training and documented acceptance, but that governs employee conduct rather than model behaviour. Searched the AI page, the security page, the pricing page, the features index and the site navigation on 4 September 2026.

AI Safety and Data Stewardship

Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.

Most of the set is published and specific, with incident practice the missing limb. Retention and deletion are stated together: on account termination customer data is retained for the period the customer requests, usually one month, then deleted, with alternative arrangements available on request. Access control is described at several levels, with role-based access at every pricing tier, an audit trail logging every access and every modification to documents, matters and deliverables, staff police checks on joining and every two years, signed confidentiality agreements for employees, third parties and contractors, and annual security training with documented acceptance. Encryption is AES-256 at rest and TLS 1.3 in transit. Infrastructure suppliers are named rather than gestured at: Amazon for application and file hosting and Object Rocket for the database, both in Australia and the United States. Penetration testing runs internally before each release and annually through an unnamed external agency. What is absent is any breach or incident notification commitment: nothing states whether, when or how a customer would be told, and no subprocessor register or change notice exists.

AI Liability and Recourse

What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.

Nothing published allocates loss, because no customer agreement is published at all. The site inventory taken from the navigation and footer across four pages on 4 September 2026 lists a privacy policy and a sitemap as the only legal documents; there is no terms of service, no subscription agreement, no end user terms and no service level document anywhere on the site. Consequently no indemnity, no liability cap, no warranty, no exclusion and no insurance position could be located, and a buyer cannot read the allocation of loss before signing even though the product offers a self-serve fourteen-day free trial and publishes per-user prices, which is the commercial posture of a product a buyer might purchase without negotiation. This is a documented absence rather than a retrieval failure: every page fetched rendered in full and the inventory is complete. The one adjacent statement located is a security-page assertion that regular audits keep the cloud environment secure, which is a practice claim rather than a recourse position.

Practice Systems Integration Depth

How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.

Named integrations with tier availability published, and no description of what they move. The pricing comparison table is the best source and lists them by tier: single sign-on through Entra, Okta, SCIM and ADFS, plus DocuSign, Outlook and Gmail at every tier, and API access at the top tier only. That is more useful to a buyer than a logo wall because it answers what a given plan actually includes. Two entries are marked as not yet available and are recorded as such rather than credited: SharePoint integration is listed coming soon at the top tier, and document libraries and knowledge base are coming soon at all three. What is missing for the top band is depth. Nothing states what synchronises, in which direction, on what trigger or what a legal team must configure, and no dedicated legal document management system is live, so a department whose matter files sit in a DMS today has no published path. An integrations page exists and was not opened in this pass.

Deployment Model and Data Residency

Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.

Residency is published concretely and tenancy is described only in outline. Locations are given for each layer rather than for the platform as a whole: application data in Amazon data centres in Australia and the United States, the database with Object Rocket in Australia and the United States, and documents in Amazon S3 in the same two regions, with multi-availability zone deployment for redundancy. Processing is addressed separately from storage, which is rare in this lane: the AI runs on Amazon Bedrock infrastructure in Australia or the United States depending on the customer's location, keeping data within the designated geographic boundary, with each model in a dedicated AWS account. Tenancy is described functionally rather than architecturally, with segregation stated to be fully supported and tenant aware, which implies a shared platform with logical separation but never says so. No single-tenant, dedicated or self-hosted option is offered, no region choice is presented to the buyer as a selectable option, and nothing changes between the three published tiers on deployment.

Security Certifications and Trust Center

Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.

Two certifications are stated with an audit cadence, and the evidence behind them is not reachable. ISO 27001 controls and the wider Information Security Management System are stated to be internally and externally audited annually, and SOC 2 internal and external audits are said to be undertaken annually with reports produced. That is more than a badge wall and it is why this sits here. Four things hold it below the top band. No auditor or certifying body is named for either. No SOC 2 type is stated, so a reader cannot tell whether the report is a point-in-time Type 1 or a period Type 2. No report date, period or scope statement appears, and no route to obtain either report is published, with no trust portal and no request form. And the ISO claim names the superseded edition, citing ISO 27001:2013 on a page last modified 21 April 2026, after the October 2025 deadline for transition to ISO 27001:2022. Recorded because a reader who checks the standard will find it withdrawn. The AWS certifications listed alongside, including FedRAMP and SOC 1, are correctly attributed by the vendor to AWS rather than claimed, and are not credited here.

Model Supply Chain Disclosure

Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.

The architecture is described in useful detail and the models and their makers are not identified. What is published is real: all AI tools run on Amazon Bedrock, each model runs in a dedicated AWS account, Bedrock does not store inputs or outputs, does not share them with third parties and does not use them for training, and inference runs in Australia or the United States according to the customer's location. A buyer can therefore establish where the model runs, on what service and under what data conditions. What is not published is whose model it is. Bedrock hosts models from several makers and none is named, no model or version is identified, and nothing states which model handles a summary as against a playbook comparison. No commitment to notify customers if the model, version or hosting arrangement changes was located. Amazon appears here as the inference service operator rather than as infrastructure, and its separate role hosting the application and database is graded on the deployment row rather than counted twice.

Commercial Transparency

Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.

A buyer can price this completely without speaking to anyone, which is rare in this lane. Three tiers are published with rates: Launch at 105 per user per month, Scale at 130 and Optimize at 160, with a currency selector offering Australian dollars, US dollars, euros and pounds. The unit is stated plainly as per user per month. What implementation adds is published rather than withheld: no implementation fee on Launch, optional customised configuration and implementation at 5,000 on Scale, and the same 5,000 as a standard inclusion on Optimize, with reconfiguration priced as a professional services fee on the lower tiers and one annual reconfiguration included at the top. A full feature comparison table runs to roughly forty rows across matter management, documents, spend, collaboration, AI, workflow, reporting, integrations, security and support, marking each as included, excluded, quantified or coming soon. AI itself is tiered transparently, with usage caps on Launch and unlimited access above it. A fourteen-day free trial is offered with self-serve signup.

Firm and Practice Coverage

Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.

The buyer is described with substance and the practice boundary is left open. The segment is stated repeatedly and consistently: in-house corporate legal departments, with the product characterised in the vendor's own strapline as a simple end-to-end matter management tool for smaller corporate legal teams, and the pricing tiers structured per user so that a team can size itself. Named customers support the claim at the mid-market and large end, with MYOB and Equity Trustees both Australian listed or substantial institutions. Coverage extends beyond legal in a way the record should carry: the same platform is sold to HR teams under a separate product line, and the vendor describes a corporate front door spanning legal, HR and finance. Practice area does not bite on a matter management product and is named rather than penalised, since the system is matter-type agnostic. What is missing is the limit: no jurisdictional coverage is stated despite Australian and United States hosting, no minimum or maximum team size is given, and nothing identifies work the product is not suited to.