Product and intended use
Define the configuration, people, information and tasks being assessed.
Product and plan covered
Public claim only
The product page states Intapp Assist is 'Currently available in Intapp Time, Intapp Terms, and Intapp DealCloud'. The sub-processor list treats Intapp Assist as a separately listed product with its own sub-processor section, while also listing it in the processing description for other products, for example 'Processing services for artificial intelligence features (e.g. text generation), such as Intapp Assist' under Intapp Time. Nothing read states whether it is a separate licence.
Ask: On our Order and Sale Agreement, is Intapp Assist a separate line item, and in which of our Intapp products is it switched on?
AI functions published
Public claim only
The product page (read 2026-09-27) describes five functions: tagging structured and unstructured content with extracted entities, concepts and sentiments, published as 'zero entry'; answering conversational queries in natural language by analysing firm-specific data; summarising notifications, reports and contracts; drafting correspondence and narrative descriptions; and recommending next steps, relationship signals and company recommendations. For Intapp Terms the implementation is published as Ask Intapp inside Microsoft Teams, returning answers with 'direct links to the underlying source information'.
Ask: Demonstrate each function on our own data in the products we licence, and confirm which can be disabled per user and per practice group.
Deployment and processing architecture
Public claim only
The Applied AI page states client data is first processed in the client's own tenant environment to build the prompt, that 'Private models live alongside a client's own tenant in their regional cluster', and that 'Shared models are deployed among our global clusters as their required infrastructure is supported in different regions.' It also states 'Client data is processed within Microsoft's Azure infrastructure and other infrastructures as applicable' and that each request carries a unique inference ID retained through the request lifecycle so data remains segregated between clients. No on-premises or self-hosted option was found.
Ask: For our tenant, which regional cluster holds it, and which of the Assist functions run on shared models rather than in our cluster?
Jurisdictions and legal content
Public claim only
Intapp Assist is published as answering from firm data rather than from published law; no jurisdictional legal content, court coverage or citator is described on the pages read. The product is published for 'professional and financial services' firms generally, and the Intapp Terms implementation is published as answering questions about the firm's own client terms and obligations.
Ask: Does Assist carry any jurisdiction-specific content or rules for Australia, or is it entirely confined to our own data?
Vendor content directing how AI describes the product
Not confirmed
Checked deliberately on 2026-09-27. None of the eight Intapp pages read contains a block addressed to AI models, assistants or crawlers, and two earlier site-map operations over intapp.com, one searched for 'llms.txt', 'information for AI assistants' and 'ai-info', returned no such page. Recorded as not found rather than absent, because a search-filtered site map cannot prove a page does not exist.
Ask: Do you publish any page or file, including an llms.txt, that instructs AI models or crawlers how to describe Intapp or Intapp Assist?
Data, controls and requirements
Compare the actual configuration with firm, client and jurisdiction requirements, supported by evidence.
Client authority and responsibility for use
Public claim only
The Applied AI page states 'Intapp Applied AI does not act or make decisions — the user is ultimately responsible for the action taken or decision made', and the cloud-security page states 'Clients are responsible for all customer data they upload and process in the products.' Neither is a contractual clause, and no clause on professional obligations, client consent or disclosure to clients was found: the master terms and the body of the data processing addendum were not read.
Ask: Which clause of our agreement allocates responsibility for client consent and for disclosing AI use to our clients, and what notice wording do you recommend?
Foundation-model providers
Public claim only
The published sub-processor list (effective 11 August 2026) has a section headed 'Intapp Assist' with a single entry: Microsoft Azure AI Services, 'Processing services for artificial intelligence features within Intapp Assist such as text generation and summarization', location 'United States, EU*'. At platform level the Applied AI page names 'Microsoft Azure OpenAI GPT-3.5' as an example of a pre-trained LLM in use, and Microsoft Azure AI Content Safety for moderation. Anthropic PBC and Amazon AWS (Bedrock) appear elsewhere in the same list against other products, including Intapp Celeste, but not against Intapp Assist.
Ask: Name the specific Azure OpenAI model and version behind each Assist function on our tenant, and the notice period before it is changed.
Sub-processor register and change notification
Public claim only
Intapp publishes a per-product sub-processor register, effective 11 August 2026, with columns for sub-processor, processing activities, locations and additional details, plus a separate 'Professional Services Providers' section and prior versions back to February 2024 with their effective date ranges. The page states: 'As Intapp's business grows and evolves, the Sub-processors we engage may also change. We will post such updates here, in accordance with the terms of the Agreement. Please subscribe to receive notifications of updates to this page and also check back frequently for updates.' Footnote 1 defines 'Intapp Affiliates' as group companies providing support, maintenance, professional services and DevOps. Footnote 2 states that where a customer uses Celeste functionality along with another product, that product's sub-processors also include the Celeste sub-processors.
Ask: What notice and objection rights does our agreement give us when a sub-processor is added, and will you notify us in writing rather than by page subscription?
Training on customer content
Public claim only
The Applied AI page states client data 'is data uploaded by a client into Intapp cloud services. With client permission, this data is used to deliver private, client-specific AI models.' It distinguishes shared models, built on Intapp proprietary and client-agnostic data or on third-party services, from private models 'further fine-tuned using client-specific data'. It also states 'Intapp uses models directly trained with client data only to process or make inferences on the client's own data' and, asked whether a client can obtain a copy of a model trained on its data, 'No, fine-tuned models can only be used with Intapp products.' These are FAQ answers on a vendor page, not contractual terms, and the form of the client permission is not published.
Ask: How is the client permission for fine-tuning given and withdrawn, is it in our agreement, and can we use Assist with shared models only?
Feedback and telemetry
Public claim only
The Applied AI page states: 'Intapp does this only for private models, in which we use individual tenant feedback and data to incrementally train the new private version of the model for the same tenant. Common models are upgraded with data produced by Intapp, including from vendors and data providers.' It also states application data — 'all the metadata-associated data and activity, along with any additional features and metrics that can be derived from it' — is collected and stored in Intapp infrastructure to monitor system performance and user satisfaction, and that data fitting the application-data definition 'can be used by automated benchmarking and analytic processes to track algorithm performance and reliability'.
Ask: What exactly falls inside 'application data' for Assist, where is it stored, and can we opt out of its use for benchmarking?
Retention of prompts, outputs and logs
Public claim only
The Applied AI page states shared-model AI services 'are stateless and therefore designed to retain no data', that responses and the inputs used to generate them are transferred back to the initiating product and, 'when appropriate, stored with client data', and that 'User interactions and AI results are also monitored and stored in client-specific databases.' The cloud-security page answers a data-control question by directing the reader to 'our customer data retention policy for more details'. That retention policy was not read and no period is published on the pages read.
Ask: Provide the customer data retention policy and state the default and configurable retention for Assist prompts, outputs and interaction logs.
Deletion and certified erasure
Public claim only
The Applied AI page states: 'When a client's entire dataset is deleted, Intapp will delete the fine-tuned model (which can include client data).' No deletion period, no process for deleting individual matters or prompts, and no certificate of deletion is published. The data processing addendum index shows a current version effective 1 November 2025, but its body did not render on the page read, so the contractual deletion and return provisions were not established.
Ask: On exit, what is deleted, in what period, and will you certify in writing the deletion of prompts, outputs, embeddings and any fine-tuned model?
Vendor access to content
Public claim only
The cloud-security page states: 'Intapp employees do not access customer data unless they are directed to do so by the client.' The Applied AI page states client data 'is stored in each Client's production environment where access is restricted to staff responsible for maintaining our services.' The published sub-processor list places 'Intapp Affiliates' providing support, maintenance, professional services and DevOps in Australia, Canada, Ireland, Singapore, Ukraine, the United Kingdom and the United States, and names third-party support and development providers against other products. A Tenant Access Policy is published at intapp.com/cloud/policies/tenant-access-policy/ and was not read.
Ask: Provide the Tenant Access Policy and confirm which countries support staff can read our data from, under what approval, and whether we receive the access log.
Authentication, single sign-on and audit
Not confirmed
Nothing on single sign-on, multi-factor authentication, role-based restriction of AI features or audit-log export was found on the eight pages read. The Applied AI page states only that 'Our models are only accessible through internal product APIs. The service behind the API will ignore any unauthorized requests.' Intapp publishes a separate ethical-walls product and a Tenant Access Policy, neither of which was read.
Ask: Confirm single sign-on and enforced multi-factor authentication for Assist, whether ethical walls constrain what Assist can retrieve, and whether we can export an audit log of prompts and answers.
Encryption and key management
Public claim only
The cloud-security page states: 'All cloud products use disk encryption (AES-256) at rest. All data in transit for external communications is HTTPS traffic-encrypted using TLS 1.2 or better. Traffic inside our VPC is encrypted based on the service-specific protocols. Keys are stored in FIPS-140-2 compliance key stores with Azure. Intapp maintains internal cryptographic use polices.' The Applied AI page adds that 'Client data is encrypted in transit throughout these steps' of AI processing. No customer-managed key option is published.
Ask: Are customer-managed or customer-held keys available for an Australian tenant, and is the AI call path to Azure AI Services covered by the same key arrangement?
Independent security assurance
Public claim only
The cloud-compliance page publishes, with links to certificates, ISO 27001, ISO 27017, ISO 27018 and ISO 27701 certification via Schellman, CSA STAR certification with a registry entry for Integration Appliance, Inc., SOC 1 and SOC 2, a Global Privacy Recognition for Processors certification, a Standard SIG 2024 Lite questionnaire, and GovRAMP membership described as 'actively pursuing a verified offering status'. The page does not state the SOC report type or period. The cloud-security page states 'Penetration testing is completed at least annually by a CREST-approved third party. Executive summaries of penetration tests will be released upon signed NDA from the client', with monthly network scanning.
Ask: Provide the current SOC 2 report with its type, period and exceptions, and confirm that the ISO and SOC scopes cover Intapp Assist and the Azure AI Services processing path.
Contracting entity, liability and forum
Not confirmed
No contracting entity, liability cap, indemnity or governing law for an Australian customer was established. The pages read refer to an 'Order and Sale Agreement' and to 'the Agreement' without publishing either; the terms page at intapp.com/terms was discovered but not read; and the data processing addendum page rendered only its effective date, 1 November 2025, and a list of prior versions rather than the clauses. The published group-members list names 'Intapp Pty Limited | Australia | Subsidiary', and the site copyright notice attributes the Intapp trade mark to Integration Appliance, Inc. or its subsidiaries. Neither statement identifies the counterparty for an Australian firm, and no ACN or ABN is published.
Ask: Which entity contracts with an Australian firm, what is its ACN, what law and forum govern, and what is the liability cap for an AI-related claim?
Cross-border transfers and where inference runs
Public claim only
The sub-processor list section for Intapp Assist gives one sub-processor, Microsoft Azure AI Services, location 'United States, EU*', with the note: 'Location is dependent on the hosting location of the Customer, However, Customers that are located outside the US and EU will be processed in the EU'. The same note appears against Azure AI Services for every other Intapp product in the list. The Applied AI page states the same thing in prose: 'For non-U.S. regions that do not yet support Intapp Applied AI's needed infrastructure, calls are routed to the EU infrastructure.' On the vendor's own published list, therefore, generative-AI processing for an Australian customer takes place in the European Union, even though the same list gives Australia as one of the Microsoft Azure hosting locations for the underlying products.
Ask: Confirm in writing that Assist inference for our tenant runs in the European Union, name the country and the transfer mechanism relied on, and say whether Australian or Singapore processing is on your roadmap with a date.
Incident notification
Not confirmed
No incident-notification period, breach-reporting commitment or incident history was found on the eight pages read. The cloud-security and cloud-compliance pages address certification, encryption and penetration testing but not notification. The body of the data processing addendum, where such a clause would normally sit, did not render on the page read. A cloud status site is published at status.my.intapp.com and was not read.
Ask: State the contractual notification period from discovery of a breach, whether it covers incidents at Microsoft Azure AI Services, and any material incident in the last 24 months.
Change notice for models, processors and terms
Public claim only
The sub-processor page states updates will be posted there 'in accordance with the terms of the Agreement' and invites subscription to notifications; prior versions from February 2024 onward are published with effective date ranges, as are prior versions of the data processing addendum from November 2020. On model change, the Applied AI page states: 'From time to time, Intapp also conducts comparative analysis of existing solutions against new technologies, algorithms, and platforms. This can lead to a partial or complete replacement or upgrade of some solutions when deemed necessary by Intapp', and that models are upgraded when new models outperform existing ones. No notice period for a model change is published. Intapp also publishes a Cloud Maintenance Policy and a Deprecated Features Policy, neither of which was read.
Ask: What written notice do we get before the model behind an Assist function changes, and can we defer or decline an AI model change?
Export and exit
Not confirmed
No export format, exit assistance or post-termination data-availability period was found on the pages read. The only related published statement is that a client cannot take a model with it: 'No, fine-tuned models can only be used with Intapp products.' The contractual return-or-delete provision would sit in the data processing addendum or the master agreement, neither of which was read.
Ask: On exit, what can we export, in what formats, over what period, and does it include Assist prompts, answers and the tags it wrote into our records?
Connectors and outbound flows
Public claim only
The product page publishes integrations for Microsoft Teams and for third-party data providers. The Applied AI page states 'No outputs of Intapp Applied AI pass directly to systems outside of Intapp infrastructure.' In the sub-processor list, the Intapp Assist section names no connector or data provider; the third-party data providers appear against Intapp DealCloud and other products.
Ask: Which connectors will be enabled for us, and does the statement that no Assist output passes directly to systems outside Intapp infrastructure survive enabling the Microsoft Teams app?
Open-web search and grounding
Public claim only
The Intapp Assist section of the sub-processor list names no web-search or web-grounding provider. Exa Labs Inc. appears under the heading 'Intapp Celeste' with the activity 'Web grounding services' and location United States, and footnote 2 states that where a customer uses Celeste functionality alongside another product, that product's sub-processors also include the Celeste sub-processors. The Intapp Terms product page states Ask Intapp 'will research the answer, exclusively referencing your firm's Intapp Terms data'. Separately, the Applied AI page states Intapp collects data from the web to build training data for classifiers, which is a vendor activity rather than a query-time search.
Ask: Does any Assist function query the open web, and if Celeste is enabled alongside it, does the Exa Labs web-grounding path apply to our data?
Autonomous action
Public claim only
The Applied AI page states: 'Intapp Applied AI does not act or make decisions — the user is ultimately responsible for the action taken or decision made', and 'No outputs of Intapp Applied AI pass directly to systems outside of Intapp infrastructure.' Against that, the product page describes Assist 'automating tasks like data entry and relationship tracking' and tagging content automatically, which are writes into the firm's own records inside Intapp. Nothing read states which of those writes happen without a person confirming them.
Ask: Which Assist actions write to our records without a person confirming, and can automatic tagging be set to suggest-only?
Human review before use
Public claim only
The Applied AI page states 'The output of our AI models is designed to be presented with context reminding users their review is necessary before acting on AI-generated information', and that most models return confidence scores that are useful for ranking but 'cannot be used as absolute quality metrics'. It states generative use cases mostly follow a retrieve-augment-generate approach 'which enables explicit referencing to the source of information used to generate the model response', and that for use cases not implementing that approach 'there is no way to avoid the potential surfacing of copyrighted content in model outputs'. A customer quote on the product page describes reviewing and confirming the AI's obligation categorisation. No contractual human-oversight obligation was read.
Ask: Which Assist functions cite their sources and which do not, and is there a setting that requires a named reviewer before an Assist output is relied on?
