Product and intended use
Define the configuration, people, information and tasks being assessed.
Product and plan covered
Public claim only
This profile covers the Appian Platform as described on the vendor's platform overview and artificial-intelligence pages (read 2026-09-27), which the vendor has also marketed as the AI Process Platform. The platform page names five parts: process orchestration, unified data (data fabric), AI, modernisation and process intelligence. The AI page names Agent Studio, AI Skills, AI Copilot, Composer and DocCenter. Appian Government Cloud is a separate offering described on the compliance page, and the DPA contemplates both a Cloud Subscription Agreement and a Master On Premise Agreement.
Ask: For our order form, which Appian release and which named AI components (Agent Studio, AI Skills, AI Copilot, Composer, DocCenter) are included, and which are separately licensed or gated behind a support tier?
Deployment model and hosting location
Public claim only
The DPA defines the Agreement as a Cloud Subscription Agreement and/or a Master On Premise Agreement, so both hosted and on-premise licensing exist. The Appian Cloud availability page (read 2026-09-27) lists the data-centre locations, which include AUSTRALIA alongside five United States locations, Brazil, Canada, France, Bahrain, Germany, India, Ireland, Japan, Singapore, South Korea, Sweden and the United Kingdom, and states that all production data is replicated across availability zones within the same region and that backups are stored in a separate isolated location within the same geographical region. The subprocessor notice states that a sub-processor's 'processing location will depend on which data center you have selected on your order form'.
Ask: Will our order form name the Australian data centre for the production site, backups and disaster recovery, and is any component of the service (including AI processing, monitoring or support tooling) hosted outside Australia?
AI features in scope
Public claim only
The artificial-intelligence page (read 2026-09-27) describes four AI groups: AI agents that 'do work like classifying, extracting, summarizing, and automating', conversational AI, AI-assisted development, and 'Private AI'. Its FAQ names Agent Studio ('building, managing, and governing AI agents'), AI Skills, AI Copilot ('finds information buried in documents, answers questions directly within applications'), Composer (application generation from natural-language prompts) and DocCenter (intelligent document processing with generative AI). The platform page names AI agents, RPA and IDP as automation components.
Ask: Which of these AI features would be enabled in our tenant by default, and can each be turned off centrally for nominated applications or user groups?
Jurisdiction and legal content
Public claim only
The platform carries no legal content library and is not marketed as legal software; nothing on the pages read claims primary-law coverage for any jurisdiction. The Australian-specific published item is on the compliance page: Appian states that its IRAP assessment means it 'has been assessed against the requirements laid out by the Australian Government Information Security Manual (ISM) up to the Protected level', produced by the Australian Cyber Security Centre. The compliance page also lists UK G-Cloud, Canada Protected B, Spain's ENS and Germany's C5.
Ask: Provide the IRAP assessment report and its date, the assessed scope (which Appian Cloud regions, which platform components, and whether the AI components are in scope), and the name of the assessor.
Vendor content directing how AI describes the product
Not confirmed
Checked on 2026-09-27 and nothing found. A site map of 40 appian.com URLs in the earlier pass surfaced no /llms.txt, /ai-info or /for-ai page, and a search of all seven Appian pages now held on disk (platform overview, artificial intelligence, the Victorian OPP customer page, the Data Processing Agreement, the security, compliance and availability pages) for an 'Instructions for AI Assistants' heading or equivalent block addressed to AI models, assistants or crawlers returned no match.
Ask: Does Appian publish any page or embedded block that instructs AI models, assistants or crawlers how to describe Appian, and if so where?
Data, controls and requirements
Compare the actual configuration with firm, client and jurisdiction requirements, supported by evidence.
Client authority and responsibility for what is put in
Public claim only
The DPA places authority squarely on the customer: 'Customer acknowledges that it has exclusive control and responsibility for determining the means and purposes and what Customer Personal Data Customer submits to Appian and warrants it has all authority, grounds, rights and consents and permissions for submission and transfer of Customer Personal Data and Processing by Appian'. The customer is the controller and Appian the processor. The DPA says nothing about legal professional privilege, court-ordered material or criminal-brief content.
Ask: For privileged and brief-of-evidence material placed in an Appian application, what notice or consent does Appian expect us to hold, and does Appian accept any restriction on processing privileged content?
Foundation-model providers
Not confirmed
No foundation-model provider is named in anything Appian publishes on the pages read. A search of all seven cached Appian pages for OpenAI, Anthropic, Azure, Bedrock, Gemini, Mistral and 'large language model' returned only navigation glossary links. The subprocessor notice (28 October 2025), which is the document that would name a model processor, lists only Amazon Web Services (hosting), Alert Logic (intrusion detection) and Snowflake (data warehousing for intrusion detection). The AI page instead describes 'Private AI' and states data stays 'within your organization's compliance boundary'. Third-party context only: the CAIDE/LIV snapshot names OpenAI/ChatGPT and Amazon Bedrock in connection with the Australian prosecution deployments; that is a university report describing customers, not a vendor statement.
Ask: Name every model used by AI Skills, AI Copilot, Agent Studio and DocCenter for our tenant, the legal entity that operates each model, the region in which inference runs, and whether any model is called outside Appian Cloud.
Subprocessors
Public claim only
Appian publishes a subprocessor notice, incorporated by clause 8.3 of the DPA, last updated 28 October 2025. It lists eleven Appian entity sub-processors with processing locations: Appian Corporation (USA), Appian Canada Corporation, Appian Europe Ltd. (United Kingdom), Appian Italy, Appian Japan G.K., Appian Mexico Software SRL, Appian Software Australia Pty Ltd. (Australia), Appian Software Germany GmbH, Appian Software International, LLC (Switzerland), Appian Spain, S.L., and Appian Computer Technologies India Private Limited. Third-party sub-processors are Amazon Web Services, Alert Logic and, from 1 January 2026, Snowflake, each with 'processing location will depend on which data center you have selected on your order form'. Clause 8.4 and the notice commit to updating the list at least 30 days before engaging a new sub-processor. No AI or model provider appears on the list.
Ask: If we enable AI agents, AI Skills or DocCenter, does any sub-processor not on the 28 October 2025 notice process our content, and will a model provider be added to that notice before it is used?
Training on customer content
Public claim only
The claim is on a marketing page, not in the contract. The artificial-intelligence FAQ (read 2026-09-27) states 'we never train our models on your data, and your data is never shared outside your organization's compliance boundary' and 'Appian doesn't share data with third parties or train algorithms on your data'. The Data Processing Agreement contains no AI or training clause; its nearest provision is clause 5(a), under which Appian 'shall only carry out any Processing of Customer Personal Data on the documented instructions of Customer' unless required by law. That limit is expressed for Customer Personal Data, so it does not by its terms cover customer content that is not personal data.
Ask: Will you warrant in the Agreement that no customer content, prompt, output, embedding or log from our tenant is used to train, fine-tune or evaluate any model, including content that is not personal data?
Feedback, telemetry and model refinement
Public claim only
Two statements on the same artificial-intelligence page sit close together and are not reconciled there. The FAQ on privacy says Appian does 'never train our models on your data'. The FAQ on monitoring lists 'Model effectiveness reviews. Monitor confidence levels, exceptions, and overrides to refine AI models over time', and lists audit trails logging 'every AI recommendation and decision'. Who performs that refinement, and on what data, is not stated on any page read, and the DPA is silent on telemetry.
Ask: Does the 'model effectiveness review' and refinement described on your AI page use any data from our tenant, and can overrides, confidence scores and exception logs be kept inside our tenant only?
Retention of content, prompts and logs
Public claim only
The DPA sets the duration of processing as 'the term and a reasonable period thereafter (within 30 days after the end of the provision of Services) in accordance with the Services relating to the Processing under this Agreement and Appian's data retention policies then in force'. That retention policy is not published on any page read. The availability page states production data is replicated across availability zones in the same region and included in nightly backups held in a separate isolated location in the same geographical region. No retention period is published for AI prompts, agent outputs, audit trails or process logs.
Ask: Provide the data retention policy referred to in clause 7(b) of the DPA, and state the default and configurable retention for AI prompts, agent outputs, audit trails, process logs and backups on an Australian site.
Deletion and certified erasure
Public claim only
DPA clause 5(i) requires Appian, upon request, to 'securely delete all Customer Personal Data within 30 days after the end of the provision of the Services relating to the Processing under this Agreement, and securely delete existing copies, unless EU law or the law of the EU member state(s) to which Appian is subject require storage'. The exception is framed on EU law only and the obligation is triggered by a request rather than automatic. Nothing on the pages read addresses deletion from backups, from AI caches or by sub-processors, or written certification of deletion.
Ask: Will you confirm in writing that deletion under clause 5(i) covers backups, indexes, AI prompt and output stores and sub-processor copies, that it applies to content that is not personal data, and that a deletion certificate will be issued?
Vendor and sub-processor access to content
Public claim only
DPA clause 5(c) requires Appian to ensure persons authorised to process Customer Personal Data are bound by confidentiality, and clause 8.2 states Appian entity sub-processors 'may respond to a customer's request for technical support and/or defect correction services'. The subprocessor notice names the eleven Appian entities and their processing locations, which include India, Mexico, Japan and Australia. The security page states 'Appian Cloud personnel located in the United States, Canada, Mexico, United Kingdom, Spain, Germany, Italy, Australia, Japan, and India', each subject to a formal screening process including a required background check. No customer-visible access log or approval workflow for vendor access is described on the pages read.
Ask: Can support and defect-correction access to an Australian site be restricted to named personnel in Australia, and will you provide us with a log of every vendor access to our data?
Access control, permissions and audit
Public claim only
The security page lists single sign-on, integration authentication, inbound API authentication, encryption of sensitive data fields in interfaces, and 'row-level data fabric security with user access preview' as included for all users, with trusted IP allow lists in premium tiers. The AI FAQ states 'Appian AI applies user permissions to make sure AI agents only access and share data with the right people in the right context' and that permissions set on a record are inherited by connected systems down to row level. Multi-factor authentication and SCIM provisioning are not mentioned on the pages read.
Ask: Confirm whether enforced multi-factor authentication and SCIM provisioning are available on our tier, and demonstrate that an AI agent cannot return content the requesting user could not open directly.
Encryption and key management
Not confirmed
The two vendor pages do not agree on key management. The security page lists 'Transport Layer Security (TLS) for end-user connections', 'Disk encryption to secure data at rest' and encrypted backups as included for all users, and places 'Bring your own key (BYOK)' and 'Database encryption' among 'Advanced security features aligned to premium tiers'. The artificial-intelligence FAQ states without qualification that 'all data is encrypted via your private key'. Neither page names algorithms or TLS versions. Appian Cloud's web tier is stated to be rated A+ by Qualys SSL Labs.
Ask: Which Appian Protect tier is required for bring-your-own-key and database encryption, and does the statement on the AI page that all data is encrypted with the customer's private key apply to a base subscription?
Independent security assurance
Public claim only
DPA clause 7.1 states the security program 'at a minimum adheres to applicable information security practices as identified in International Organization for Standardization 27001 (ISO/IEC 27001)'; clause 7.2 commits Appian to maintaining an annual SOC report and not materially reducing the overall level of security set out in its SOC 2 report; clause 7.3 provides current audit reports and certificates on request under confidentiality and supports one customer security review a year on 30 days' notice, while recording that for Amazon Web Services Appian 'cannot commit to procure AWS's compliance with any specific audit requirement from Customer'. The compliance page (read 2026-09-27) lists SOC 1 Type II/ISAE 3402, SOC 2 Type II, a publicly available SOC 3, PCI DSS, FedRAMP Moderate and High, DISA IL2/IL4/IL5, GovRAMP Moderate, ISO/IEC 27001:2022, 27017:2015 and 27018:2019, ISO 9001:2015, Canada Protected B, UK Cyber Essentials and Cyber Essentials Plus, Spain's ENS, C5 and an IRAP assessment up to PROTECTED. HITRUST appears in the security page's certification list, but the compliance page's HITRUST entry describes the framework without stating that Appian holds the certification, so the HITRUST position is not established.
Ask: Provide the current SOC 2 Type II report, the ISO 27001:2022 certificate with its scope statement, the IRAP report, and a direct answer on whether Appian holds a HITRUST CSF certification and for which scope.
Contract stack, contracting entity and forum
Public claim only
The DPA was read in full (2026-09-27); the Cloud Subscription Agreement it depends on was not, so liability, service levels and export obligations were not established. The DPA states the processor contracting entity is 'either: (a) Appian Corporation; or (b) Appian Software International LLC, as more particularly described in the Agreement with the Customer', names no Australian contracting entity, and takes its governing law, dispute resolution and liability cap from the Agreement. It provides that the DPA prevails over the Agreement in the event of conflict, and that 'The Competent Supervisory Authority shall be the supervisory authority of the EU Member State where the Customer is headquartered', which does not describe an Australian customer.
Ask: Which entity will sign our subscription agreement, under which governing law and forum, and will you issue an Australian addendum covering the Privacy Act 1988 and the Australian Privacy Principles in place of the EU-framed DPA?
Cross-border transfers and APP 8
Public claim only
The DPA's transfer machinery is European: 'Applicable Data Protection Laws' is defined to include the EU GDPR, the UK GDPR and Swiss data protection laws and does not mention the Australian Privacy Act; clause 3.1 relies on the EU-US, UK Extension and Swiss-US Data Privacy Frameworks, and clause 3.2 applies the EU standard contractual clauses (Module Two) and the UK Addendum to restricted transfers. Clause 8.1 authorises transfers of Customer Personal Data 'to Appian branches, Affiliates and selected third parties ... within and outside the EEA or Switzerland'. Australia is an available data-centre location and sub-processor processing location follows the data centre chosen on the order form, but no APP 8 mechanism, no Australian transfer clause and no statement of where model inference runs appear on any page read.
Ask: For an Australian site, list every country in which our content may be stored, processed, backed up, supported or used for inference, and state the APP 8 mechanism you rely on for each disclosure.
Incident notification and history
Public claim only
DPA clause 7.4 requires Appian to notify the customer 'without undue delay (unless compelled by law otherwise), and, where feasible, not later than 72 hours after becoming aware' of a security incident affecting Customer Personal Data processed by Appian or its sub-processors, and excludes incidents 'attributable to Customer or Customer's Users'. Clause 5(g) adds notification in writing without undue delay and records that notification is not an acknowledgement of fault. The security page states the Security Incident Response Team investigates submissions and 'Appian notifies affected customers'. No incident history and no Notifiable Data Breaches scheme commitment appear on the pages read.
Ask: Will you commit contractually to supporting our Notifiable Data Breaches assessment within a fixed period, and list any material security incidents affecting Appian Cloud in the last 24 months, including at sub-processors?
Change notice for processors, models and controls
Public claim only
The DPA and the subprocessor notice commit to updating the published sub-processor list at least 30 days before Appian engages a new sub-processor, and clause 3.1 commits to at least 30 days' notice before Appian withdraws from its Data Privacy Framework self-certification. Clause 7.2 permits Appian to 'replace or upgrade controls to maintain or increase the security level' provided the overall level in its SOC 2 report is not materially reduced. The notice records a planned change of its own: from 1 January 2026 customer sites transition from Alert Logic to Snowflake, running concurrently for about six months. No notice commitment for a change of model, model version or AI feature behaviour was found, and the DPA gives no objection right to a new sub-processor beyond the 30-day posting.
Ask: What notice will we receive before a model, model version or AI feature behaviour changes, and do we have a right to object to, or suspend use of, a newly listed sub-processor?
Export and exit
Not confirmed
No export or return-of-data obligation was found. The DPA deals with deletion (clause 5(i), within 30 days of the end of Services on request) but not with export, format or assistance on exit; those terms would sit in the Cloud Subscription Agreement or Master On Premise Agreement, neither of which was read in this pass. The terms of use linked in the site footer at https://appian.com/legal/terms-of-use was also not read.
Ask: What does the Cloud Subscription Agreement require on termination: in what formats are applications, matter records, documents and audit trails returned, over what period, and at what cost?
Connectors and what AI agents can reach
Not confirmed
The platform's design point is reaching other systems: the platform page describes data fabric as giving 'employees and AI agents a secure way to access and update information across your ecosystem-without data migration', and the AI FAQ says agents apply user permissions and that record permissions are inherited by connected systems. The security page lists integration authentication, inbound API authentication and, in premium tiers, AWS PrivateLink and VPN connectivity. No list of available AI connectors, and no statement of whether an agent can call an external service the firm has not approved, appears on the pages read.
Ask: List every connector an AI agent could invoke in our tenant, and show the administrative control that prevents an agent calling a system or external endpoint we have not approved.
Open-web retrieval
Not confirmed
Nothing on the pages read states whether any AI component retrieves from the open web. The artificial-intelligence FAQ addresses the opposite direction only, saying data stays 'within your organization's compliance boundary' and that AI 'runs in a secure, isolated environment, ensuring sensitive data never leaks to public models'. No search, browsing or retrieval toggle is documented on any page read.
Ask: Can AI Copilot or an AI agent retrieve content from the open internet, and if so can outbound retrieval be disabled for our tenant and logged when used?
Agent actions and confirmation
Not confirmed
The same artificial-intelligence page describes both autonomous action and pre-action review. Its first FAQ says AI in Appian includes 'AI agents that autonomously make decisions and take action', and a later answer says 'Appian AI agents can automate complex work based on context, data, and prior outcomes-without human input. If a problem arises, the agent can adapt in real time using complex reasoning to adjust its course.' The governance answer states 'Human-in-the-loop oversight. Users can review, validate, or override AI recommendations before action is taken', alongside business rules, thresholds and escalation paths. Which behaviour is the default, and which is configurable, is not stated.
Ask: In Agent Studio, what is the default for an agent that writes to a record or sends external correspondence: does it execute or queue for confirmation, and can a mandatory confirmation step be enforced centrally and audited?
Human review before an output is used
Public claim only
The governance controls listed on the artificial-intelligence page are human-in-the-loop oversight, permissions management, business rules and policies constraining outputs, traceability 'showing when and why decisions were made, what data was used', and audit trails in which 'every AI-driven action is logged'. The platform page states Appian 'captures every human, AI, and system action for a centralized view'. These are marketing statements; the DPA contains no human-oversight clause, and nothing published says a reviewer step can be made mandatory or that review is recorded against a named person.
Ask: Which product setting forces a named human reviewer before an AI output leaves the system, and can we export an audit trail showing who reviewed which output and when?
