Product and intended use
Define the configuration, people, information and tasks being assessed.
Product and plan covered
Public claim only
Brightflag is sold as ELM covering e-billing, spend, matters and legal-service requests, plus invoice AI and Ask Brightflag. No public plan matrix. Customer MSA and Trust Center are gated.
Ask: Confirm which modules (invoice AI, Ask Brightflag, Connect, service requests) are on our order form and attach the MSA/DPA.
Deployment model
Public claim only
Multi-tenant AWS SaaS. Security: data centres in Ireland, the United States and Australia, chosen by customer location. Privacy: platform information can be hosted in the EEA; websites default to the EEA unless the customer elects US hosting. No on-premises option was found. Checked 2026-10-10. Vendor page: “Our data is stored in secure data centers in Europe (Ireland), the United States and Australia.” (https://www.brightflag.com/security). Recorded as Australian region available.
Ask: Will our tenant, backups, logs and Ask Brightflag indexes sit only in Australia?
AI features in scope
Public claim only
Invoice AI reads every invoice, emails a work summary and billing-guideline hits, and allows approve-from-inbox. Ask Brightflag opens an LLM window on the current page over matters, vendors and spend, including drafting vendor emails. Navigation describes models and agents trained and tuned for legal teams. A site AI agent answers product questions.
Ask: Which of invoice AI, Ask Brightflag, email-approve and the site agent will be on for our tenant?
Jurisdictions and legal content
Public claim only
Content is the customer's invoices, matters and vendors, not a licensed Australian case-law library. Website terms: Ireland law and Irish courts. Privacy last modified 23 July 2026 (GDPR/CCPA/CPRA/VCDPA).
Ask: Confirm which Brightflag entity and law govern our MSA, and that Ask Brightflag will not browse open-web Australian law.
Data, controls and requirements
Compare the actual configuration with firm, client and jurisdiction requirements, supported by evidence.
Client authority and professional use
Not confirmed
Invoice reviewers may be finance or legal ops, not only lawyers. Ask Brightflag can draft vendor emails. Enabling Connect on Gmail and approve-from-inbox is the customer's authority decision. Website terms do not address professional use.
Ask: Who may approve invoices from email, and can we restrict Ask Brightflag so it cannot draft outbound vendor mail?
Foundation-model providers
Not confirmed
Ask Brightflag is described as an LLM. Privacy says customer data may be processed by AI subprocessors listed in the DPA and at trust.brightflag.com, which did not scrape. No named OpenAI, Azure or Anthropic entity appears on the public AI or security pages.
Ask: Name each model, legal entity, region and contract used for invoice AI and Ask Brightflag, and the notice period before a model is swapped.
Subprocessors
Not confirmed
Privacy 1.7: all subprocessors, including AI subprocessors, are listed in the DPA and published at trust.brightflag.com. That Trust Center URL failed to scrape (likely gated). AWS is named as the host on the security page.
Ask: Provide the current dated subprocessor register, including AI processors, with entity, country, purpose and objection rights.
Training on customer content
Not confirmed
Navigation markets 'models and agents trained and tuned for legal teams' without saying on whose data. Privacy uses analytics on how customers use the Services to make them more intuitive. No clause prohibiting foundation-model training was found in the public Privacy Notice or website terms.
Ask: Does the signed DPA/MSA prohibit training of Brightflag or third-party models on invoices, prompts, outputs and embeddings?
Feedback and quality monitoring
Public claim only
Privacy: generative AI may be used for support and to analyse crash information; usage analytics to study how customers use the Services. Whether invoice accept/reject labels train a shared model is not stated.
Ask: Do invoice-review decisions and Ask Brightflag feedback stay in our tenant only?
Retention of prompts, outputs and matter indexes
Public claim only
Security: retention and deletion configurable. Privacy: keep personal data in active systems only as long as needed. No numbered default for Ask Brightflag chats or invoice-AI logs.
Ask: State default and configurable retention for invoices, Ask Brightflag chats, embeddings and AI-subprocessor logs on an Australian tenant.
Deletion and certified erasure
Public claim only
Security: records not retained longer than needed; dSAR support. Privacy: right to erasure with legal exceptions; Brightflag may delete collected information per retention policies. No certified-deletion SLA covering AWS or AI subprocessors was found in public terms.
Ask: On exit, what is deleted (invoices, chats, embeddings, subprocessors), in what period, and will you certify deletion in writing?
Vendor and verifier access to content
Public claim only
Privacy: access only to employees and consultants with a need for a lawful purpose; staff are trained. Support may use generative AI. Trust Center is staff-gated.
Ask: Who can read our invoices and Ask Brightflag chats, from which countries, under what approval, and can we see those access logs?
Access control, SSO and audit
Public claim only
Security: SAML SSO and SCIM; roles and permissions; every action captured for audit. Approve-from-inbox is an access-control choice.
Ask: Confirm SSO/SCIM and MFA on our plan, and whether we can export a complete Ask Brightflag and invoice-approval audit log.
Encryption in transit and at rest
Public claim only
Security: AES-256 at rest with AWS KMS; HTTPS with minimum TLS 1.2 in transit. Customer-managed keys were not offered on the page read.
Ask: Confirm encryption of Ask Brightflag indexes and the path to any AI subprocessor, and whether customer-managed keys are available for Australia.
Independent security assurance
Public claim only
ISO/IEC 27001:2013 (older edition than 2022), SOC 1 Type 2 and SOC 2 Type 2 across all five Trust Categories, CREST penetration testing. Reports live in a gated Trust Center and were not inspected.
Ask: Provide the current ISO 27001 (state the edition) and SOC 2 Type 2 reports covering Ask Brightflag and invoice AI, with scope and exceptions.
Contract, DPA, liability and forum
Not confirmed
Public website terms (Ireland law) govern the marketing site. Privacy says processor activity is under a DPA with the customer. That DPA and the product MSA are not published. Group entities: Shine Analytics Limited, Brightflag Inc, Brightflag Pty Ltd.
Ask: Provide the current MSA/DPA for an Australian tenant, with liability, training and AI-subprocessor terms.
Cross-border transfers and APP 8
Public claim only
Security offers Ireland, US and Australia data centres. Privacy: EEA hosting option; websites in the EEA unless US elected; CRM and conferencing in the US with SCCs (not DPF). AI subprocessors are not listed on the public page. Marketing and support tools may keep contact data in the US.
Ask: For invoice AI and Ask Brightflag on an Australian tenant, list every country for storage, inference, backups and support.
Incident notification and history
Not confirmed
No contractual 72-hour clause in website terms. Vulnerability reports to disclosures@brightflag.com. Incident history not published. DPA (gated) may contain the SLA.
Ask: State the contractual notification period from discovery, NDB/OAIC cooperation, and any material incidents in the last 24 months.
Change notice for models, terms and processors
Public claim only
Privacy: material privacy changes notified by email or in-product; continued use is acceptance; unacceptable changes require the customer to stop using the Services. Website terms likewise. Subprocessor change notice lives in the gated DPA.
Ask: What written notice will we get before a new AI model or subprocessor is added, and can we object?
Export and exit
Not confirmed
Privacy includes a portability right for certain personal data. Export formats for invoices, matters and Ask Brightflag history, and certified deletion of AI subprocessors, are not in public terms.
Ask: What export we receive for invoices, matters and Ask Brightflag history, and will you certify deletion within 30 days including AI subprocessors?
Connectors and third-party AI tools
Not confirmed
Gmail Connect and finance integrations are optional. Enabling them is the customer's choice. Privacy: Google API Limited Use for Connect. Website terms disclaim third-party sites.
Ask: List every connector that will be on for our tenant and the data each receives.
Open-web search
Not confirmed
Ask Brightflag is described as using Brightflag as the system of record. The site AI agent answers product questions. No open-web toggle was found.
Ask: Does Ask Brightflag browse the open web, and can we disable any such retrieval?
Agent actions and confirmation
Not confirmed
Invoice AI can email a summary and let a reviewer approve from the inbox in a single click. Ask Brightflag can craft vendor emails. Workflow rules are customer-configured. Those are action gates the firm must set.
Ask: Can we disable approve-from-inbox and block Ask Brightflag from sending or drafting vendor email without a second Legal confirmation?
Human review before use
Public claim only
Invoice AI is framed as assisting a reviewer who can open the invoice. Approve-from-inbox still requires a human click, but not necessarily a lawyer. Ask Brightflag chat does not require a lawyer to approve an answer before it is shown.
Ask: Can we require every invoice AI recommendation and Ask Brightflag answer to be approved by a named lawyer before it is acted on?