Read for intent
Identify what a message asks for, not simply which keywords appear.
A communications framework for organisations that operate at scale. Intelligent classification can bring the right correspondence to the right team—while sensitive matters remain subject to meaningful human judgment.
Exania works with government bodies, research organisations and enterprises. Their messages may involve technical information, procurement, personal data, safety, legal rights or confidential research. Not all correspondence should be handled in the same way.
This policy explains the standards Exania applies when assessing and routing communications, including any AI-assisted triage configured for an approved channel. A classification is a recommendation for handling—not a factual verdict about the sender or an automatic decision about their rights.
Operational details may differ by channel, contract and jurisdiction. The interactive examples on this page use fictional messages and do not display an actual inbox or live company systems.
Identify what a message asks for, not simply which keywords appear.
Limit collection and exposure of information to the people authorised to address it.
Ambiguous, urgent or consequential requests require review, not confident guesses.
Authorised specialists approve substantive replies in sensitive or high-impact matters.
Explore how a structured communications process can validate, classify and route an enquiry while preserving the opportunity for a person to intervene.
The receiving channel captures the information necessary to deliver and manage the enquiry. Messages from public forms, authorised email addresses and institutional contacts can follow different intake rules.
A specialist triage system may assign one or several labels. The absence of a category—or low certainty—should direct a message toward review rather than cause it to disappear.
Tenders, regulated engagements, public-sector research, official requests and institutional verification.
ROUTE / INSTITUTIONAL TEAMScientific collaborations, model evaluation, deployment constraints, integration and technical questions.
ROUTE / TECHNICAL REVIEWAccess, correction, erasure, objections, complaints and requests concerning personal information.
ROUTE / PRIVACY REVIEWPotential vulnerabilities, suspected misuse, compromised accounts, exposures and urgent incident information.
ROUTE / SECURITY REVIEWCorporate engagements, formal partnership proposals, press questions and public statements.
ROUTE / AUTHORISED TEAMRoutine enquiries, requests for documentation, status checks, event invitations and other messages.
ROUTE / RELEVANT TEAMChoose a sample message. The interface demonstrates how the topic, potential impact and need for human review affect its proposed route.
A formal research or procurement enquiry merits contextual review by an authorised institutional contact.
This is an interactive demonstration built from synthetic messages, not a live email service, a measured classification score or a claim about production-level automated handling. Actual processing is subject to approved infrastructure and operating procedures.
Priority is determined by potential harm, urgency and the nature of the request—not just the sender’s title or the presence of words such as “urgent”.
Suspected security incidents, imminent safety concerns, verified access compromise and sensitive disclosures may require prompt specialist assessment and restricted handling.
Privacy rights, legal correspondence, complaints, formal notices and contractual matters are referred to an authorised reviewer to assess obligations and applicable deadlines.
Government enquiries, research proposals and enterprise requirements are assigned according to mission area, confidentiality and relevant subject-matter expertise.
Public documentation enquiries and general guidance can be directed to published resources, with escalation available where context remains unclear.
A message may meet multiple criteria. When priority is uncertain, the handling route should favour safe review rather than unattended automation. This page does not promise a fixed response time.
AI can help identify the subject, summarise context or prepare material for assessment. It must not substitute for an authorised professional where judgement, confidentiality, rights or substantial organisational commitments are involved.
Review the original communication, relevant records and uncertainty before taking action.
Legal, security, privacy and specialist technical communications require suitable authority and expertise.
A reviewer can change the classification, request clarification, reject a draft, or escalate to another team.
Where applicable, record routing and approval decisions in accordance with access and retention rules.
When a channel is equipped for assisted triage, a flagged case can enter a restricted review queue and trigger a notification to the responsible team’s approved mailbox or case-management system. The handoff should preserve enough context to act without disclosing more than necessary.
Record topic, urgency signals, reason for escalation and uncertainty.
Send a controlled case reference and summary; provide access to the original through an authorised channel.
The assigned reviewer checks the source, determines next steps and authorises any substantive response.
Where enabled, receipt confirmations may acknowledge that a message reached the right channel. Such acknowledgements do not mean that an enquiry has been accepted, that a request has been verified, or that any commitment has been made.
AI-assisted drafting may organise information or suggest neutral wording, but substantive statements on contracts, policy, data rights, security incidents, regulatory matters, pricing and institutional commitments require appropriate review before being sent.
Messages sent externally should identify the organisation and an appropriate reply channel. No email should be treated as an official legal or contractual acceptance merely because it was generated or routed by a model.
Neutral acknowledgement and reference information. No substantive decision or promise.
Drafting assistance or topic summaries submitted to a responsible reviewer.
Approved statements and actions communicated through an authorised corporate channel.
Classification does not make a message safe. Controls on channels, storage, access and approved processing environments matter throughout the lifecycle.
Collect and process what is needed to understand, route and respond. Avoid unnecessary personal or confidential information in general forms.
Security reports, sensitive datasets and restricted government material should use approved channels, not a public contact form or an unapproved AI tool.
Access should be limited to authorised teams. Logging and retention are subject to purpose, applicable law and documented organisational policy.
Data must not be repurposed for model training solely because it appeared in an email. Any training or reuse requires a separate lawful and approved basis.
Links, attachments and embedded instructions may be malicious. AI-generated classifications are not security approvals and must not override validation rules.
Exania operates within Exohood Labs Inc. arrangements; EXO POWER COMPUTING SERVICES LTD may provide computing services where engaged under suitable terms. Roles depend on the actual contract.
For institutional enquiries, partnerships and research discussions, begin at the Exania contact page. For privacy rights, corrections and data-processing concerns, contact privacy@exania.ai.
Security information should be limited to the minimum needed to initiate a secure handoff. Do not include passwords, authentication tokens, full confidential datasets or classified records in public web submissions.
If you believe a message was misclassified, delayed, mishandled or directed to an unsuitable team, request manual review and describe the original channel and reference, if one was provided.
These answers explain the boundaries of AI-assisted communications, manual review and expectations when contacting Exania.
Not necessarily. Handling depends on the channel and approved operating configuration. Some messages may be routed through conventional rules or directly to a person. This policy describes the intended standards when AI-assisted classification is used.
A neutral acknowledgement may be automatic where configured. Substantive replies concerning privacy, contracts, security, financial commitments or institutional decisions require appropriate human authorisation.
A reviewer should be able to correct a classification or route. Send a follow-up using the original channel or contact Exania with the reference and the requested correction.
No. Topic labels help organise work. They should not be treated as an assessment of a person, an automated refusal of rights or a final determination of an institutional request.
Only where permitted by the receiving channel and relevant controls. An attachment may be restricted, rejected, isolated or referred for manual security review; do not assume it will be analysed automatically.
Yes. Use the Exania contact page for general escalation, or privacy@exania.ai for privacy matters. Include sufficient non-sensitive context to explain the concern; a public website is not an appropriate channel for classified information.
No fixed service-level response time is established by this general page. Binding timelines may arise from law or an applicable agreement. Urgency labels are handling aids, not guarantees of delivery or resolution.
This policy does not authorise automatic training on correspondence. Any model training or additional use of communications requires a documented basis, approved safeguards and disclosures where legally required.
Tell us the nature of your request, who you represent and what expertise or decision it needs. The right response begins with the right context.
Contact ExaniaOnly essential website functions are active by default. Choose which optional services may collect information during your visit. Change your preferences at any time from the footer.
Essential website functions and storage of your choice.
ALWAYS ACTIVEPreferences are retained for up to six months in this browser. Read the privacy policy for details.