AI Usage Policy

Last updated: 22 October 2026

Gevu is a trading name of Seawolf Facilities Services Ltd (“Seawolf”, “we”, “us”, or “our”), a company registered in England & Wales (no. 15028189). Gevu provides AI-assisted features powered by third-party large language models (“AI Features”). This policy applies to all AI Features across all parts of the Service — including any AI functionality added in the future — and governs all use regardless of which feature, role, or device is used to access it. This policy forms part of, and is expressly incorporated into, our Terms of Service. By enabling or using any AI Feature, by proceeding past any AI consent notice, or by benefiting from any AI-assisted output, you confirm that you have read, understood, and agree to be bound by this policy in full. If you do not agree, you must not use AI Features.

1. AI Features offered

The following platform features are currently powered by AI (Google Gemini via the Gemini API, or, where a Customer connects its own AI provider key, the provider that Customer chooses; see section 10A). This list may change as features are added, modified, or withdrawn:

• RAMS generation: AI-assisted drafting of Risk Assessment and Method Statement documents for cleaning operations.
• Proposal and scope generation: AI-assisted drafting of scope-of-work descriptions, cleaning specifications, and commercial proposal content.
• Safety data sheet (SDS) parsing: AI extraction and summarisation of chemical safety information from uploaded SDS documents.
• CRM contact research: AI-assisted summarisation of publicly available information about prospective clients.
• Marketing content generation: AI-assisted drafting of campaign emails and outreach copy.
• Survey generation: AI-assisted creation of operational survey and inspection templates.
• Weekly insights and operational analysis: AI interpretation of scheduling, payroll, absence, and compliance data to surface trends and suggestions.
• Inventory and task suggestions: AI-assisted recommendations for cleaning inventories and shift task lists.
• COSHH assessment generation: AI-assisted drafting of Control of Substances Hazardous to Health (COSHH) assessments from a product name, work area, and method description.
• Survey-map photo analysis: AI review of site survey photos and annotations to suggest hazards, COSHH considerations, or relevant inventory items for a marked location.
• AI search and operational assistant: a natural-language search across your workspace’s own records (staff, clients, invoices, quotes, shifts, incidents, feedback, and similar), including financial summaries, and AI-assisted drafting of automation rules from a plain-language description.
• Marketing send agent: AI-drafted outreach emails to leads and contacts as part of a marketing campaign you configure. Each campaign defaults to requiring your approval before any email is sent; you may instead configure a campaign to send AI-drafted emails automatically, without a per-email human review step, in which case you are solely responsible for the content sent under your account.
• RAMS renewal and site RAMS: AI-assisted updating of an existing RAMS when it is due for renewal, and drafting a RAMS for a whole site from its survey.
• Task wording: AI-assisted rewriting of a cleaning task description to make it clearer, and suggestions for an area’s tasks from its survey photos.
• Policy drafting assistant: AI-assisted drafting and editing of your company policies in the policy editor.
• Policy translation for staff: on the signing screen, a staff member can ask for an AI translation of a policy into their own language to help them understand it. It is labelled as an AI translation that may contain mistakes; the English version is always the one they sign and the one that applies.
• Prospect dossiers: AI-assisted summaries of a prospective client business, built from public sources such as Companies House and the business’s own website.
• AI receptionist (phone): where you connect your own Retell AI account, an AI voice assistant answers your business phone line, takes messages and passes urgent calls to you. Every call starts with a fixed notice that the caller is speaking to an AI assistant and that the call is recorded. The voice model is provided by Retell under your own agreement with it, not by Google; we store the call summary and transcript in your workspace.
• Recording transcription and summaries: when you tap Transcribe on a site survey recording, an area voice note or a meeting recording, the audio is sent to Google’s Gemini API to produce a transcript and a summary, and for meetings suggested action points that you review before any become tasks. Transcripts can mishear words, label speakers wrongly or leave gaps marked [inaudible], so check them against the recording before relying on them. You must tell people before you record them; see our Privacy Policy for how long recordings are kept.
• Website live chat assistant: on our public website, an AI assistant may answer chat messages from visitors, in and out of opening hours, using only our public website content. Its replies are always labelled as AI, it cannot see any account or workspace data, and it hands the chat to our team when asked, when a team member replies, or when a chat is flagged.
• Cleaning Manager assistant: an optional, per-workspace conversational assistant, described separately in section 1A below because it can take actions in your workspace, not just generate text.

AI Features are available subject to subscription tier eligibility and daily token limits. We reserve the right to add, modify, restrict, or withdraw any AI Feature at any time without notice.

1A. The Cleaning Manager assistant — agentic actions

Where an owner or admin enables it, the Cleaning Manager assistant is a persistent, named AI assistant for a single workspace. It can read operational, staff, incident, and financial data (including revenue, cost, and profit figures) within that workspace to answer questions, and can take a limited set of actions on your behalf — currently limited to creating an incident report. Depending on the “autonomy” level the Customer configures for the assistant, an action it proposes is either carried out immediately or held for a human in the workspace to approve or reject first. Every question asked and every action taken or proposed is logged and reviewable by the workspace’s owners and admins.

You are solely responsible for: (a) choosing an autonomy level appropriate to your business and reviewing that choice periodically; (b) reviewing any record the assistant creates, whether immediately or after approval, for accuracy before relying on it; and (c) ensuring only appropriately authorised staff can approve actions the assistant proposes. We may extend the assistant’s capabilities over time, including to other record types; any new capability is governed by this policy from the point it is enabled, whether or not this page has yet been updated to name it specifically, consistent with section 14 below.

1B. Telling people when they are dealing with AI

Where AI Features communicate with people on your behalf, you must not present them as a person. Our website chat assistant and the AI receptionist always say they are AI, and that notice cannot be removed. If you use AI-drafted content in messages sent to clients, staff or prospects (for example automatically sent marketing emails), you are responsible for any disclosure that the law, a platform’s rules or your industry codes require, and for honouring opt-outs.

No AI Feature makes decisions about people on its own. AI output about staff (for example operational insights mentioning absence or hours) must not be used as the sole basis for any decision that significantly affects someone, such as discipline, dismissal or pay, without a person reviewing the underlying records.

2. AI Features are tools, not professional services

AI Features are software tools designed to assist your human judgment. They are not a substitute for professional advice of any kind. In particular:

• Not health and safety advice: AI-generated RAMS, risk assessments, method statements, and SDS summaries do not constitute professional health and safety advice and have not been prepared, reviewed, or approved by a health and safety professional.
• Not legal advice: AI-generated contract language, policy text, or regulatory summaries do not constitute legal advice and should not be relied upon as such.
• Not financial or accounting advice: AI-generated insights, cost analyses, or payroll summaries do not constitute financial, accounting, or tax advice.
• Not medical advice: No AI Feature generates or is intended to generate medical advice of any kind.

Seawolf Facilities Services Ltd is a software company. We do not provide professional services in any regulated field. Nothing generated by AI Features constitutes advice, a professional opinion, or a compliance assurance from Seawolf Facilities Services Ltd or any affiliated entity.

3. No warranty on AI output accuracy

AI-generated content is produced by probabilistic machine learning models. Outputs may be:

• factually incorrect, incomplete, or outdated;
• internally inconsistent or self-contradictory;
• plausible-sounding but entirely fabricated (“hallucination”);
• inappropriate for your specific site, jurisdiction, or operational context;
• non-compliant with applicable law or regulation;
• different in quality or content from one generation to the next, even for identical inputs.

Seawolf Facilities Services Ltd makes no warranty, express or implied, as to the accuracy, reliability, fitness for purpose, completeness, legality, or suitability of any AI-generated output. AI outputs are provided “as-is” and “as available” without representation of any kind. We do not guarantee that AI output will meet any particular standard, pass any audit, or satisfy any legal or regulatory requirement.

4. Your sole responsibility to review AI output

You are solely and entirely responsible for reviewing, verifying, correcting, and approving all AI-generated content before: (a) relying on it yourself; (b) applying it operationally; (c) sharing it with any third party; or (d) incorporating it into any contract, document, or communication. You must not deploy AI-generated output without human review by a person with the competence and authority appropriate to the content type. Specifically:

• RAMS documents: Must be reviewed, adapted to actual site conditions, and approved by a competent person as defined under UK health and safety law before being applied to any worksite, shared with any client, or relied upon by any worker.
• SDS summaries: Must be cross-checked against the original manufacturer’s Safety Data Sheet before being used in any COSHH assessment or chemical handling procedure. AI summaries may omit critical hazard information.
• Contract and quote language: Must be reviewed by an appropriately qualified person before being presented to a client or incorporated into any legally binding document.
• Marketing content: Must be reviewed for factual accuracy and compliance with applicable advertising standards (including ASA/CAP codes) before distribution.
• Operational insights and analysis: Must be treated as indicative only and verified against primary records before being used for payroll decisions, disciplinary action, compliance reporting, or any other consequential purpose.
• CRM research: Must be verified for accuracy and must not be relied upon as complete or definitive information about any individual or business.

5. Health and safety — critical legal notice

The following applies to all AI Features that generate or contribute to health and safety documentation:

• Under the Health and Safety at Work etc. Act 1974, employers have a non-delegable duty to ensure, so far as is reasonably practicable, the health, safety, and welfare of their employees and others affected by their work.
• Under Regulation 3 of the Management of Health and Safety at Work Regulations 1999, employers must carry out a suitable and sufficient risk assessment. An AI-generated document does not constitute a “suitable and sufficient” risk assessment under this regulation unless it has been reviewed, customised to actual site conditions, and signed off by a competent person.
• Under the Control of Substances Hazardous to Health Regulations 2002 (COSHH), employers must conduct their own COSHH assessment based on actual substances in use. An AI-parsed SDS summary is not a COSHH assessment.
• Under the Construction (Design and Management) Regulations 2015 (CDM 2015), specific duties apply to the preparation and content of method statements. AI-generated method statements do not satisfy these duties without competent review.
• Seawolf Facilities Services Ltd is not a health and safety consultant, is not registered with IOSH, NEBOSH, or any regulatory body, and does not accept any responsibility for the content of AI-generated health and safety documents or their fitness for use at any worksite.

6. No liability for AI output

Nothing in this section excludes our liability for death or personal injury caused by our own negligence, for fraud or fraudulent misrepresentation, or for any other liability that cannot lawfully be excluded under English law.

Subject to the above, and to the fullest extent permitted by applicable law, Seawolf Facilities Services Ltd, its officers, directors, employees, agents, contractors, and licensors accept no liability whatsoever — whether in contract, tort (including negligence), breach of statutory duty, or otherwise — for any loss, damage, injury, death, claim, regulatory penalty, fine, enforcement action, or other consequence of any nature arising directly or indirectly from:

• any error, omission, hallucination, inaccuracy, or unsuitability in any AI-generated output;
• your reliance on AI-generated content without independent human verification;
• the application of an AI-generated RAMS or risk assessment to any worksite without competent review and approval;
• use of an AI-generated SDS summary in any COSHH assessment or chemical handling context;
• any decision made, contract entered into, communication sent, or action taken based on AI-generated content;
• failure of AI-generated content to satisfy any legal, regulatory, contractual, or certification requirement;
• any claim by a third party (including an employee, client, subcontractor, regulator, or insurer) arising from AI-generated content produced through your account;
• the unavailability, interruption, degradation, or withdrawal of any AI Feature;
• changes to the behaviour, accuracy, or outputs of the underlying AI model;
• any regulatory enforcement action, prosecution, or civil claim arising from your use of AI Features.

This exclusion applies regardless of whether the loss was foreseeable, regardless of the form of action, and regardless of whether you or any third party notified us of the risk of such loss. Your use of AI Features is entirely at your own risk. If a court decides that any part of this exclusion is not enforceable, our liability for AI Features is limited as set out in the limitation of liability section of our Terms of Service.

7. Third-party reliance

Where you share, distribute, or make available AI-generated content to any third party (including clients, subcontractors, employees, regulators, or insurers), you do so entirely at your own risk and at the risk of those parties. Seawolf Facilities Services Ltd accepts no liability to any third party for AI-generated content produced through your account. You agree to indemnify and hold harmless Seawolf Facilities Services Ltd from any claim, loss, cost, or liability brought by any third party arising from AI-generated content produced through your account, including any claim by an employee, worker, or member of the public in connection with health and safety documentation or any other AI-generated content you have distributed.

8. Ownership of AI-generated content

Subject to the terms of the underlying AI model provider (Google LLC), AI-generated outputs produced through your account are owned by you as the Customer. Seawolf Facilities Services Ltd does not assert ownership over AI-generated content in your workspace. However, because AI models may generate similar or identical outputs for different users from similar inputs, we make no representation that AI-generated content is unique, original, or free from prior use by others. You are solely responsible for assessing whether AI-generated content infringes any third-party intellectual property right before use.

9. Your regulatory compliance responsibility

You are solely responsible for ensuring that your use of AI Features complies with all applicable laws and regulations, including but not limited to: UK data protection law (the UK GDPR and the Data Protection Act 2018, as amended by the Data (Use and Access) Act 2025); health and safety legislation applicable to your industry and jurisdiction; advertising standards and consumer protection law; professional regulatory requirements applicable to your business; and any emerging AI-specific regulation applicable to your use case. We make no representation that AI Features, or your use of them, will comply with any specific regulatory requirement. If your use of AI Features is subject to regulation (for example, if you are deploying AI in a regulated industry), you are responsible for obtaining your own legal and regulatory advice.

10. Data sent to AI models

When you use an AI Feature, operational data from your workspace — which may include site names, task descriptions, shift data, chemical names, contact names, document text, financial figures (such as revenue, cost, and profit), incident reports, and other operational content — is included in prompts sent to Google’s Gemini API. For features that can take an action rather than only generate text (see section 1A), the model’s output may also be used directly by our system to create a record in your workspace, in which case Google’s role is limited to generating the proposed content or decision, not to storing or executing it. By using AI Features, you: (a) consent to this data being processed by Google LLC under the Gemini API Terms of Service and Google’s Privacy Policy; and (b) confirm you have a lawful basis under UK GDPR for any personal data included in prompts. Seawolf Facilities Services Ltd is not responsible for Google’s handling of prompt data. Prompts and responses are not used by Gevu to train AI models. We use Google’s paid Gemini API service, under whose terms Google does not use prompts or responses to improve its products; Google may keep them for a limited period to detect abuse. Where a feature analyses a photo, a document or an audio recording (such as site survey photos, an uploaded safety data sheet or a survey recording), that photo or file itself is sent to the Gemini API for the analysis. We keep a log of AI requests and outputs for quality, cost control and dispute resolution; the log does not keep the photos or audio.

10A. Using your own AI provider key

On the Pro and Elite plans, an owner or admin can connect the Customer’s own API key from Google (Gemini API), OpenAI or Anthropic (Claude) in the workspace’s AI settings. While a working key is connected:

• Where the data goes: prompts for AI Features, including the operational data, personal data, photos and documents described in section 10, are sent to the provider the Customer chose, using the Customer’s key and under the Customer’s own agreement with that provider. That provider acts on the Customer’s instructions and is not a sub-processor of Seawolf Facilities Services Ltd. Features that only Google Gemini supports (currently voice transcription and web research) continue to run on our own Gemini account, as described in section 10.
• Your responsibilities: the Customer is responsible for its agreement with the provider, including any data processing terms, international transfer safeguards, and choosing a paid or business service whose terms do not allow the provider to use prompts to train or improve its models (for example, Google’s free Gemini API tier allows this and must not be used with staff or client data). We ask you to confirm billing is set up before a key is saved.
• Costs: the provider bills the Customer directly for all usage on the key. We do not apply a monthly limit to usage on your own key (we still apply per-user rate limits), so you should set a spending limit in your provider account. Charges made by the provider are your responsibility and are not refundable by us, except to the extent they were caused by our negligence, subject to section 12 of our Terms of Service.
• The key itself: we keep the key in restricted storage that only our servers can read, never show it again in full after it is saved, and use it only to run your AI Features. We check it with the provider when you save it and periodically after that, using the provider’s free model listing.
• If the key stops working: if the provider refuses the key (for example it is revoked, out of credit, or over quota), we stop using it, tell the workspace’s owners and admins, and AI Features go back to your plan’s included allowance until you save a working key again.
• Output: sections 2 to 9 apply in the same way whichever provider generates the output. Output quality, speed and available features can differ between providers and models, and the model you choose is your responsibility.

You can remove the key at any time in the same settings; we delete it from our systems when you do. Any discount linked to using your own key is described in section 4D of our Terms of Service.

11. Consent recording

For AI Features that generate a formal document for worksite or client use (currently: RAMS generation, SDS parsing, proposal and scope generation, and area analysis), you will be presented with a per-feature consent notice before first use. Your acceptance is recorded in your account with: a timestamp, your IP address, your device user agent, and the feature name. Consent is valid for 90 days and must be renewed thereafter. Consent records are retained as evidence of your acceptance of this policy and may be reviewed by our compliance team, and constitute your binding acknowledgement of the terms of this policy for that feature. Other AI Features display an inline “AI-generated, review before use” notice with a link to this policy rather than a separate consent screen; by proceeding past that notice, or by otherwise using or benefiting from any AI Feature, you accept this policy for that feature regardless of whether a formal consent screen was shown, as set out in the introduction above.

12. Insurance

You are solely responsible for ensuring that your use of AI Features — and any reliance on AI-generated content by you or your employees, clients, or subcontractors — does not invalidate, reduce, or otherwise affect any insurance policy you hold. We make no representation that AI-generated health and safety documentation, or its use at a worksite, will be accepted by your insurer. You should obtain confirmation from your insurer before relying on AI-generated RAMS or other safety documentation for insurance or liability purposes.

13. Prohibited uses of AI Features

You must not use AI Features to:

• (a) generate content that is false, defamatory, misleading, threatening, harassing, or intended to deceive any person or entity;
• (b) reproduce, closely imitate, circumvent, or substitute for any content, work, or database protected by copyright, database rights, trade marks, or any other intellectual property right of any third party — including using AI to paraphrase, restructure, or reformulate protected works in a manner that constitutes infringement. You are solely responsible for ensuring that prompts you submit and content you publish do not infringe third-party intellectual property rights;
• (c) gather, compile, or synthesise competitive intelligence about Gevu, its features, workflows, pricing, architecture, or business model, or to use AI-generated outputs to inform the development, improvement, or marketing of any competing product or service;
• (d) generate health and safety documentation — including RAMS, risk assessments, method statements, or COSHH assessments — that is deployed, published, or relied upon at any worksite without competent human review and written approval;
• (e) circumvent any legal, regulatory, contractual, or professional obligation, or use AI output as a substitute for compliance with any such obligation;
• (f) process personal data of any third party without a valid and documented lawful basis under applicable data protection law;
• (g) generate, distribute, or act on AI output in any regulated professional context as a substitute for advice from a qualified solicitor, accountant, health and safety professional, medical practitioner, or other regulated professional;
• (h) generate content for use in legal proceedings, regulatory submissions, or insurance claims without independent human expert verification; or
• (i) probe, reverse-engineer, extract, replicate, or attempt to derive the underlying AI model weights, training data, system architecture, or system prompts used in the Service.

Breach of any provision of this section constitutes a material breach of these Terms and may result in immediate termination without refund, reporting to relevant authorities, and civil or criminal liability.

14. Changes to AI Features and this policy

AI Features, the underlying models, output quality, daily token limits, and available functionality may change at any time, including as a result of changes made by the AI model provider. We will aim to notify you of material changes through the platform or by email. Continued use of AI Features after any change constitutes acceptance of the updated policy. Because AI model behaviour may change without notice from the model provider, we cannot and do not warrant that AI output quality or characteristics will remain consistent over time.

15. Contact

For questions about AI Features or this policy: support@seawolfservices.co.uk
Seawolf Facilities Services Ltd, registered in England & Wales (no. 15028189).

← Back to policies