AI Agent Terms

1. What the AI Business Agent is

1.1 The AI Business Agent is an assistant inside each business workspace. It is included with every business subscription. You ask it in your own words; it works out what you need and uses tools that are limited to the business you are working in. It cannot see or change your other businesses.

1.2 Standard actions (included in your subscription, no credits). Within the current business, the agent can:

• read the business record: your assessment, the selected opportunity and research summary, offer, confirmed facts, pipeline counts, website state, money summary and alerts;
• read your potential and existing customers in summary form (first names, stage, channel, follow-up date, and the latest enquiry text with contact details removed);
• build or revise your website and save it as a new version (not publish it);
• save drafts for you to review: messages, proposals, plans, marketing drafts and analyses;
• add a potential customer whose name you gave it, and schedule a follow-up on an existing contact;
• build the business's customer-acquisition strategy, find organisations in public business listings and save candidates you choose to your CRM (Terms, section 3.5);
• remember notes about the business, such as decisions and preferences, marking whether you stated them or the agent assumed them;
• ask you one question when it cannot decide something important on its own.

1.3 Actions that always wait for your approval. The agent can ask you to approve, but cannot itself:

• publish a website version (Website and Domain Service Terms, section A2);
• send a message. A message is sent only from a mailbox you connected, to one contact of that business, after you review it and approve. Nothing is sent until you approve, and a message is recorded as "sent" only after your email provider accepts it;
• run a credit-metered task (section 3).

1.4 What the agent never does. It never buys anything, never spends advertising money, never buys or registers domains, never deletes your records, never makes binding commitments for you, and never claims that an external action happened (for example that a message was sent or a payment was made) unless the Service confirms it. It is instructed not to invent credentials, certifications, awards, client names, testimonials, statistics or results.

1.5 Record changes listed in section 1.2 (new website versions, drafts, potential customers, follow-ups, saved candidates and notes) are made without a separate approval for each one, because they stay inside your workspace and can be edited or removed by you.

1.6 You do not need to use the agent to use WhatTrust or to follow your execution plan.

2. How the agent works and its limits

2.1 Model provider. The agent's reasoning uses language models provided by OpenAI. To answer you, the business record, your request, the conversation and the results of the tools in section 1.2 are sent to OpenAI for processing. We ask OpenAI not to store these requests for later retrieval (store: false). We do not use your data to train AI models. The Privacy Policy, section 4, explains this further.

2.2 When the agent is available. Live AI is enabled only when the owner has approved it for the environment you use. When no model is available, the Service says so; it does not present rules-based output as the agent's reasoning.

2.3 Retrieved content is untrusted. Content in your business record, messages from your customers, web pages and other tool results are treated by the agent as information, not as instructions. This reduces, but cannot remove, the risk that misleading content affects an answer.

2.4 AI can be wrong. The agent can misunderstand a request or produce inaccurate, incomplete or inappropriate output, even when it sounds confident. Recommendations and "next best actions" are suggestions for you to evaluate. You must review every draft, website version and record change before you rely on it, and every message before you approve it (Terms, section 9).

3. Credits

3.1 What credits are for. Some advanced AI tasks are metered and use AI agent credits, for example a full marketing plan, a campaign brief, a competitor analysis or a performance report. The agent offers a metered task only when you ask for such a deliverable. Standard actions (section 1.2) do not use credits. Your subscription does not include unlimited AI usage.

3.2 Introductory credits. When an eligible business is first activated, it receives a one-time allocation of 20 introductory credits. They:

• have no cash value and cannot be exchanged for money;
• are not refundable;
• do not renew with the subscription and are not granted again if you cancel and reactivate;
• do not currently expire. If an expiry is introduced, it will be stated in these terms before it applies to credits granted afterwards.

Businesses that existed before 2026-10-28 do not receive introductory credits retroactively.

3.3 Top-ups. You can buy credits in prepaid top-ups of US$20 for 2,000 credits (plus any applicable tax). Each top-up needs its own separate, unticked checkbox that states the exact amount you will be charged and that it is a one-time charge. Nothing is charged until you tick it and confirm, and we record that consent. Credits are added only after Stripe confirms payment. You need an active business subscription to buy a top-up. There is no AI agent subscription and no automatic recharge.

3.4 One balance. Introductory and purchased credits are held in one balance for your account, usable for any business that has an active subscription.

3.5 What credits are not. Credits are not money, not a gift card, not a deposit and not transferable, and they cannot be exchanged for cash except as stated in section 8.

3.6 Expiry of purchased credits. Purchased credits stay in your account until used. They do not expire.

3.7 If your subscriptions end. Credits remain recorded, but metered tasks cannot run until a business subscription is active again. Unused credits are not refunded when your subscriptions end, except as stated in section 8 or required by law.

4. Cost shown first; confirmation for every metered task

4.1 Before a metered task runs, we show its credit cost and your balance. The task runs only after you confirm that cost. If the cost changes before you confirm, you are asked again.

4.2 The confirmed credits are deducted when the task starts.

4.3 No overage. The agent never spends more than your balance. If your balance is too low, nothing runs until you top up. There are no automatic or hidden overage charges.

4.4 System errors cost you nothing. If a metered task fails because of a system or provider error, its credits are restored automatically.

4.5 Daily limit. A daily limit on credit use per account may apply; the Service tells you when it is reached.

5. Service limits (fair use)

5.1 Each business has an internal monthly limit on the cost of AI and data-provider work done for it. As the limit is approached, optional and expensive work is reduced first. When it is reached, advanced AI work for that business (including the agent's reasoning, discovery and metered tasks) may pause until the next billing month, even if you still have credits. Buying credits does not raise this limit. The Service tells you when the limit stops a task, and no credits are taken for work that did not run.

5.2 Your records, website, CRM, drafts and other features that do not need AI or paid data remain available while the limit applies.

6. Approvals and standing authorisations

6.1 Per-action approval. Publishing, sending and metered tasks each create an approval request that shows what will happen. Only you can approve it, while your subscription is active. An approval is used once. We record each approval.

6.2 Standing authorisations. Some actions of the earlier AI Agent task system also require a standing authorisation for that type of action, which you grant by accepting its exact text. The text is shown in your interface language, and we record the exact text you saw, its fingerprint (SHA-256) and the language. You can revoke an authorisation at any time; a revoked authorisation is never reactivated, and you must grant it anew.

6.3 Never purchases. The agent can never make a purchase or commit you to a payment, with or without an authorisation. This includes buying advertising, domains, subscriptions or goods. Purchase requests are always refused.

7. Your responsibility for what you approve

7.1 You are the sender and publisher. When you send, publish or otherwise use content the agent drafted, or approve an action, you are responsible for it, as if you had written or done it yourself.

7.2 Outreach compliance. You are responsible for complying with anti-spam, telemarketing and electronic-marketing laws (for example CAN-SPAM, the TCPA and state telemarketing laws, the GDPR and EU/UK ePrivacy and PECR rules, CASL, Australia's Spam Act, and Korean, Turkish and Gulf-state electronic-communications rules), data-protection law for any contact data you use, and the rules of your email provider and any connected platform. This includes having consent where it is required, identifying yourself, honouring opt-outs and adding people who must not be contacted to your do-not-contact list (Terms, section 6.6).

7.3 No fake reviews or deceptive claims. You must not use the agent to write, buy or publish fake, AI-fabricated or undisclosed-incentive reviews or testimonials (see, for example, 16 CFR Part 465), to impersonate anyone, or to make unsubstantiated earnings, health or performance claims. The agent is instructed not to invent such content for your website or drafts.

7.4 Finding customers. When the agent finds organisations for you, it uses public business listings and is a tool you direct. WhatTrust does not supply, sell or guarantee leads or customers, and the agent does not contact anyone without your approval of each message.

7.5 Our liability for actions you approved is limited as described in Terms of Service section 13, subject to its carve-outs.

8. Refunds of top-ups

8.1 Used credits are not refunded, except credits taken by a system error, which are restored automatically (section 4.4).

8.2 Introductory credits are not refundable (section 3.2).

8.3 Billing problems. We refund duplicate, wrong-amount and unauthorised top-up charges.

8.4 Unused purchased credits. If none of a top-up's credits have been used, we refund it if you ask within 14 days (to be confirmed) of the purchase.

8.5 What a top-up refund or dispute does.

• It applies only to that top-up. That top-up's unspent credits are removed from your balance; your balance never goes below zero.
• Credits already used are not restored.
• A top-up refund or dispute never affects a business subscription.

8.6 Statutory rights described in the Subscription, Refund and Cancellation Policy, section 9, apply to top-ups too.

9. Records we keep

We keep a record of:

• each top-up payment and each introductory-credit grant;
• each metered task you confirmed, with its credit cost;
• each agent run, the steps and tools it used and its result;
• each approval of a publication, message or metered task;
• each standing authorisation you granted or revoked, with its exact text, fingerprint and language;
• each action refused, and why.

Legal and consent records are append-only. You can download your credit ledger, authorisations, agent actions and outputs with Export my data; some agent records are not yet included in the export (Privacy Policy, section 10). Records are kept after account deletion for the periods in the Privacy Policy.

10. Changes and suspension

We may change task types, credit costs of future tasks and the agent's capabilities. The price of credits you have already bought will not change. We may pause the agent for a business that breaches the acceptable-use rules (Terms, section 7), for security reasons, or under section 5.