myS1 includes AI features. The main one is Ask Otto, the assistant you can ask about your business data. This page explains how these features handle your information. It is a description, not a contract. The binding terms are in the Terms of Service and the Data Processing Addendum.
1. The short version
- Ask Otto works with the access of the person asking, and nothing more.
- To answer a question, the records Ask Otto looks up are sent to a third-party AI model provider.
- Some contact identifiers in the question you type are masked by default. The records Ask Otto looks up are not masked.
- Ask Otto proposes actions. A person approves them.
- Your administrator decides who may use AI and how much it may be used.
2. What Ask Otto can read
Ask Otto reads only what the user who is asking is permitted to see. It looks up records through the same checks as the rest of myS1:
- it stays inside your workspace;
- it can only use the modules your plan includes; and
- it can only read what that user’s own permissions allow.
Ask Otto does not get a copy of your database. It looks up what a question needs, one lookup at a time, and each lookup is limited in size.
3. What is sent to the AI model provider
Answers are generated by a third-party AI model provider. For that to work, content has to be sent to the provider. When a user asks Ask Otto a question, the following is sent:
- The question and the conversation so far.
- Context about the request. This includes the user’s name, the page they are on, any text they highlighted and the key details of the record they have open.
- The records Ask Otto looks up to answer. These are sent as they are stored. Depending on the question, they can include names, company details, deal values, notes, the content of emails, and the contact details inside those records.
We say this plainly because it matters: when you ask Ask Otto about your business data, the data it looks up for that question does reach the AI model provider.
4. What is masked, and what is not
Before a question leaves myS1, some contact identifiers are replaced with placeholders. The real values are put back on our side when the answer returns.
Masked by default, in the question you type and the conversation:
- email addresses;
- phone numbers in common formats;
- South African ID numbers; and
- card numbers.
Not masked by default:
- names of people and companies, amounts and other free text in your question;
- the records Ask Otto looks up to answer, including any email addresses, phone numbers and other identifiers inside those records; and
- phone numbers written in a format the masking does not recognise.
The setting can be changed. Masking is a workspace setting. A workspace administrator can change it, including turning it off.
5. Actions need a person’s approval
Ask Otto cannot change your data by itself. When it suggests an action, such as updating a record or sending an email, it shows a proposal. Nothing happens until a person decides.
- Sending an email and scheduling a meeting must be approved every time.
- A proposal that nobody answers expires.
- Where the app offers the option, a user can choose to stop being asked about a specific kind of low-risk change. That choice belongs to that user, can be withdrawn at any time, and is never available for sending emails, scheduling meetings or other actions that cannot be undone.
The person who approves an action is responsible for it, so read the proposal before you approve it.
6. Accuracy
AI output can be wrong, incomplete or out of date. Ask Otto checks the figures in its answer against the data it read and adds a note when it cannot find a figure there, but this does not guarantee that an answer is correct. Review what Ask Otto tells you before you rely on it, and check anything important against the record itself.
Do not use AI output as the only basis for a decision that has legal consequences for a person or affects them to a substantial degree.
7. Content from outside your workspace
Some of the records Ask Otto reads contain text written by people outside your business, such as incoming emails. That text can contain instructions aimed at an AI assistant. myS1 marks such content as untrusted and scans it for known patterns of this kind. If it finds one, it flags it and treats any outgoing action proposed while answering that question as higher risk. This reduces the risk. It does not remove it, which is one more reason actions need your approval.
8. What is logged and stored
- Conversations. Ask Otto conversations are stored in myS1 so that the user can return to them. They are private to that user and are kept for a limited period set by retention settings, which an administrator can shorten.
- Usage. Every AI call is logged with the user, the feature, the model used, the amount of usage and the cost. Administrators can see AI usage for their workspace.
- An audit record for Ask Otto. myS1 keeps a separate record of what Ask Otto looked up and who approved which action.
9. Controls your administrator has
- Who may use AI. An administrator controls which users have permission to use AI features.
- Spending. An administrator can cap AI spend for the workspace.
- Masking. An administrator controls the masking setting described in section 4.
- First use. Each user must read and acknowledge an AI notice in the app before using Ask Otto for the first time.
If you do not want content from your workspace sent to an AI model provider, an administrator can withhold that permission so that the AI features are not used.
10. Other AI features
Other parts of myS1 also use AI, depending on your plan. Examples are record summaries and suggested next steps in myS1 CRM, drafting help in Email Studio, building an automation from a description, and AI steps inside automations.
These features also send the content they work on to the AI model provider. The same workspace masking setting applies to the text these features send. It covers only the identifiers listed in section 4. AI usage by these features is logged in the same way as Ask Otto.
AI steps inside an automation run without a person present. Design those automations with care, and follow the rule in section 6 about decisions that affect people.
11. The AI model provider
The AI model provider is a third party that processes content for us in order to generate responses. It acts as our sub-processor under the Data Processing Addendum, and the Sub-processors page describes the sub-processors we use.
The provider may process content outside South Africa. Our Privacy Policy and the Data Processing Addendum explain the basis on which personal information is transferred to other countries.
12. Questions
If you have a question about how myS1 uses AI, email info@mysolutions.co.za.