1. What the AI features do
Message drafts: you choose what the message is for, how it should sound and how long it should be, and you can add one short line and choose to use your notes about the person. An AI model writes a draft that you can copy, change and send yourself.
Practise a conversation: an AI plays the person you want to talk to, from a few simple facts, so you can rehearse what to say. At the end, it can give you tips.
Reply help: when you ask for help to answer a message you received, you pick what the answer should do, for example thank, accept or propose a time, and an AI model writes a draft. We send the message you answer, from the words you chose to keep if that choice includes reply help, or what you paste, and you see it, with the credit cost, before anything is sent. You change the draft and send it yourself.
Placing your companies: when you start a scan, or after each import if you switch that on, an AI model places in one of our industries the companies of your network that our own information cannot place. It sees only each company's name and up to three words for the kinds of jobs people there do, never a person. A second, separate AI check must agree before we keep an answer. It uses credits only for the companies it places, and you can change any industry.
Finding new people: an AI model searches public web pages for people who fit one of your goals. Before you see a person, our software checks that the page names them and their organisation, and a second, separate AI check must agree. You open their page, and you decide whether to add them or write to them. On every plan, and without AI, “Find new people” also prepares searches that you open yourself (see the privacy notice).
Suggestions in the rest of the app, such as your daily steps and the best fits for a goal, are worked out by our own software, not by AI. Industries that AI placed stay in your workspace, where you can change them; suggestions do not use them yet.
2. Who provides the AI
We use two AI providers, Anthropic and OpenAI, with models we choose, and either may answer a request: the main one first, and the other only when the first refuses the request before doing any work (a request the first provider may already have worked on is never sent to the other). The same data is sent to whichever one answers; a request the first refused has already reached it, so then both received it, and while the main one is known to be down, a request goes to the other one first. For a draft, a reply or a practice conversation, “What we send” names both (“Anthropic or OpenAI”) before anything is sent. “Find new people” uses only the main provider. We do not train our own models. If we change a provider, we update this notice, the privacy notice and the sub-processors page.
3. What AI can and cannot do
AI output can be wrong, can sound sure when it is not, and can state things that are not true. It does not know the real person, their situation or your history beyond the few facts it receives. It is not legal, financial, career or other professional advice.
No person at TapForNetwork reads AI output before you do. For drafts, replies and practice, our software checks each answer for some problems before you see it, for example a web link, an e-mail address, a phone number or markup, and refuses it if it holds one; a refused answer costs you no credits. The company scan can only answer with an industry from our own list, and a second AI check must agree with it. “Find new people” shows a person only after the checks described above.
4. How AI output is marked
Every draft and reply is marked as made by AI, with a short note that invites a quick read before you use it. A practice conversation says that it is with a simulated person, not the real one, so you always know when you are talking to an AI. The results of a company scan and the people “Find new people” finds carry the same note.
A message you send based on a draft is your message.
These marks are shown on screen. The text you copy carries no hidden mark.
5. AI never decides about people
AI in TapForNetwork writes text for you to read and change, places companies in industries, and looks for public pages that fit a goal you chose. It makes no decision about you or anyone else with legal or similarly serious effects, and it does not read faces, voices or emotions. Do not use it to make decisions that seriously affect people; the acceptable use policy forbids that.
6. What is sent, and what is kept
For a draft, a reply or a practice conversation, you can see what will be sent, and to which provider, before anything is sent. We send only what the feature needs: the name you use for the person, their title and company, simple facts such as when you connected, the words of a goal you choose, what you type, and your notes about that person only when you tick “use my notes”. We do not send the e-mail addresses or profile links from your import, or your imported messages, except the one message you ask reply help to answer. For the company scan we send only company names and job-role words. For “Find new people” we send the kind of people you are looking for, picked from your goal, never your notes, your network or a person's name.
We do not save drafts or what you type in a practice conversation; we keep only a record of each AI request without its content. A draft stays in our server's working memory for at most 10 minutes, so that a request sent twice is charged once. The privacy notice, under “AI features” and “People found on the public web”, gives the details.
7. Limits and credits
Each paid plan gives a number of AI credits a month, and each feature you start shows its cost before it runs; the company scan uses credits only for the companies it places, never more than you have left. There are also daily safety limits for each account and for the service as a whole, so AI can pause for a while; unused credits stay until the end of your billing month. Nothing from a Free plan is sent to an AI provider.
8. Your part
We keep improving our AI, but anything it makes can contain mistakes, so give it a quick read before you use it: you decide what you use, and the messages you send are yours. Check a person's page before you write to them. Do not use AI drafts to pretend to be someone else or to mislead anyone, and do not present what an AI said in practice as the real person's words. Keep sensitive details about people out of what you type. The acceptable use policy applies.
If an AI answer is harmful, offensive or clearly wrong, tell us at support@tapfornetwork.com. Describe what happened, without copying sensitive details.
9. Changes to this notice
When we change what the AI features do, the provider or what is sent, we update this notice first. Each version has a number and the date it takes effect.
What changed in this version
- Corrects one sentence: suggestions do not use the industries that AI placed yet. Those industries stay in your workspace, where you can change them.