AI Replies and Handover

Gene can reply to leads on its own, and a person on your team can step in at any time. Use this page to see who is answering a conversation, take a thread over, and hand it back to Gene.

The Gene switch#

Every conversation has its own Gene switch, beside the Send button in the composer. It decides who answers that one contact and never changes any other conversation.

A new WhatsApp number you have never spoken to is held back too. It lands under New senders, and Gene does not auto-reply until you classify the sender (see New senders).

Take over or hand back a conversation#

• Turn Gene off for this conversation (the Gene switch, beside Send): hands the thread to your team. Switching off asks you to confirm first.
• Turn Gene back on (the Gene switch): hands the conversation back to Gene. This is immediate, with no confirmation.
• Turn off Gene and send (confirmation dialog): when you press Send while Gene is on, choose this to switch Gene off for this conversation and send your message in one step. Keep Gene on cancels.
• Rerun the AI on the last message (More actions, the three dots in the thread header): Gene takes another pass at the most recent message and replies shortly. Use it when a reply did not land, or when you changed something Gene should now know.
• Rate a Gene message (thumbs up or down under an AI bubble): sends feedback to your team for review, so replies improve over time.
• See who sent each outbound message (hover a sent bubble): a Gene mark means Gene sent it, a person icon means a teammate sent it. The send time and delivery status sit next to the icon.

What Gene knows about how a lead reached you#

On every reply, not only the first, Gene reads how the lead reached you: the source and channel, the campaign, ad set, ad and form, the click-to-WhatsApp, Messenger or Instagram ad they tapped before writing, the page they came from and its UTM tags, the contact's referral label, and who wrote first. So on the tenth message Gene still knows which ad, and so which property and price, the lead is asking about. When the ad, campaign or form names a property or project, Gene answers about that one first, matched to the listing in your knowledge, and does not start over by asking what the lead is looking for or suggest other properties before they ask.

• A lead you already know who taps a different click-to-WhatsApp ad is asking about that ad. Gene switches to the newest ad they tapped, so a lead who wrote about one villa and comes back through another villa's ad gets answers about the second one. The first ad they ever came through is still kept on the contact.
• Contacts transferred from GoHighLevel or Gene 2.0 keep which ad they came from (see Transfer), so Gene reads the same thing for a lead who started the conversation in the old CRM.
• The instructions you wrote for the lead's source and for its Meta ad form (see Flows > Inbound) also stay in front of Gene for the whole conversation. Only First-reply guidance in AI > Sales agent is limited to the first reply.
• Gene knows which messages in the thread were sent by an automation (a campaign, an automatic follow-up, a welcome message, the follow-up to a repeat inquiry, an appointment reminder) and which a teammate wrote. When the lead answers one of those, Gene reads the answer against that message, and it never replies as the teammate. On email, Gene goes by the message the lead's reply names, even when something else was sent after it.
• For website and Gene form leads, UTM tags in the address of the page the form was filled in on are saved on the contact, even when the form had no hidden field for them. For lead webhook leads, UTM tags and a page address sent as fields (utm_campaign, Utm Medium, page_url) are saved the same way.
• The source Gene reads is the one the lead's latest inquiry came through. A webhook lead is named by its connection ("Landing page Meta Ads (webhook)"). A contact first added by hand or imported, who later fills in your website form, is read as a website lead, with how the record was first created noted beside it.

When Gene hands over on its own#

Gene switches itself off automatically (the same as flipping the switch) when:

• The lead books an appointment. A person takes the meeting, so Gene steps back, moves the lead to the Selection stage, and notifies the assigned salesperson (in-app and by email unless they changed it in Settings > Notifications).
• The lead asks to speak to a human. Gene hands the thread to the team. A lead who only asks whether they are talking to a person, a bot or an AI ("am I talking to a human?") is asking a question, not for a person: Gene answers it the way your Extra rules say and keeps helping. Gene also does not hand over on its first reply just because it could not answer something; it answers what it can and asks one useful question first, unless your rules say to hand over.

A detail Gene could not find, or two sources that seem to disagree, is not a reason to hand over on its own. Gene first reads the project record, your FAQ and your uploaded files and answers with what they show. A fact that only a brochure or price list has is still an answer, and a handover date inside a stated quarter is the same fact, not a conflict. Gene says your advisor will confirm anything still missing, notes the gap and keeps helping.

In both cases the conversation is assigned to its owner and Gene keeps the full history. You can turn Gene back on whenever you want and it picks up with full context.

Gene hands a conversation over once. While a handover from the last 24 hours is still waiting on a person, it does not hand over again, so your team is not notified twice for the same lead.

When Gene tells a lead when they will hear back, it goes by your team's own timezones: the timezone each active teammate set with their working hours (Settings > Calendars > Availability), or the one on their My Profile if they set no hours. A handover outside working hours in every one of those timezones is answered the next morning. Gene never says your team is based in a city or country, because the company location on the AI page is where the business is registered, not where everyone works. If nobody has set a timezone yet, Gene only says your team will reply as soon as they are online.

Gene always knows who a lead is assigned to, so it can tell the lead who will take over ("let me connect you with Omar"). It refers to that person by first name (or by their role if it does not know the name), in the handover reply and every reply after it, and never as "he", "she", "him" or "her". The reason Gene notes for your team when it hands over stays with the team: if its reply to the lead repeats that note, the repeated sentence is taken out before it sends.

When a message fails to send#

Gene also switches itself off for a contact when a message to them did not send. Rather than keep writing into a channel that just failed, it leaves the conversation to you:

• The Gene switch reads off, and the thread says that the last message did not send.
• Turning Gene back on resumes it straight away. Failures from before that no longer hold it back.

Some failures say nothing about the channel, so they leave Gene on:

• a closed 24-hour window
• a file that could not be sent
• Meta holding back a marketing message to that one person (its per-person marketing limits, or the lead having turned off marketing messages from you)

In those cases the failed message says what happened, and Gene answers as soon as the lead writes. When Meta's per-person limit held back a template, Gene also retries it a day, three days and a week later, with a note in the conversation each time (see When Meta holds a template back).

What happens if nobody answers a handover#

A handed-over conversation stays in the Open tab until a person replies, and Gene keeps chasing until one does:

1. After 3 hours, the assigned person gets a second alert, in-app and by email by default (each person picks the channels for Unanswered hand-over in Settings > Notifications).
2. At your handover window (Settings > AI > Follow-ups, 24 hours by default), their manager is told, so it is no longer one person's private to-do.
3. At twice that window, the lead is posted in the team WhatsApp group, where the whole team sees it.
4. After 72 hours of total silence, Gene starts answering the lead again rather than leaving them with nobody.
   • You get a note on the contact and a message in the group when this happens.
   • You can take over at any time by replying.

Any of these stops the chain immediately: a human reply, snoozing an opened conversation, or ticking off the handover task.

Both windows are configurable. Setting Gene takes the lead back to 0 hours switches step 4 off entirely.

New senders#

A WhatsApp number you have never spoken to shows a banner at the top of its thread. Gene does not auto-reply until you pick one:

• Add as lead: makes them a real lead so Gene can engage.
• Block: adds the number to your blocklist and removes the contact. It asks for a second confirmation because it cannot be undone.
• Not now: just closes the thread for now.

To find unknown senders waiting to be classified, use the New senders filter in the conversation list.

Test your agent#

The flask icon in the Inbox header opens a practice conversation with your own agent. Nothing is delivered on any channel.

Start a test#

1. Click the flask icon in the Inbox header.
2. Pick a scenario:
   • Lead writes first: a lead messages you first, or
   • Website form: a lead fills in a form, and the agent opens with the welcome message it would really send.
     • You fill in the lead's form yourself: First name, Budget, Timeline, Area and Notes. The chips above the fields fill them with a ready-made lead, and every field stays editable, so you can change or clear anything a chip filled in. Only what you see in the fields is sent with the test.
     • If your website form source has a first message typed in its Welcome flow tab, that message opens the test as written, with the lead's name filled in. No AI writes it and it uses no test reply.
     • On email, the source's first email opens the test instead, under its own subject. A designed email shows as its text.
     • If the first message is empty, Gene writes the opener for this lead from the form answers and the source's instructions, as it would for a live lead. If Gene cannot write it, the short generic welcome a live lead would get goes out instead (also free).
3. Pick the channel in the select beside the scenario tabs:
   • Lead writes first: WhatsApp, Email, Instagram, Messenger or Telegram.
   • Website form: WhatsApp or Email.
   • SMS, iMessage and website chat are not offered.
4. Pick how the lead opens. A badged TEST lead appears in your Inbox, on the channel you picked.
5. Play the lead from the bar above the composer. The bar names the channel ("Practice lead on Email").

The composer itself still speaks for your team, so taking over mid-chat is part of the test.

Choose or write the lead's first message#

On the Lead writes first tab you have two options:

• Pick a starter: one of the ten ready-made openers (asking the price, wanting a viewing, saying it is too expensive, writing in Spanish, asking you to stop, and more).
• Write your own (above the starters): type the lead's first message yourself, the way your own leads write it. Any language works; your agent answers in the language the lead writes in.

You can also add a first name and a phone number (pick the country, then type the whole number). Both are optional.

• The number lets you try the check your agent makes before offering times ("I see a Dutch number, are you in that timezone now?", see Appointments).
• Nothing is sent to that number, and it is never saved as the lead's phone. A number the lead gives your agent during the test is kept off the Phone field the same way.
• Your draft stays in the box when the popover closes, so trying the same message from another country is one edit.

Start test opens the same kind of test a starter does.

Replies on each channel#

Your agent writes for the channel the test runs on, exactly as it would for a live lead there:

• On WhatsApp, Instagram, Messenger and Telegram, replies are short chat messages, formatted the way that app shows them.
• On email, each reply greets the lead and signs off, and it joins the thread under "Re:" and the subject of the email before it.

Nothing is sent on any channel, and the practice lead has no real phone number, email address or account.

What happens during a test#

The agent answers for real, against your real properties, FAQs and files. So the CRM reacts the way it would with a live lead:

• the contact fills in
• the temperature turns
• the stage moves
• a property the lead names is linked under Interested in (with property links on, see Leads interested in a property)

Actions that would reach outside the practice lead (booking, sending documents, portal access, handover) are simulated. Each one shows in the test chat under the reply, marked as a test:

• a file card with the name of the file a real lead would get (if sharing is off for that file, the card says a real lead would not get it)
• a note when the agent would have handed over to your team, with its reason
• a note when it would have booked, moved or cancelled a meeting

Your agent treats each of these as done, as it would with a live lead. A file the practice lead already got is not sent again unless the lead asks for it again, and a handover is not repeated while the last one (within 24 hours) is still open.

Several messages in a row#

The lead box never locks. Send as many messages as you like, the way real leads write. Your agent answers everything sent before its reply lands in one reply, as it does with live leads: it waits a few seconds after the last message, and if the lead writes again while it is still writing, it starts over and answers them together. One batch is one test reply.

Long replies#

A reply that needs several tool calls (searching your properties, checking times) can take a few minutes. The chat shows Gene thinking until the reply lands, so there is no need to send the message again. If one step runs slow, Gene stops there and writes the reply with what it has already found, rather than leaving the lead with nothing. If you leave the chat, the reply still arrives and is there when you come back. If a reply takes too long to finish, the lead's message says "No reply: the run took too long. Send it again." and that test reply is not counted or charged.

What Gene saw#

Under each reply, What Gene saw opens what went into it besides your Settings > AI: the lead's messages it answered, how many messages from before them it read, what it knew about the lead (form answers and contact details), how the lead reached you, what the lead is replying to (an automation or a teammate's message), what it had noted earlier in the chat, your first-reply guidance and the lead source's instructions, and the tools it used. First-reply guidance and lead source instructions are always listed: "None set" when there are none. Past the first reply, First-reply guidance shows a note that it is only used on the first reply; lead source instructions are used on every reply. For a website form test, it shows what the welcome message was written from.

On every channel except email, a reply or a welcome message in several paragraphs shows as several bubbles, one per paragraph, exactly as a live lead would receive it (see How messages are split). An email stays one message. Older test replies saved as one block show split the same way.

Each start opens its own chat, so a form lead and a lead who wrote first, or the same lead on WhatsApp and on email, can sit side by side and be compared. Each test chat deletes in one click.

What tests cost#

Note

50 test replies a month are free, per workspace, resetting on the 1st. Past 50, a test reply costs 1 credit, the same as a reply to a real lead.

• The lead bar and the Test your agent popover both show how many free replies are left, and that a reply costs 1 credit after that. The count updates after each reply.
• Gene staff testing on your workspace are free and are not counted against your 50.
• Translating a test chat is free, whether it is the automatic translation of the lead's messages or one you ask for.
• You are told once in the conversation when you cross over from free to paid.
• A reply that shows as several bubbles is still one test reply.
• A reply your agent fails to write never costs anything.
• A reply that repeats the previous one never costs anything either. Gene redrafts once; if it still repeats, the reply is not shown, the test reply is not counted or charged and a ticket is opened for the Gene team.
• A website form test's welcome message only counts when Gene writes it. A typed first message, or the generic welcome, is free.
• If the free 50 are gone and there are no credits left, the test is refused rather than answered.

Popups and dialogs#

• Gene is on for this conversation. Opens when you try to send while Gene is on. Turn off Gene & send switches Gene off here and sends; Keep Gene on cancels. Internal notes are never blocked by this.
• Turn Gene off for this conversation? Confirms before switching Gene off. Gene stops replying here until you turn it back on.
• New sender banner. Sits above an unknown WhatsApp thread with Add as lead, Block, and Not now. Block asks for a second confirmation because it cannot be undone.

Good to know#

• There is no separate filter for "Gene on" versus "Gene off". The switch state lives on each conversation.
• Turning Gene off never loses anything. The full history stays, so you can hand a thread back to Gene whenever you want.
• The send hold exists so you and Gene are not typing over each other. When you press Send with Gene on, it offers to switch off and send together so the handover is clean.
• Hovering a sent bubble is the fastest way to confirm whether a reply came from Gene or a teammate.