Outbound should keep moving after setup, without turning the founder into the scheduler.
Outbound software usually gives you a place to create a campaign, a place to watch it send, and a place to answer the replies.
Then it hands the actual job back to you.
you still decide who to contact.
you still write the messages.
you still keep the list alive.
you still notice when a sender breaks.
you still decide what to do with a reply.
you still remember to follow up.
you still build the next campaign when the first one runs out.
That is a collection of tools.
It is not an outbound company.
The campaign is not your job
Instaboost V3 starts with one-time business context.
You connect the accounts and give the system the outcome you care about.
The website does the heavy lifting during onboarding. Instaboost crawls up to 10 pages and extracts the ICP, offer, tone, and objections from the business itself.
Your website becomes the source for the agent's understanding of who you serve, what you sell, how you speak, and which concerns the buyer is likely to raise.
The final onboarding step is called Campaign, but that does not mean you become the campaign builder.
It means you describe the direction once: book discovery calls, re-engage old leads, share a lead magnet, or announce a product.
The agent turns that direction into the campaign and owns the campaign work.
It chooses the acquisition path, finds eligible leads, creates the sequence, writes the variants, chooses senders, applies the timing rules, and decides when the campaign can start.
The human input is the business decision.
The system owns the campaign work.
That distinction is the entire point of V3.
The agent becomes the outbound team
The V3 loop runs in one order.
The agent can discover leads, build campaigns, write copy variants, make the sender choice, send, follow up, reply, book meetings, and optimize the next move.
scrape and discover leads.
build campaigns.
write copy variants.
pick senders.
send.
follow up.
reply.
book meetings.
optimize.
Then it starts the next useful action without waiting for you to reopen the dashboard.
A campaign running low on leads does not become a silent emergency. The agent can refill the runway with eligible leads and keep the active campaign moving.
That top-up is part of campaign ownership. It does not quietly rewrite the strategy or restart the machine. It adds eligible prospects inside the existing policy and records what happened.
The same principle applies when a campaign finishes. Instaboost can suggest an adjacent ICP, a re-engagement path for people who went quiet, or a lead-magnet follow-up for people who engaged without booking.
In Autopilot, an eligible next campaign can move from suggestion to launch inside policy.
The user does not log in to create that campaign.
The user does not log in to keep the old campaign alive.
The user does not log in to move every reply to the next stage.
If the user has to log in to do manual work, V3 failed at the job it was built for.
Four channels, four native loops
Instaboost runs email, LinkedIn, X, and Instagram.
Those channels belong in one product, but they should not be flattened into one generic automation script.
Email has sender health, DNS, warmup, unsubscribe handling, bounce detection, and deliverability checks.
LinkedIn has its own account session, connection flow, messages, and follow-ups.
X has its own account health, pacing, daily limits, and send path.
Instagram has its own connected-account path, lead identifiers, DM worker, account limits, and reply polling.
The campaign can use shared business context while each channel keeps its native sending and conversation mechanics.
That is important because the agent is responsible for the outbound outcome, not for pretending every platform behaves the same way.
The conversation stays where it started.
The business context stays useful across campaign creation.
The system does not invent a cross-channel thread that does not exist.
Autopilot is the default operating model
V3 has three autonomy modes because trust should be adjustable without turning the product into a manual workflow.
Autopilot is the default. The agent creates campaigns, chooses eligible work, sends, follows up, responds, books, tops up, and continues the next eligible action automatically.
Guardrail keeps the agent active but pauses before sensitive moves that need confirmation.
Pilot is the deliberate opt-out from sending. The system prepares the work and orchestrates the campaign, but the user reviews every send before it leaves.
Pilot is not the normal way to use V3. It is for someone who wants to inspect every message while the system handles everything around the send.
The goal of the dial is not to make the user operate three different products.
The goal is to let the user choose how much sending control to retain while the rest of the outbound work remains autonomous.
Trust is enforced by the system
Set-and-forget does not mean ungoverned volume.
The agent cannot launch past hard rails and safety controls.
DNS and deliverability checks sit before sending.
Warmup state matters.
Compliance and unsubscribe rules apply to the message path.
Verified sending capacity matters more than the number of connected accounts.
Sender health and session health can hold an account back.
Pacing and channel limits are enforced per channel.
Pattern-guessed email addresses are blocked instead of being treated as leads.
Suppression and deduplication rules protect the campaign from contacting the same person through overlapping work.
A global pause can stop the outbound workers when the user needs everything held.
These are system-enforced controls, not reminders placed next to a launch button.
A connected account is not automatically an eligible sender.
A guessed address is not an acceptable import.
A campaign that fails a hard gate does not get a hidden launch-anyway path in Autopilot.
New sending ramps conservatively. A clean health window can promote the campaign after 72 hours. A regression holds the current tier and creates an anomaly entry instead of silently increasing volume.
The kill switch is always available.
Pause everything means pause everything across the outbound workers. Resume continues from the held state instead of forcing the user to reconstruct the campaign by hand.
That is what makes strict autonomy usable.
Timing beats another list
A static list tells you who might fit.
It does not tell you when the conversation makes sense.
V3 adds an intent layer around free-source signals such as funding, hiring, executive changes, website visits, content engagement, competitor interaction, company news, and technology adoption.
The agent can score a lead against the campaign ICP and use an approved signal to prioritize or enroll the lead.
The signal is not a promise that the person will buy.
It is a reason to consider timing before sending another generic message.
The same attention goes into local timing.
When the lead's timezone is known, the system schedules around the prospect's local working hours instead of only the sender's clock.
When an out-of-office reply includes a return date, the system can pause that follow-up path until the named date.
The user does not need to create a reminder and remember to restart the conversation later.
These mechanics are easy to overlook because they are quiet.
They are also the difference between an agent that runs a continuous system and a batch job that fires messages until someone notices the problem.
You teach the system in plain language
The product should not require a founder to translate every preference into a nested rule tree.
You can teach the agent with a sentence:
“if someone asks about pricing, send this answer.”
“do not respond to complaints. flag them for me.”
“if they are interested but not ready, follow up after the window I set.”
The teach-the-agent flow turns those sentences into reply policy.
The agent combines that policy with confidence checks. When intent is unclear, it does not invent certainty. It takes the conservative path or escalates the conversation.
The AI command bar gives the user the same direct interface for system actions.
“pause the email campaign”.
“how many replies this week”.
“draft a reply to the Acme lead”.
“reconnect my LinkedIn account”.
The command bar can answer, draft, navigate, or take a permitted action. Destructive actions stay outside the command registry, and permitted actions are recorded in the activity history.
The user should be able to state an obvious request without learning where every setting lives.
When something genuinely breaks
An autonomous system should not turn every small problem into a human task.
Reply drops, bounce spikes, sender problems, and ramp regressions become plain-language anomaly alerts with the reason and the next action.
Autopilot can apply eligible fixes within policy.
Guardrail and Pilot can surface the fix for review.
The blocker center is reserved for the work that truly needs a person: reconnecting an expired session, fixing DNS with provider access, changing billing or plan state, connecting a new channel, or approving review-mode work.
When something genuinely breaks, the system pings the user once with the next step. One ping, one fix, and the loop is back to running without the user.
That is the correct role for a notification.
It should return ownership of one blocked dependency, not make the user resume ownership of the entire outbound system.
Outcomes replace activity
A campaign can optimize for more than one thing.
The outcome toggle lets the user set the current priority: more replies, more meetings, or less spend.
That direction changes how the agent evaluates reasonable next actions.
A meeting-focused campaign should not optimize only for cheap activity.
A spend-sensitive campaign should not behave as if every action has infinite budget.
When a lead shows buying intent, the agent can move toward calendar booking.
Meeting attribution can connect the meeting back to the campaign and channel that produced it. The system can attribute a calendar booking to the work that created it.
The system can learn from replies, booked calls, and winning variants.
That evidence can refine the ICP, improve the next sequence, change the campaign direction, and suggest the next eligible campaign.
“make another like this” carries the useful structure of a winning campaign into a fresh one without pretending the new audience is identical to the old one.
This is what learning should mean in an outbound product.
It should change what the system does next.
The dashboard is not the product
The home screen should make status easy to understand.
It should show system health, current blockers, recent activity, and the outcome being optimized.
The activity history makes the work explainable. The user can see why a sender paused, why a lead was added, why a variant changed, or why a campaign was held.
That is visibility, not a new job description.
The user does not log in to perform the work or become a full-time campaign coordinator just to feel informed.
The product exists to remove the daily outbound checklist, not to give that checklist a better interface.
Why I built it this way
I did not want to build another sequencer with an AI button on top.
I wanted the system to understand the business, create the work, protect the sending accounts, handle the normal conversations, book the right meetings, learn from what happened, and keep going.
That means the agent owns campaign creation too.
It owns the list.
It owns the variants.
It owns the sender decision.
It owns the send, the follow-up, the reply, the booking, and the next campaign.
The user provides the business context and the outcome.
Then the user walks away.
Meeting results still depend on fit, timing, message, and offer. Instaboost does not guarantee a result.
It owns more of the work that has to happen for a result to be possible.
That is the set-and-forget outbound loop I wanted to exist.
The Bottom Line
You should not have to run outbound to get the benefit of outbound.
You should set the business context once, choose the outcome, and let the agent create and run the work from discovery through booked meetings, with the goal to book more meetings.
Meeting results still depend on fit, timing, message, and offer.
Join the V3 waitlist at getinstaboost(.io).