The situation
The client was launching cold outreach on two channels at once: LinkedIn, through four sender profiles in HeyReach, and email through Apollo. They already paid for a tool that pushed replies into Slack, but they wanted something built around their own rules.
Two rules mattered most. Never contact an existing customer. And when someone says stop, stop everywhere, on both channels, right away.
The tools didn't make that easy. Each platform kept its own lead lists and its own record of who had opted out. Nothing told one channel what had happened on the other. A person who asked LinkedIn to stop could still get the next email.
What I built
Replies in Slack, with context. When a lead replies on either channel, a card lands in the sales team's Slack channel showing who replied, what they said, and a link straight into that conversation. It posts on the first reply only, so a back-and-forth doesn't flood the channel.
One-click Do Not Contact, with undo. Each card has a Do Not Contact button. One click removes that person from both outreach platforms and records them on a master suppression sheet the client owns. If someone clicks by mistake, Undo reverses all three. Every button click is signature-checked, so nobody outside the workspace can fake one.
Existing customers screened out automatically. When a company on the active-client list shows up in an outreach list, everyone from that company comes off, with no human step. Leads are screened the moment they arrive, not in a weekly cleanup.
Research that cites its sources. Reps get a short, dated, recent fact about each company to open with. The checks run in code, not on the AI's word: the source link has to load, mention the company, come from somewhere other than LinkedIn or a data broker, and fall inside the last 12 months. On a 100-company sample, 80 came back with at least one fact that passed, and about one in four had a real merger or acquisition.
I tested three ways to run that research. All three matched on quality, so I recommended the flat-priced option and kept a pay-per-use version built as a fallback. Researching individual people didn't earn its cost, so I recommended dropping it and using a free job-change flag the client's email platform already provides.
Vendor tools don't always tell the truth. So every add, remove and opt-out is confirmed by reading the record back from the platform, not by trusting the reply.
How I made sure it works
One of the platforms reports success on requests that change nothing, and takes a while to reflect its own updates. Another leaves out, in one of its lookups, the very fields the opt-out design depends on. Both would have become silent bugs. Instead, every write is confirmed by reading the record back.
Everything was tested on synthetic leads first. A separate audit workflow checks that no real lead was touched, and it runs before every demo. That testing caught six defects in the customer-screening flow before it went near real data.
Where it stands
The full chain is built and proven end to end in the client's live accounts on test records: reply, Slack card, Do Not Contact across all three systems, and undo. The client has a rehearsed demo of the whole story. Production launch follows their own campaign schedule.
The bigger result is a sales team that owns its rules. The suppression list lives in a sheet they control, the logic lives in a workflow they can read, and adding a third channel later means adding one more connection, not rebuilding the system.