Revenue operations

Outreach at scale, with a human gate on every send.

An outreach orchestrator that researches prospects, drafts in the founder's written voice, and sequences follow-ups across LinkedIn and email. Nothing is sent until a person approves it.

Shipped to a client
DomainRevenue operations
ChannelsLinkedIn and email
StackPython · FastAPI · Claude
GateEvery message approved by a person
523Prospects prescreened
165Drafted
116Sent after approval
45Stopped at the human gate
10Replies

Figures from ResonAi's own pipeline, the first place the system ran.

01
The problem

For a small firm, outreach competes with the work that pays for it.

Research, first-contact drafting, channel choice and follow-up timing are each manual and judgment-heavy, and each loses the calendar fight against client delivery. A strong week of prospecting is followed by three weeks of delivery, follow-ups quietly expire, and the pipeline resets.

Good first-contact messages are not a mail merge. Written to one named reader, with the right opening and the right ask, they take time. Done as a template, they read as one.

Pain pointWhat it costWhat the build does
Manual research
An hour per prospect, or a generic opener.
Research assembled before drafting starts.
Drafting from scratch
Every message competes with billable work.
Drafts written inside the encoded writing rules, ready to edit.
Follow-ups expire
Warm threads die during busy weeks.
Sequencing holds the cadence and resurfaces threads on time.
Tone drift
Rushed messages slide into generic phrasing.
The rules travel with every draft.
02
The approach

Encode the discipline. Orchestrate around the human.

The writing rules came first, as a spec: one named reader per message, the stake before the frame, the frame before the mechanism, and a test every sentence must pass (would it work unchanged on a website? Then it is too generic for a one-to-one message).

The orchestrator applies the rules. A person approves every send.

The approval gate is the centre of the design. Whether this prospect should get this message this week is a human call. The system does everything around it, so the founder's job shrinks to one decision per message.

Six agent roles handle research, fit scoring, strategy and drafting, with model choice per job.

03
How it works

Five stages. One human gate.

Each stage compresses the work around the person's judgment instead of replacing it. Click a stage.

04
What it shows

What this work shows.

45 of 116 drafts stopped at the gate. That is the gate doing its job.

Encode before you orchestrate

Written standards fail under calendar pressure. Encoded, they become the floor.

Automate around judgment

Relationship calls stay human by design.

Cadence is where outreach dies

Follow-up timing was the most valuable part of the build.

We run what we sell

Proven on ResonAi's own pipeline first.

05
See it working

Try something like it.

PythonFastAPIClaude (model per job)LinkedIn and emailCSV and JSON records

Which of your standards survive contact with a busy week?

The same pattern fits proposal follow-up, client communications and recruiting. Start with the discipline your team wrote down but can't run consistently.

📅 Book a call