The brief
A growing marketing agency was running its outbound by hand. The volume looked impressive on paper. The reality underneath was broken.
- 28,000 emails sent per month, manually.
- 10,000+ leads sourced by hand, copy-pasted between Google, spreadsheets, and the CRM.
- 54% of their 65,000 contacts were eventually deleted because of bad data entry.
- Most of the email traffic was landing in spam.
The team was spending 140 hours a month on the workflow and getting punished for it. Senior reps — who cost the agency between $120k and $180k a year — were doing data-entry work that should have been automated three years ago. The campaigns were running, but every campaign cost the agency more in senior-rep hours than the campaign returned in pipeline.
The owner’s diagnosis was wrong at first. He thought the problem was list quality. After OFO Collective ran the audit week, the actual problem became clear: the workflow itself was the bottleneck. Every step was manual, every step had a quality failure built in, and every step compounded the failures of the steps before it.
What we found in the audit
The audit week (week one of the trial) mapped the existing workflow end-to-end. Seven manual handoffs. Three different systems holding pieces of the same data. Zero validation before write. A senior rep doing roughly the work of two junior reps and one operations analyst combined.
The audit also surfaced two facts the team had not articulated:
- The agency was already paying for Apollo, HubSpot, and a Google Workspace stack that could do 80% of the work — they just were not wired together.
- The owner was spending 8 to 10 hours of his own time per month “checking” the manual workflow. That was the hidden tax nobody had measured.
The build target locked at the end of week one was clear: rebuild the outbound engine end-to-end on the existing stack, wire Apollo and HubSpot together, add the AI layer (OpenAI GPT for personalisation and reply triage), and ship before day 30.
What we built
We replaced the entire workflow with a single rep-facing interface and an automated pipeline running underneath.
Before: Manual search → copy to spreadsheet → manual upload to CRM → manually email each lead → manually update each lead.
After: Automated lead search and save to CRM → automatic sorting → automatic enrolment in email sequences → lead status updated as replies come in.
The interface for the team is one screen. A sales rep picks the target industry and screen location for the month. Hits Start Sequence. The system does the rest. Interested replies land in the rep’s inbox. Out-of-office and disinterested replies are handled automatically. The rep makes the call and closes.
The architecture, on a page:
- Rep opens a one-screen interface, picks an industry and a location
- Apollo runs a programmatic search and pushes raw leads into the pipeline
- Leads are enriched, deduplicated, and validated against the existing HubSpot database
- Clean leads write to HubSpot with the right lifecycle stage
- An AI layer trained on the agency’s historical reply data generates personalisation tokens for each lead
- HubSpot enrols the lead into the relevant 4-touch sequence
- Replies route based on intent — interested replies to the rep, the rest handled automatically
The stack
- Apollo for lead sourcing and enrichment — programmatic queries triggered by the rep-facing interface
- HubSpot as the CRM and sending platform — restructured lifecycle stages, sequence enrolment logic, and reply-routing
- OpenAI GPT for AI personalisation and reply triage — trained on the agency’s historical reply data
- A dedicated sending domain properly warmed before the first send
- n8n for the orchestration layer between Apollo, HubSpot, and the AI services — including deduplication and validation before write
Total monthly platform cost after the build: $2,400 AUD. The agency was paying $2,100 AUD before the build for the same set of tools used worse.
The results
- 94% time saved. 140 hours per month down to 5.
- At the same cost as the old manual workflow.
- Capacity for 100,000 emails per month — roughly 4x what they were doing before.
- 20,000+ leads sourced and uploaded automatically in the first month after ship.
- End-to-end fully automated. Search, sort, send, follow up, status.
- Clean CRM, reliable data. The 54% deletion rate stopped accumulating from the day the system shipped.
- Owner time recovered. The 8 to 10 hours of monthly “checking” went to zero.
What we learned shipping it
Two things broke in the first month and are worth naming, because they would break in any similar build:
The first reply-triage model over-routed. Our first version of the AI reply-triage was too generous — too many “polite no” replies were being routed to the rep’s inbox as if they were interested. The rep’s inbox got noisier for a week. We retrained on a fresh batch of labelled replies, tightened the threshold, and the noise dropped to expected levels.
Sequence delivery was punished by domain reputation. The agency’s domain reputation was already taking hits from the previous manual workflow. Even with clean data, the first two weeks of sequences saw lower-than-expected inboxing rates. We rotated to a dedicated sending domain, warmed it properly, and the inboxing rates climbed back over the following month. A clean stack does not undo a poisoned reputation overnight.
Both fixes were inside the 30-day trial. Neither cost the client a separate engagement.
What changed for the team
The reps stopped being data-entry clerks. They became closers.
The agency now runs more campaigns at the same time, with less effort, on better data. The volume scales with the system, not the headcount. Senior reps are spending their afternoons on the calls that close deals, not on the data entry that fills the CRM with bad leads.
The owner’s 8 to 10 hours of monthly “checking” went to zero. He uses that time on growth conversations instead — partner deals, agency-of-record renewals, the work that actually moves revenue.
Most importantly, the build did not require headcount change. The team that ran the manual workflow runs the automated one. The system replaced the work, not the people.
Why this build matters for Australian B2B agencies
Most Australian B2B agencies are doing some version of this manual workflow today. The pattern is familiar: a tool stack that mostly works, a CRM that mostly holds the right data, a team that mostly executes the campaigns. The “mostly” is where the hours go.
The build above is replicable. Apollo and HubSpot are platforms most agencies either already pay for or could adopt for under $3k AUD per month. The AI personalisation and reply-triage layer is the part that needs custom build — it is not a SaaS product you buy off the shelf, because the value comes from training it on your agency’s historical data.
OFO Collective ships this kind of build on the 30-day trial. The deeper write-up of the AI lead research workflow is at /blog/ai-lead-research-marketing-agency-case-study. To talk through a similar build for your stack, book a call.