Case study · 25 August to 24 September 2026
One person, a set of agents, and an ABM campaign that outran its own calendar
For thirty days the founder of Digiton ran account-based outbound across 24 countries with no sales team. The agents did the sourcing, the counting, the follow-ups and the paperwork. Then the calendar became the bottleneck.
The setup
One founder, no SDR, no agency, and a few years since he had done this himself.
Digiton sells an owner-acquisition engine to short-term-rental managers and a zoning-intelligence platform to municipalities and developers. Both are sold to people who get pitched every day, in nine languages, across markets where the buying rules change from one street to the next.
The constraint was headcount. So the campaign was built the way we build for clients: a set of agents around one person, each agent owning one job, every action logged against the company it touched. The founder wrote the strategy and took the calls. Everything before the call was the agents' job.
The machine
Eight jobs, one person in the middle, two channels out.
Segmentation
Every target came from public registers and listing data an agent could read and re-read on a schedule. No bought lists. For each company the agent recorded the city, the size band, the segment, and the reason it was on the list at all. That reason became the first line of the first message: the email to a Manchester operator opened with Manchester's numbers.
Decision makers
112 companies had more than one person mapped: founder, head of sales, operations, a regional lead. Each got a different angle on the same company, and a reply from one person paused the sequence for the others.
Artifacts per company
A company that replied got something built for it before the first call: 18 proposal pages published as live web pages with selectable options and a live price, and 22 decks, each for a single company. A Bristol manager saw Bristol's owner-run homes counted. The artifacts were generated from the same data the segmentation used, so they took minutes.
Two channels
Email carried the volume. LinkedIn carried the deals.
Sequences
Up to four touches per company, each one different, each one referencing the last. Warm follow-ups, a message into a thread where someone had already replied, converted at 31%.
Thirty days
Eight waves out, eleven meetings held.
What came back
13% replied. Price in the first touch cut that to 3%.
Two first-touch variants are worth naming. The one with no price, one case study and one ask replied at 11%. The same audience with a price line in the first touch replied at 3%. And when a price is corrected the day before a call, the call dies. Price belongs on the call.
MQL: a company that replied with anything other than a no, an opt-out or an auto-reply. SQL: a company that agreed to a call, held one, or asked for a price.
What broke
The machine booked faster than one calendar could sit.
It kept going while the founder was in the calls it had already booked. Replies waited a median of 25 hours for an answer, and 26 of them waited more than three days. Invites went out the moment a slot was approved, so he prepared for calls that moved at the last minute, calls that never existed because they sat on someone's confirmation, and a few that overlapped after a client asked for an hour earlier or later.
The leads were there. The bottleneck was the human in the loop.
What changed
Three rules, now enforced by the agents instead of remembered by a person.
Same-day replies
Every inbound reply is answered inside the working day. The reply desk flags anything older than that in the morning view.
Confirmed times only
No invite goes out until the other side has agreed to a time. An invite nobody accepted is flagged the day before, never treated as a meeting.
One pipeline, refreshed every three hours
Every company, every contact, every touch with the message and the figure it carried, every meeting with its acceptance status. No model tokens spent on the refresh.
The next 30 days run the same machine on half the list, with more humans on the calls and the founder's time spent only on the conversations that exist.
For a client
This is the engine we install.
Sourcing from public data, multi-contact account maps, two-channel sequences, artifacts generated per company, and a pipeline view that tells you the truth every morning. Your team takes the meetings. We do everything before that, for property managers, developers and B2B service firms.
How many people ran this campaign?
One. The founder wrote the strategy and took the calls. Agents handled sourcing, segmentation, sequencing, artifact generation, reply detection, logging and follow-up scheduling.
Where did the target list come from?
Public registers and public listing data, read by agents on a schedule. Nothing was bought. Each company carried a recorded reason for being on the list, and that reason opened the first message.
Which channel worked best?
Email carried the volume and LinkedIn carried the deals. LinkedIn was 7% of touches and produced five of the seven short-term-rental meetings that led somewhere, plus the one contract signed in the window.
What would you do differently?
Answer every reply the same day, keep price off the first touch, and never let an unaccepted invite sit in the calendar as if it were a meeting. All three are now enforced by the system rather than by memory.
Can Digiton run this for my company?
Yes. We install the same engine for property managers, developers and B2B service firms. Your team takes the meetings. Everything before the meeting is ours. Start at digiton.ai/contact.
Next step
Want this running for your company?
Tell us who you sell to and in which cities. We come back with the public sources we can read for that market, the account map we would build, and the first message we would send.
Talk to DigitonDigiton Dynamics. AI agents and AI search optimization, built in Lisbon, deployed across 8 countries. Campaign window 25 August to 24 September 2026. Published 2026-09-24.