For Ben Stammers, after Tuesday's call. What changed on LinkedIn today, then the three things you asked for in writing: the booking links, the cold email build, and the tools. Costs are in the email, not here.
00 What changed on LinkedIn today
Done just before midday, in place. Everyone already in a sequence simply gets the new text at their next step. Nothing was restarted and no invitation was withdrawn.
Seat
What is live now
You, Software Partnerships NA
Søren's shape. Two sentences, what Digica has delivered in their area, then 20 mins, with your cal.com/digica-ben link on its own line.
Mike, UK Defence, Semi and Medical
Same shape, his targeting untouched. The "false alarms" line is gone from every step, and so are "full IP transfer" and "300+ deployments".
Mike, Counter-UAS
Same shape, same targeting. No booking link yet because he has no page yet.
Sandra
Written the same way and proven on a dry run. It goes on the minute you have spoken to her, which is your own rule. It also takes the client name out of her step 2.
Søren, Rafał, Lukasz
Untouched. Søren's is the model. Rafał's is working, so it stays as it is.
The rules now in every sequence we touched
No opener about them. Nothing about their role, their company, or what "I see" them doing. Every other bot does that and people know it. Straight in with what Digica has delivered in their area.
The case study does the matching. One of the cleared lines from your section 05, picked for their niche, nothing invented. Project matching still does the work, it just does not announce itself.
No pain-point opener. If the pain is not theirs you have discounted yourself in line one.
No name at the top of the later steps. It is already a chat. No "opinion on something". No "I'll leave it there".
20 mins, and a booking link where the person has one.
"300+ AI projects completed", never "300+ deployments". Your rule, applied.
The AI tells you called out are checked on every message. No statement, colon, statement. No dash asides. Plain words a busy person would type.
Still open on LinkedIn
Your hand-written examples fold in on top of this by the same route when they arrive.
The connection request note on your seat. Raised on the call, not decided. Your own note from section 01 is ready to go on if you want that test, with Lukasz as the control.
Ramp to 20 accounts once the numbers are right. That is a seat and licence question, and the first week on the new copy tells us when.
The team. You said you would take Sandra, Mike and Søren through it yourself. The two-line version: the direct shape is what is getting Søren replies, so everyone moves to it, and each person's own examples shape their version. Happy to sit in on that call if it helps.
01 Booking links on cal.com
You made cal.com/digica-ben on the call and deleted the 15 minute option, so the link now lands straight on your 30 minute page. Four more things to do.
Swap Cal Video for Google Meet. Event type, Location, remove Cal Video, add Google Meet. Connect your Google account when it asks. Every booking then carries a Meet link in the invite.
Add a photo and a one-line bio on your profile. It is what a prospect sees before they pick a time.
One page per person. Mike, Sandra, Søren, Rafał and Lukasz each need their own account and their own 30 minute event, so the sequences can carry each person's link. The paid team plan puts them all under one team page and lets you see the whole team's bookings in one place. Names to use: cal.com/digica-mike and so on, same pattern as yours.
The clean URL. To send digica.com/ben instead of the cal.com address, add a redirect on the digica.com website: /ben forwards to https://cal.com/digica-ben, one line per person. That is a five minute job for whoever runs the site and it needs no self-hosting. Self-hosting cal.com is possible (it is open source) but it is a project, not a setting, so do the redirect first.
Why it mattersEvery click between "yes" and a booked slot loses people. One link, one page, one calendar picker. Our appointment setter books people onto our page while she has them on the phone, which is the other way to use it.
Cal.com also has a proper API, so once the pages exist we can read bookings out of it and push them into your CRM or the weekly report without anyone copying anything.
02 The cold email build
What you said you want: about 3,000 sends a day, 100,000 a month, in-house, with the infrastructure part done properly so it is not your time. This is the shape. Costs are in the email.
The shape
Piece
What it is
Domains
About 66 plain, generic .co names. Never a look-alike of Digica or anyone else. One send domain never touches digica.com.
Mailboxes
50 per domain, about 3,300 in total. Each mailbox sends one email a day, so to the receiving side each domain looks like a 50 person company doing ordinary email.
Isolation
One tenant per domain. If a domain ever gets flagged you lose that one domain and its 50 boxes, replace it, and nothing else is touched.
Warm-up
Three weeks before the first cold send, on every mailbox. Not negotiable, it is where most people burn their estate.
Daily checks
Placement tests every day on a sample of mailboxes, so a problem is found the day it starts, not when replies dry up.
Sending
The sequencer runs on our workspace, which has the capacity spare. Plain text, no links, no images, no signature blocks in the first email. Business addresses only.
Leads
A three step cadence at 100,000 sends a month is about 33,000 new people a month. Wider titles than the LinkedIn work, manager level up, then work up the chain from the replies.
Data
A bulk data source first, then a waterfall enrichment pass on the rows that need more (see the tools section). The more columns you have, the more personal the email can be without anyone writing it by hand.
The rules that keep it alive
The five things the estate I looked at yesterday got wrong, plus ours. Any one of them on its own is enough to end the channel.
DKIM, SPF and DMARC set up properly on every send domain before a single email goes.
Never the main domain, and never a subdomain of it. A subdomain still burns the parent. Send domains are separate, generic names with no brand in them, because the filters now link every domain that shares a name.
Plain text. No HTML, no signature block, no images, no links, no attachments in a cold email. All of them read as phishing now. The 2020 email you showed me would not deliver today for that reason alone.
Business addresses only. A batch to personal mailboxes lands in spam and takes the domain's reputation with it.
Never a whole database on a newsletter. And read the out-of-office count: 7,000 sends and 16 out-of-office replies means the mail is not landing, whatever the tool says.
Warm-up first, three weeks. Placement tests daily after.
Why vendors burn estates. A vendor is paid to use cheap infrastructure because the margin is theirs, and when it burns they do not fix it. Under about £5,000 a month for this volume that is what you are buying. It is why this is done at cost, on the vendors we use ourselves.
What it feeds
Interested replies for a person to call, not a link for them to clickOn our own LinkedIn engine, 27 interested replies in the last 7 days and 88 in 30. Our appointment setter started on the 15th and had booked nine by the time we spoke, by phoning the warm replies and booking them onto the page while she has them on the line. People who say they are interested rarely book themselves off a link, and they do book when someone calls. 100 interested replies a day is what the estate is sized for. If a setter reaches 30 of those and books half, that is 15 appointments a day.
What we do and what you keep
We set the workspace up, buy and configure the domains and mailboxes on our bulk rates, run the warm-up, and hand you the rules and the vendors. You own all of it. Nothing outside vendors we have used ourselves or been introduced to by people we trust.
Once it is running it is a daily check, not a daily job. If something goes wrong the checks say which domain, and the fix is a replacement, not a rebuild.
When it is live I will add you to a couple of operator groups where the changes get talked about before they are public. That is where most of what is above came from.
Start
Say go and the domains and mailboxes get bought this week, the warm-up runs while you are away, and the first cold sends land mid October. The list build and the copy happen inside the warm-up window, not after it.
03 The tools you asked about
The short version of each, what we use it for, and where to read more.
Jev, by TypeSafe
A model that only answers typed questions about a piece of text. Yes or no with a probability, one of a list, or a position on a scale. It does not write anything back. It answers in about a third of a second and costs a hundredth of a cent a call, so it can look at ten thousand things where a normal model would look at ten.
What it is good at: deciding. Is this reply interested or not. Is this person a buyer or a vendor. Which of these five companies fits the brief. Does this page say what we think it says.
How we use it: the code owns the loop, Jev makes every repetitive call, a person keeps every decision that ends in a send. My mailbox clean-up example from the call is exactly that: four options per email, the whole inbox in seconds.
For your prospecting tool: it is the piece that ranks the people once you have found them. Give it the profile text and the role brief, ask "does this person buy this", and you get a score per person instead of a pile.
Docs: docs.typesafe.ai. There are Python and JavaScript SDKs and the pricing is per token, not per seat.
treg
One API in front of a catalogue of data and enrichment tools, metered per call. People and company enrichment, email finding and verifying, LinkedIn post and profile data, web search, scraping, and the image and video models. You search the catalogue by what you want to do rather than by vendor, it tells you which providers cover it with a measured success rate and price, and you call the one you pick. Your keys stay in one place.
What it is good for: the waterfall. Try the cheap provider first, fall through to the next only for the rows that came back empty, and stop when the column is filled.
For your prospecting tool: the finding half. Search, enrich, verify, all through one key, with Jev doing the judging in between.
The recipe I mentioned. Give it a show and a date, and it searches LinkedIn posts about the show, asks Jev which posts are on topic, pulls the people who wrote or reacted to those posts, asks Jev which of them hold a buying role against your brief, and writes a page that ranks who to contact and why. One run this week read 235 posts and 636 people and kept 18, for less than a dollar. Nothing is contacted from it automatically and nobody is told how they were found.
CRM enrichment pass
The one I described for a manufacturing client. Take every company and contact record, fill the blanks (direct email, phone, headquarters number and address), re-verify every email, and find additional contacts where a record had none. Their coverage went from roughly 70 percent to roughly 96 percent on about 20,000 companies, and it cost about a thousand dollars in data. We then ran it on our own CRM and found about 80 people we had spoken to on LinkedIn but never had an email for. The same pass on your HubSpot is a few days of machine time and a data bill, not a project.
Decks
What you saw today is a template, not a designer. A deck is written as a short brief, rendered into the house template as a web page, and published to a private link in minutes. The same for one-page maps and leave-behinds. For your salespeople the useful version is a deck per prospect that pulls the right case study for that company automatically. Send me what the manual version looks like today and I will show you the shape on Wednesday.
Skills for your brain
You asked whether there are others worth sharing. Your brain has nine today: set-up, the content engine, CRM update, the list builder, meeting prep, reactivation, the reply desk, the voice gate and the weekly report. Ours has several hundred, and the thing you asked about, whether that slows it down, is answered by how they load: a small set loads every session, the rest sit in an index and fire on a plain phrase such as "prep me for my call with X". For the two things you described, the ones worth porting are:
A prospecting skill on top of your list builder. Find (treg), rank (Jev), enrich the keepers (the waterfall), and hand the salesperson a short page of who and why. The trade show finder is one input to it.
A deck skill. Brief in, house template out, published to a link, the right case study picked per prospect.
The CRM enrichment pass, run once on HubSpot and then monthly.
A post-call summary that writes the actions and the follow-up email from the recording.
Call prep that reads the diary and writes the sheet on the morning of the call.
Send me the two builds in a paragraph each and I will bring the shape of both on Wednesday.
04 Before Wednesday, 10:00
The two things you are building. The prospecting tool for the salespeople and the slide creation. A paragraph on each is enough, what goes in, what should come out, who uses it.
Your hand-written LinkedIn examples, when you have them. Your sequence is on Søren's shape now and your own examples refine it.
Mike's booking link, once his page exists, so it can go into his sequence the same way yours is in yours.
A line when you have spoken to Sandra, and her sequence goes on the same minute.
Go or not on the cold email build, so the buying starts this week and the warm-up runs while you are in Spain.