The plan, soup to nuts
Salesforce for everything. Copilot on every desk. LinkedIn accounts working every week. Warmed mailboxes. A trade show calendar. Reps who know the market. Training handbooks already written and years of machine knowledge sitting in folders and mailboxes. The one thing missing is the machine that joins them up, so a lead moves from first message to closed deal without a person carrying it between systems by hand. This page shows what that machine looks like, workflow by workflow, and where it already runs.
So close, so far
Scroll and watch what the build does. Each line is a real connection: an API user on Salesforce, the LinkedIn accounts and mailboxes we already run, the show lists, and a workspace that remembers everything it has seen.
What Tim has already built
Most companies that ask for AI have to spend the first six months writing down what they know. IR has already done it. The training handbooks exist. The machine knowledge is in folders and mailboxes going back years. Hundreds of thousands of conversations with engineers, buyers and suppliers sit in Salesforce and the inboxes. The reps carry the rest in their heads and their trip notes. The build reads all of it once, keeps it, and puts it to work.
Zach or Graham asks "what tube fits this system and what did we quote the last three times?" and gets the answer from the handbooks and the quote history, with the source, instead of asking a colleague between calls.
An engineer emails about a system down. The draft comes back with the part, the alternative that worked at another site, and the last price, from the mailbox history and the handbooks. The rep checks it and sends.
Laura asks for a piece on a common failure and its fix. It is drafted from the handbooks and the real cases, in IR's voice, ready for the LinkedIn page, the blog and the pre-show sequence.
A request comes in for a part IR has quoted forty times. The draft quote carries the right spec and the right number, and the rep approves it.
This is the same job we did for BOY Machines on Dynamics 365: five thousand trip notes and close to a hundred thousand CRM activities read once and turned into something to say about more than five thousand accounts, before a rep ever asked a question. IR's raw material is deeper.
The workflows
Each one plays when you reach it. The grey lane is how it runs now, with the places it waits or drops. Then press the teal button for the same job with the machine in the middle, or flip the tabs at any time. Every teal lane is something we run already, either for our own pipeline or for a client.
What you already have from this
Both platforms, all five seat lanes (Courtney's account, Tim's and Mike's accounts running for Tanya and Arynn on the old platform, then Arynn's and Tanya's own accounts since June), January to the end of August 2026. Our own readbacks. They stop at the point a person replies, because that is where the work leaves our system and lands in a notification. Tim's read on the 28th: real business has come from it, the dollars are hard to tie back, and not every interested reply gets worked.
Historical results are what these accounts have produced, not a forecast. The point is the three dashed boxes: the build makes them solid.
Where this already runs
Salesforce is the one platform we have not yet wired, and it is the one with the cleanest API of the three. Microsoft Dynamics 365 is a far harder platform to work with, and it runs.


The return
Two places the money sits. Change the inputs to your own figures; nothing here is a promise, it is arithmetic on what you tell it. The pre-filled numbers are placeholders to show the shape.
How it is worked out: attendees x accept rate x meet rate = meetings; meetings x quote rate x average quote = pipeline per show; x close rate x shows = closed. Backlog: replies x reached-and-quoted x average quote x close rate. The accept rate on your own accounts since January has run at 29 percent across both platforms (2,731 of 9,554), so the 20 percent placeholder is conservative; the rest are yours to set, and the machine reports the real ones every week.
What this leaves out on purpose: the cost of the show itself does not change, the reps' time on the floor is the same time spent better, and the same machine keeps working the leads long after the show.
Salesforce stays
Salesforce keeps every record. The AI inside it can give you a list. The brain can read the list, write the emails, create the leads, log the calls, route by territory, chase the silence, and remember every conversation it has had with you. It works through a standard Salesforce API user, the same route the platform's own partners use, scoped to leads, contacts, accounts, opportunities, quotes and activities.
We have wired this into HubSpot and into Microsoft Dynamics 365, which is a far harder platform than Salesforce to work with. The part of the Salesforce spend that goes on "someone to make it do what we want" is the part this replaces. Copilot stays as the front door if you want it to.
Innovative TeleRadiology

ITR has no outbound today, no CRM, one part-time marketer, and a service that imaging centers and women's health practices need at the moment their turnaround, cost, capacity or quality slips. The build points the IR machine at a second list in the same weeks, with the routing rule ITR chose: an interested reply lands with Heather or Jen as an email with the conversation and a number to call.
A message goes out from Jen's account to the practice manager at a women's imaging center.
"Our turnaround has slipped to a week, what's yours?"
Reply read, the center created in the CRM, the conversation attached.
Heather gets the email: who they are, what they said, the number.
She calls. The call and the outcome go on the record.
Quiet for a week? The retouch goes out on its own.
The call
Eleven in the morning if you are still in San Francisco, seven in the evening for Dan. If Dreamforce runs late, Monday the 21st or Tuesday the 22nd at the same time. Josh, Tim, Kyle, Mike and Laura.