1. Home
  2. AI use cases
  3. Relationship intelligence

AI use case · built in Our own sales desk

A sales desk that reads, remembers and reminds

Relationship businesses run on who you know and when you last spoke. We built a desk that turns screenshots into records, drafts considered replies, and makes sure no conversation goes quiet by accident.

The problem

Most firms keep their relationships in three places at once: a CRM nobody updates, a spreadsheet someone does, and a partner's memory. Typing a LinkedIn profile into a CRM takes minutes per person, so it doesn't happen, and the follow-up promised in a meeting slips.

The work that wins business (a thoughtful comment, a timely email, a reason to call) is exactly the work that gets squeezed out by data entry.

How it works

Each step is done by whichever is right for it: a model where something has to be read or drafted, tested code where a rule applies, and a person where a decision is made.

  1. 01InputScreenshotsProfiles, search results, company pages, several at once
  2. 02AIClaude reads themName, role, company, country, sector and size, never a guessed link
  3. 03Rules in codeDe-duplicateSame URL or same name at same firm is flagged; only empty fields are filled
  4. 04A personYou correct and saveNothing is written until you confirm the list
  5. 05OutputPipeline and roundNext check-in dates, and who to engage today

AI reads, plans or draftsTested code applies the rulesA person reviews and decides

Live today

What it already does

Labelled honestly: AI where a model does the work, Rules where tested code does.

  • AI
    Screenshot to recordPeople and companies read from LinkedIn screenshots by Claude on the server, about a cent per image, with nothing stored by the model.
  • AI
    Reply draftingFive reply formats (appreciate, compliment, encourage, curious question, view and question) drafted from the post itself.
  • Rules
    Guardrails in codePolitical, religious and personal posts are skipped whatever the model says; a reply that pitches, links, tags or argues cannot be copied.
  • Rules
    One person, many conversationsThe same contact can be at a different stage for each business you run, each with its own value and next step.
  • Rules
    Need-to-know accessAn assistant can work the daily round with her own login and see names and timing only, never notes or addresses.

What AI adds next

Builds we can scope on the same foundations, each kept inside the same rules and the same review step.

Through the lens of your sector

The same build answers a different problem in each business. Pick yours.

The problem
A fund raise means hundreds of LPs, family offices and placement agents met at conferences and on LinkedIn, tracked in a spreadsheet that is out of date by the next roadshow.
Live today
Conference attendee lists and LinkedIn searches screenshotted in batches and turned into a clean investor list, each with a check-in date.
What AI adds next
Signal watch on LP allocations and team moves, and a briefing pack before each investor meeting.
Stays with a person
Who gets a call from the partner, and what is said.

Want your version of this?

Tell us how the work is done today. We will tell you which parts a model should do, which parts need rules around them, and what a pilot on your own data would look like.

Get a quote