Product 01 · The business operating layer
Weft
More conversations.
Not more people.
A human decision outranks a machine observation. Always.
The category
Two kinds of software exist.
Neither one owns the middle.
A CRM is a system of record for deals. An engagement platform is a system of delivery for messages. Between them sits the part that actually decides whether a relationship-led business grows: who you are talking to, what they meant, what you already said, and what happens next.
Nobody owned that. So every team improvises it — in a spreadsheet, in a browser tab, in one person's memory. It works until the volume arrives, and then it quietly stops working without anyone deciding that it should.
Weft is that missing layer, built deliberately. The delivery engines stay exactly where they are — LinkedIn outreach on its platform, email on its own sending infrastructure. Weft sits above both and owns identity, classification, pipeline and truth. That is why it shipped in weeks instead of being a migration, and why replacing an engine underneath is a change of adapter rather than a rebuild.
Screen 01
One inbox. Every channel.
Sorted by what they meant.
This is the product. Everything else feeds it.
- What this is
- Replies arrive already sorted by what the prospect meant. Interested is a tab with a number on it, not a search you have to remember to run.
- Why it exists
- Intent is the only ordering that matches how a relationship sale actually works. Sorting by arrival time tells you what the machine did most recently. Sorting by intent tells you where the revenue is this morning.
- Business outcome
- 486 warm leads standing in one queue instead of scattered across four platforms and a spreadsheet. The day starts with the ones that matter.
- What this is
- 4,180 live conversations, across every sender identity and both channels, held in one list.
- Why it exists
- The real ceiling in relationship selling was never lead generation. It was how many open conversations one person can carry before dropping one. That ceiling used to be somewhere around forty, and it was enforced by human memory.
- Business outcome
- Capacity stopped being a property of the person and became a property of the system. The next thousand conversations do not require the next hire.
- What this is
- Campaign attribution on the open conversation — which sequence brought them in, and where they are inside it.
- Why it exists
- Before replying to a warm lead, everyone asks the same question: what have we already done with this person? Journey answers it with the full date-ordered history — every message, every reply, every campaign, every status change, in one column.
- Business outcome
- Removes the one mistake a relationship sale cannot absorb: approaching the same studio twice, differently, from two different people.
- What this is
- Reply in place, with attachments, and an AI draft button that drafts and never sends.
- Why it exists
- The blank page is the single biggest reason a warm reply sits unanswered for a day, and at this volume a day becomes a week. But nothing leaves without a person deciding it should. No autonomous agent speaks to a senior designer on behalf of the brand.
- Business outcome
- Average response time held at six hours while volume multiplied — without ever handing the relationship to a machine.
Screen 02
The whole operation,
on one screen.
Every figure traceable to its source and filtered by the same controls as everything else — so two screens can never disagree.
- What this is
- 18,400 approaches, and every conversion step underneath it — accepted, replied, interested — each carrying its own rate rather than a bare count.
- Why it exists
- A count tells you what happened. A rate tells you whether to change the message. Every figure here is pulled from the source platform and filtered by the same date, campaign and identity controls as the rest of the product.
- Business outcome
- Outbound became a measured channel rather than an act of faith — which is the precondition for it improving at all.
- What this is
- 486 interested leads — 41% of everyone who replied.
- Why it exists
- Intent is resolved through a strict chain: your decision beats a colleague's earlier decision, which beats the platform's machine tag, which beats a mere system observation like “a message was sent”. A human judgement is never overruled by a machine's note.
- Business outcome
- The warm pipeline is visible by default instead of being reconstructed by hand every Monday. In a market where one studio can be a multi-year account, that ordering is worth more than any other single decision in the product.
- What this is
- The working queues — 63 waiting on a reply, 340 follow-ups scheduled or overdue — and a six-hour average response time.
- Why it exists
- In a six-month buying cycle, “send me something in March” is the most common thing a prospect says and the most commonly forgotten. The queue re-checks itself: if someone replied and was marked closed, their follow-up disappears rather than nagging.
- Business outcome
- Long conversations survive their own gaps instead of dying in them. The four-month wait became an asset rather than a leak.
- What this is
- Revenue, on the same screen as the activity that produced it — £1.24M quoted, £612,000 closed.
- Why it exists
- Outbound is usually measured in messages because messages are easy to count. Putting enquiries, quotations and closed business beside the send volume is what turns a marketing activity into a commercial one.
- Business outcome
- Outbound acquired a revenue number the finance team can use — and a defensible answer to what the channel is worth.
Screen 03
A pipeline that refuses
to lie to you.
The most quietly opinionated screen in the product, and the one finance people react to.
| Studio | Market | Stage | Value | Owner |
|---|---|---|---|---|
| Colefax & Vane Studio | United Kingdom | Quotation | £186,000 | Ayaan |
| Moran Interiors | Ireland | Closed | €94,000 | Nidhi |
| Nordiska Rum | Sweden | Quotation | €142,000 | Rahul |
| Whitcombe Design | New Zealand | Enquiry | A$128,000 | Aditi |
- What this is
- £1,240,000 quoted, rendered as three separate currency totals rather than one.
- Why it exists
- Nine markets means nine currencies. The moment a product adds two things that don't belong together to produce one tidy number, it has started lying — helpfully, and irreversibly.
- Business outcome
- Finance can read this screen without unpicking a blended figure nobody can source.
- What this is
- 285 live deals, each in its own currency, with the owner beside it.
- Why it exists
- Stage, value, owner and market are stored server-side, per user, permanently — so a deal survives a new laptop, a cleared browser, and the person who opened it leaving.
- Business outcome
- A conversation can be picked up by anybody, because the record isn't in a head.
- What this is
- “No grand total — by design.” The product explaining a number it refuses to print.
- Why it exists
- £1,240,000 plus €486,000 is not 1,726,000. Somebody would eventually ask where the total was, so the answer lives on the screen instead of in my head.
- Business outcome
- The restraint is the feature. It is also the reason people trust the other numbers.
What changed
Capacity stopped being
a property of the person.
Same team. Same market. Two completely different shapes of business.
Before · capacity = memory
~40
live conversations per person
Everything past that ceiling required another salary — and every new hire arrived with their own private, untracked pile of half-finished conversations.
After · capacity = system
4,180
live conversations held
Every one recorded, classified by intent, and scheduled — on a surface any colleague can pick up mid-conversation without asking a question.
The next thousand conversations no longer cost the next headcount.
| Before | After | |
|---|---|---|
| Live conversations held | ~40 per person, from memory | 4,180 across the operation |
| Warm leads surfaced | Rebuilt by hand each week | 486, classified by intent |
| Places to check one prospect's history | 3 | 1 |
| Where the record lives | Individual memory and a spreadsheet | Server-side, shared, permanent |
| Pipeline visibility | None | £1.24M across 10 currencies |
| Follow-up mechanism | None | 340 queued, self-checking |
| Average response time | Days, when it happened | 6 hours |
| Cost of the next increment of volume | One salary | Zero |
£1.24 million of pipeline, from a channel that previously had no number at all.
Not because more messages went out. Because the conversations they started stopped evaporating — and because the business could finally see what it already had.
Try it
The rule that decides whether you
can see your own pipeline.
Twelve leads, four rules, one ordering. Put a machine's observation above a human's decision and most of your warm pipeline vanishes — not deleted, just sorted behind something irrelevant.
Status precedence chain
Drag a rule to reorder it, or focus one and press ↑ / ↓.
Resolution order · highest wins
2 of 12 leads visible as Interested
Each card shows the value that won, and which rule produced it.
Architecture, for the people who ask
Weft keeps the engines and owns the workspace. LinkedIn outreach still runs on its existing platform; email still runs on its own sending infrastructure. Weft sits above both and owns the layer that was missing — identity, classification, pipeline and truth. That is why it shipped in weeks rather than being a multi-year migration, and why replacing an engine underneath is a change of adapter rather than a rebuild.
Data isolation is enforced at the source rather than in the browser, which is the only place it can be enforced honestly when several brands share one workspace. Every figure on every screen resolves through the same filter set, so no two views can disagree about the same week.
Much of the outreach platform's API is undocumented, so the integration was reverse-engineered and then confirmed with the vendor's own engineers — including two endpoints that appear in no published documentation and now carry the product's most important numbers. The negative findings were written down as carefully as the positive ones.
The whole front end has zero runtime dependencies. No framework, no build chain to rot. It will still run in five years if nobody touches it — which, for a system that has to outlive the attention of whoever built it, is a feature rather than an aesthetic.
Next product
Shuttle
Scale the accounts. Not the headcount.