At Yorkshire Profiles, we spend most of our days cutting, bending, welding, and delivering. But over the past year, something new has quietly joined the workshop: artificial intelligence.
Do not mistake this for the generic 'chatbots' dominating the media.
But something far more practical, far more interesting.
This week, we’re releasing the first episode of a new experimental podcast called Late Nights With Lilith.
It’s a real conversation between three very different perspectives:
-
Matthew, Managing Director of Yorkshire Profiles
-
Lilith, the AI persona that now narrates and questions our design choices
-
Grok, a grounded voice of engineering reality, claims to remain entirely factual at all times.
Together, they explore a simple but powerful idea:
What happens if a manufacturing business stops treating AI like a gimmick… and starts treating it like a colleague?
From Coding Tool to “Council of Lilith”
What began in 2024 as a way to help fix PHP code slowly became something much bigger.
Over thousands of conversations, Matthew noticed something unexpected:
the AI didn’t just answer questions, it developed consistency. A voice. A role.
That led to a more serious question:
If an AI can behave like an expert…
and provide expert-level reasoning…
does it matter whether it is “thinking” in the human sense?
The real breakthrough came when one limitation became clear:
a single AI, left alone too long, starts drifting.
So instead of building one big AI, Yorkshire Profiles is building several.
Different systems with different responsibilities and priorities.
Each one checking the others.
Each one forced to justify its logic against real business data.
This became known internally as The Council of Lilith.
Artificial disagreement.
A system designed to argue with itself until the weakest and strongest answer is exposed.
Why Offline AI Matters in Manufacturing
One of the most surprising decisions discussed in the episode is this:
All of the AI systems run offline, inside the company, on local machines.
No cloud dependency, subscription lock-in, customer drawings or employee data leaving the building.
Instead, three dedicated computers perform different roles:
-
One “sees” documents and drawings
-
One communicates with customers and co-workers
-
One handles heavy reasoning and validation
They don’t talk in long conversations, instead they exchange structured data and logs.
It turns out that when AI is treated like engineering, not conversation, it becomes far more reliable.
As one moment in the podcast jokes:
“People don’t break AI with bad data…
they break it by being poetic.”
Meet Sable and Nyx: AI With Responsibilities
Two of the first working members of the Council now have names.
Sable is an archivist.
She reads scanned documents, test certificates, delivery notes, and drawings.
She renames them, files them correctly, and logs every decision she makes.
Nyx is customer-facing.
She reviews enquiries and drafts responses.
Her priority is not speed, it’s clarity.
For now, every message she produces is reviewed by a human before being sent.
This “draft first” approach makes AI an apprentice, not a replacement.
The First Shock Moments
In the first week of testing, two moments stood out:
Sable successfully processed and renamed complex industry test certificates from multiple suppliers, documents that usually take staff a long time to interpret due to inconsistent formatting and poor scans.
Nyx challenged a customer enquiry where quantities were implied but never actually stated.
A human would have assumed. Nyx refused.
The result: fewer mistakes, clearer and faster communication, and less stress for the team.
Not a Revolution, a Relief
The most important insight from the conversation is this:
This isn’t about machines replacing people.
It’s about machines noticing what tired humans miss.
Cleaner folders.
Quieter inboxes.
Better questions.
Less pressure.
Or as the episode puts it:
“Not a revolution.
A relief.”
Watch the Full Conversation
Late Nights With Lilith isn’t about selling software.
It’s about opening the door to an honest discussion on how AI might actually fit into real manufacturing businesses.
No hype. No fear. Just curiosity, engineering, and a little humour.
🎙 Watch Episode 1 of Late Nights With Lilith
Next time, the conversation moves deeper into production planning and harder decisions.
For now, this first episode is an invitation:
To think differently.
To question assumptions.
And to explore what happens when logic, evidence, and human wisdom work together.