An AI company building agents for manufacturing sales
Delynt AI builds AI agents and AI SaaS for manufacturing sales and marketing. It starts with the mailbox, because that is where quote requests actually arrive.
What Delynt AI is
An AI company. What that means concretely: software agents that connect to systems a shop already runs and do work that currently falls to whoever has time for it — delivered as AI SaaS rather than a consulting engagement or a model you have to operate yourself.
One word needs pinning down before anything else, because it means something else in this industry. In manufacturing an agent is a person: a commission rep who carries your line into a territory. That is not this. Nobody at Delynt AI contacts your buyers, negotiates on your behalf, or takes a percentage of what you sell. The agents Delynt AI builds are software, running against a mailbox you connected and can disconnect.
The starting point is the sales mailbox, and that is a deliberate choice rather than the easiest thing to build. It is where quote requests actually land, it is the same in every shop, and it is upstream of everything else — a system that never sees the RFQ cannot help with any step that follows it.
Why this problem first
One quote request, sitting unopened since Tuesday. That is the observation, and it is not a market thesis — it is one specific thing that keeps happening in shops that are otherwise run well.
I came to this with a master's in data science and the specific lesson that comes with it: a technically correct model is worthless if it arrives after the moment it was supposed to inform. A message scored perfectly on Thursday, about mail that landed on Tuesday, is a research artifact rather than a working tool.
The version of that problem I kept running into was not about data at all. Small manufacturers are not short on information about their quote requests — the quote request is sitting right there, insales@, in full. What is missing is anything that reads it before a person gets to it. So a shop that would have won the job never bid on it, and never knew that was the thing that happened. The failure leaves no trace: no lost-bid report, no entry anywhere, just a buyer who went quiet.
That is a narrow problem and a boring one, which is precisely why it is a good one to start with. It does not need an agent that understands manufacturing. It needs one that reads the mailbox as mail arrives, drops what is obviously junk, and puts the rest in front of whoever quotes with a reason attached.
Who builds it
Owais Ahmad Khan. MSc Data Science, University of Verona. Around eight years in search and digital marketing before this, which is where the habit of checking a claim against a source comes from, and where I learned how much of what gets published about a product is written by someone who has never used it.
I write the code. Not "led the team that wrote the code" — there is no team. That is a real limitation, and it is also the reason a design partner gets the founder on the phone rather than a support queue.
Proof, not summary
How it gets built, in public
Every technical claim on this site describes something running in production. The build log is where those claims are set out in enough detail that you can disagree with them.
Agentic email triage: how the pipeline was built
The architecture, the trust-versus-spam scoring model, and the parts that broke. The longest and most checkable thing on this site.
Read itIs it safe to connect your inbox to an AI agent?
What security architecture actually protects a mailbox connected to an automated system, and what to ask about.
Read it
Where it goes next
Stated as roadmap, because that is what it is. More agents across manufacturing sales and marketing, working outward from intake into the steps that follow it. After that, other verticals with the same shape — regulated, document-heavy, and full of work that is currently done by whoever has a free afternoon.
None of that is available. It is planned. The distinction matters on a site whose entire argument is that what it says is checkable, and the honest version of a roadmap is one you can hold the company to later.
What is not true yet
This belongs on the about page as much as anywhere else, because the about page is where a founder is most tempted to imply momentum that doesn't exist.
- One agent is shipped — the one that reads a sales mailbox. Everything else described as next is planned, not available.
- No paying customers. Delynt AI is pre-launch and the site says so on every page that could be read as implying otherwise.
- No revenue, and therefore no pricing. Any number would be invented.
- No Outlook or Microsoft 365 support, and no date for it.
- No case studies, no testimonials, no named logos. There is nothing to show yet, and an illustrative scenario dressed up as a result is still a fabrication.
Talk to the person who wrote it
No sales script, because there is nobody to read one. Twenty minutes on your own sales mailbox.
Last reviewed 6 August 2026