Make artificial intelligence
something everyone can actually use.
Daedalus.AI was founded in May 2025. We start from how a business actually runs, distil auditable, reproducible workflows from it, and turn those into systems that hold up in production. Today that means three product lines: RPA, the Chengguan workflow engine, and an open-source CRM. Customers so far include Aomanlin and JOMI.Liberate.
Technology earns its value
by lowering the barrier.
AI should be something anyone can understand and use — amplifying what people can do, unlocking what an organisation is worth, and earning trust through delivery that is close to deterministic.
01Put AI within reach
AI should not be a complex system only a handful of technical teams can operate. It should be something everyone on the front line can understand, use and benefit from. The value isn't in showing how powerful a model is; it's in lowering the barrier far enough that colleagues in finance, supply chain, operations and HR get the efficiency directly.
02Amplify people, don't replace them
Machines are good at volume, repetition and anything with a pattern. People are good at judgement, creativity and situations nobody planned for. We embed AI into real workflows not to move people aside, but to free them from tedious execution and give that time back to the work that genuinely needs experience and judgement.
03Unlock what the organisation is worth
Repetitive, low-value work should go to AI — organising data, running processes, checking results. Those steps consume enormous amounts of people's time today and ought to be handled steadily by a system. People's value belongs in the decisions and creation that matter, not buried under the same task repeated daily.
04Deliver reliably
For a business, AI you can demo and AI you can depend on are entirely different things. We aim for delivery that is close to 100% deterministic — every run reproducible, auditable and traceable. That isn't a slogan; it's the engineering standard we hold every workflow and every engagement to.
People define the rules. Results have sources.
These three aren't values on a poster. They're practices you can check on any project.
The rules behind restocking, clearance and invoicing come from the people who do that work. We write them down in readable form and hand them to the system — we don't leave it to a model's own judgement.
The execution log records step by step and a failure can be replayed in place. Data writes only append — never alter, never delete — so you can always find who changed what, and when.
What isn't working yet is labelled as not working yet. Every status on this site — live, open source, private beta — matches the repository and the deployment.
A very small team.
Building and testing the same question at once: can AI really enter the daily work of a business, rather than stopping at the demo?
Yuan QuanBecky
Founder · CEO
Business and engineering
University of Southampton
MEng Computer Science with Artificial Intelligence
Huang ZijianSam
Co-founder · CTO
Engineering
University of Sheffield
MSc Computer Science
Liu YuxuanWill
Co-founder · Head of Growth
Marketing and sales
University of Manchester
MSc Fashion Management and Marketing
Want to talk through your process?
Walk us through it and we'll tell you which one or two product lines fit, and which steps should not be handed to AI at all。