Service
Agents that do one named job and hand back at the edge of it.
Language models put to work on specific tasks: reading a document, answering from the company’s own material, qualifying an inbound lead and sending it to the right queue. Every agent has a scope, a fallback, and a point where it stops and a person takes over. This is the application layer; the infrastructure it runs on is local AI.
Focus
Agents, retrieval, extraction, qualification and routing, and internal assistance.
- Service
- AI engineering
- Deliverables
- Named and scoped below
- Bought as
- A defined build, or run under management
- Ownership
- Every account and platform stays in the client’s name
Scoped agents, not a chatbot
An agent gets one job, the tools it needs for that job, and a defined handoff. Anything outside the scope goes to a person instead of being improvised.
Retrieval and extraction on real documents
Answers come from the business’s own contracts, records, and transcripts. Fields the model cannot read confidently stay empty rather than filled with a guess.
Qualification and routing
Inbound gets read, scored against criteria the business set, and sent to the right queue with the reasoning attached, so a human can disagree with it.
Scope
What gets built.
Outcome
What changes.
Reading and re-keying stops being somebody’s afternoon.
Answers come from the company’s own material instead of the model’s memory of the internet.
Every agent has a boundary, and crossing it returns the work to a person.
Model behaviour gets measured against fixed cases before it reaches a customer.
More
The other focused engagements.
Start
Start with what’s broken.
Send the situation in a paragraph. Strygon comes back with a read on what’s likely wrong and what it would take to fix, before anyone talks about price.
Most builds start withleads dying in an inbox., three half-finished pipelines., follow-up nobody owns., numbers that never agree., four vendors blaming each other.
Or email hello@strygon.com
- What to send
- A paragraph. What broke, and where it shows up.
- What comes back
- A read on what is likely wrong and what fixing it takes.
- Price
- The last conversation, not the first