Ryan Rose / Software engineering

AI agent development.

AI agents connected to your business workflows.

I build AI agents that work with your knowledge, records and tools. Start with one defined task, then connect the agent to the information and actions needed to complete it with appropriate human review.

Book a 30-minute discovery call ↗

Discuss your requirements and a practical next step.

Give the agent a concrete job

An agent is most useful when the task and output are clear. Examples include gathering missing enquiry details, drafting a response, summarising a call or preparing a record for a team member to review.

We define the inputs, permitted actions and situations that require a person. A useful first version should make the process easier to assess, rather than hide it behind an open-ended chat interface.

Knowledge, tools and human approval

I connect agent workflows to source documents, CRM records, databases and APIs. Access should follow the task: an agent preparing a draft does not automatically need permission to send it or change a customer record.

Evaluation uses representative examples, including incomplete information and ambiguous requests. Review points, source references and failure handling are designed into the workflow.

Explore a worked example

The enquiry-handling demonstration shows an incoming enquiry, missing-detail collection, a drafted response and a task awaiting human approval. It uses fictional data and does not send a real message.

My engineering work also includes AI-enabled call logging and scoring, knowledge-connected systems and Underpin Optimise, a project for selecting source context before model calls. A discovery call can help identify a practical first workflow for your business.

Try the illustrative enquiry-handling demo ↗

Before we start.

Will the agent send responses automatically?

Only if that behaviour is explicitly designed and approved. A first implementation can keep outgoing messages and record changes behind human review.

Can it work with our existing tools?

That depends on their APIs, permissions and data quality. We assess those interfaces before deciding how the agent should retrieve information or prepare actions.

Talk through your project.

Bring the process, system or idea you want to improve. We’ll discuss requirements, constraints and whether a focused first build makes sense.