01
Enquiry and triage agents
Read incoming requests, gather the relevant context, categorise and prepare a response or hand-off for your team.
AI agent development · UK
AI agents.Connected systems that take the next step.
An AI agent reads the context, decides between permitted next steps and acts through your existing tools: preparing a reply, updating a record, raising a task. I build agents with clear limits, approval points and an audit trail, so you stay in control of what they do.
Free first call · Fixed price after scoping · Sometimes the answer is a simpler tool
New enquiry: “Could you quote for 40 units, delivered by March?”
- Read and classify the request
- Check account history in the CRM
- Check availability and lead times
- Draft a reply and create a task
Reply drafted. CRM task created.
Waiting for approval before anything is sent.
Limited to permitted actionsApproval before consequential stepsEvery action logged
Is it right for you?
When an agent
is the right tool.
Agents suit work that needs judgement across several steps and systems. When fixed rules would do the job, simpler automation is cheaper and more predictable, and I’ll recommend it.
A good fit when
- The process involves reading, checking and deciding, not only fixed rules
- The work spans several systems that people currently switch between
- You can say what’s allowed and who approves what
- There’s enough volume that consistent, well-documented handling matters
Probably not yet when
- A fixed rule-based workflow would do the job. I’d build that instead
- Nobody can define what a good outcome looks like
- The action can’t be reversed and there’s no room for approval
What I build
Agents that work
inside your operation.
Each agent has a defined job, a defined set of tools and a defined point where a person steps in.
02
Operations agents
Chase missing information, reconcile records between systems and keep work moving, within the limits you set.
03
Research and preparation agents
Gather information from your systems and approved sources and prepare a briefing, comparison or first draft for a person to finish.
04
Agents inside your product
Let customers ask for something in their own words and have the product carry out the permitted steps on their behalf.
Agent or automation?
Choose the simplest thing
that does the job.
Agents are powerful but not always the right answer. The distinction helps decide what to build.
Workflow automation
Follows fixed rules: when this happens, do that. Predictable and inexpensive to run. Best when the steps never need interpretation.
AI agent
Interprets language and context, then chooses between permitted actions. Best when inputs vary and the next step depends on what they say.
Often, both
Many good systems use automation for the fixed steps and an agent only where judgement is needed. I design for that balance.
Keeping agents in check
Freedom to help.
Limits you set.
An agent that can take actions needs clearer boundaries than one that only answers questions.
Permissions and limits
Each agent gets only the tool access its job needs. Actions outside that list aren’t possible, not merely discouraged.
Approval points
Actions with consequences, like sending, paying or changing records, wait for sign-off until you decide otherwise.
Audit trail
Every step is logged: what the agent saw, what it decided and what it did. You can review and explain any outcome.
Shadow running
New agents run alongside your existing process first, so you can compare their decisions with your team’s before relying on them.
Ownership and running costs
Yours to run.
Costs you can see.
Deliverables
What you receive
You own the custom work I build for you.
- A working agent, connected to your agreed systems
- A documented list of permissions and approval points
- Logs and a way to review what the agent did
- Documentation, test scenarios and results
- A handover session with your team
Costs
What it costs to run
Running cost depends on how often the agent runs and how much it reads. I estimate it during scoping.
- Hosting and infrastructure
- AI model usage, which scales with how much the system is used
- Any third-party licences or APIs
- Optional ongoing engineering and support
How a project runs
Map. Shadow. Own.
A focused first version. Tested in the real world. A clear path to what comes next.
01
Map.
I map the process, the systems, the decisions and the approval points, then agree a defined first version and fixed price.
- Process & decisions
- Permissions & approvals
- Costed first version
02
Shadow.
The agent runs alongside your team on real work. Every proposed action is reviewed until it’s consistently right.
- Build & integrate
- Shadow running
- Agreed acceptance criteria
03
Own.
I move the agreed actions into everyday use, document the limits and hand over. Ongoing support is available if you want it.
- Go-live
- Documentation & training
- Support options agreed
What is an AI agent?
Software that uses an AI model to interpret a situation and choose between a set of permitted actions, such as looking something up, drafting a message or updating a record, then carries them out through your existing tools.
What's the difference between an AI agent and automation?
Automation follows fixed rules. An agent interprets language and context to decide what to do next, within limits you set. Many useful systems combine both, using an agent only where judgement is needed.
Can the agent act without approval?
Only where you agree it can. Most agents start with approval on every external action. After a period of shadow running, some low-risk actions may be allowed to run automatically; that is your decision.
Which tools can it connect to?
Most business software with an API: CRMs, email, calendars, accounting and practice-management systems, databases and file stores. I confirm your specific systems during scoping.
What happens when the agent is unsure?
It hands over to a person with the context it has gathered, rather than guessing. Where that threshold sits is agreed and tested during the build.
What does it cost?
After scoping, we agree a fixed price for the defined build. Running costs are estimated separately. Talk to me about your project.
Related
Other ways
I can help.
A useful first conversation
Is there work that needs judgement, not just rules?
Describe the process, the systems it touches and where people step in today. I’ll talk through whether an agent fits and where its limits should be.
Discuss your AI project Read the free RAG guide Or email hello@niroxai.comIndependent AI engineering. Based in Hull & East Yorkshire, working across Yorkshire and the UK.