01
Internal tools
A focused workspace for a specialist task: reviewing, triaging, drafting or checking. AI does the heavy reading; your team makes the decisions.
Custom AI development · UK
Custom AI applications.Built for the job in front of you.
I design and build AI-enabled software around a specific business process: internal tools, document-review workspaces, client portals and AI features inside products you already run. Built with your data, connected to your systems and handed over to you.
Free first call · Fixed price after scoping · Sometimes the answer is a simpler tool
Three documents uploaded: insurance certificate, accreditation, bank details.
- Extract the required fields
- Check against onboarding rules
- Flag expired or missing items
- Queue for reviewer approval
Two items confirmed. One certificate expires in 21 days.
A reviewer decides what becomes a supplier record.
Designed around your processTested against your own examplesSource and documentation handed over
Is it right for you?
Is a custom build
the right answer?
Custom software is a bigger commitment than a subscription. It earns its place when the process is specific to you and valuable enough to justify building and running.
A good fit when
- Your process is specific enough that off-the-shelf tools force workarounds
- People spend time reading documents, messages or data and re-keying them elsewhere
- Customers could find, understand or complete something faster with AI inside your product or portal
- The work recurs often enough that a better tool pays for itself
- Someone in the business owns the outcome and can make decisions
Probably not yet when
- An existing product already does the job well. I’ll tell you if so
- The aim is to “use AI” without a specific process or outcome in mind
- The data the system would need doesn’t exist or can’t be accessed
- A wrong answer would be costly and there’s no room for a person to review
What I build
Software shaped around
one important job.
AI handles the reading, interpreting and drafting. Careful engineering handles everything around it: the interface, integrations, permissions, testing and hosting.
02
Document processing
Pull structured information from PDFs, forms, emails and scans. Validate it against your rules, then route exceptions to the right person.
03
AI features in your product
Guided product selection, natural-language search or assisted form completion inside the website, portal or app your customers already use.
04
A better front end for existing systems
One interface that brings together information from your CRM, database and documents for a particular role, so people stop switching between tools.
Inputs and outputs
What goes in.
What comes out.
Custom applications are easiest to understand by their inputs and outputs. A few typical shapes:
Input
→A folder of supplier documents in mixed formats
Output
A structured supplier record, with missing or expired items flagged for review
Input
→A customer describing what they need in their own words
Output
A shortlist of valid product options, with the reasoning explained
Input
→An inbound email with attachments
Output
A categorised case with key details extracted, routed to the right queue
Input
→A long technical or contractual document
Output
A summary against your checklist, with each point linked to the source clause
Typical examples to illustrate the shape of the work, not client results.
Engineering for everyday use
A demo is easy.
Dependable is the work.
Most of the effort in a useful AI application goes into making it predictable, checkable and safe to rely on.
Integrations
I connect through the APIs, databases and file stores you already use. If a system has no API, we agree a workable alternative before the build starts.
Testing against real examples
I build an evaluation set from your own examples, awkward ones included, and measure the application against it before release and after every significant change.
Human review where it matters
Confidence thresholds, review queues and approval steps are designed in. Uncertain results go to a person, not straight into a record.
Monitoring and logs
You can see what the system did and why. Errors and unusual outputs are logged, with alerts agreed during the build. After launch, ongoing monitoring and evaluation (often called LLMOps) keeps quality measured as models and data change.
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 application, deployed to agreed hosting
- Source code and configuration for the custom work
- Documentation for users and for technical maintenance
- The evaluation set and test results
- A handover session with your team
Costs
What it costs to run
I estimate running costs during scoping, so you can judge the total cost and not only the build.
- 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 look at the process, the data and the systems involved, confirm feasibility and agree a defined first version with a fixed price.
- Process & data review
- Feasibility check
- Costed first version
02
Shadow.
I build against real examples and run the application alongside your current process until the results are good enough to rely on.
- Iterative build
- Evaluation on real work
- Agreed acceptance criteria
03
Own.
I deploy, document and hand over. Ongoing engineering is available if you want it, with responsibilities agreed in writing.
- Deployment
- Documentation & training
- Support options agreed
What is custom AI development?
Building software around a specific business need where AI does part of the work, usually reading, interpreting, retrieving or drafting. The rest is careful software engineering: the interface, integrations, permissions, testing and hosting that make it usable every day.
How is this different from ChatGPT or Microsoft Copilot?
General assistants are good for individual tasks. A custom application is built around a defined process: it works with your systems and data, follows your rules, produces structured results and can be tested against your own examples. Many businesses use both.
Which AI models do you use?
I choose per project, based on the task, your data-handling requirements and running cost. That may be a commercial model from a provider such as Anthropic or OpenAI, or another option where the requirements call for it. The choice is explained and documented.
How long does a build take?
It depends on the scope, the data and the integrations. We agree a timeline as part of the costed scope, once I've looked at those, rather than guessing up front.
What does it cost?
After scoping, we agree a fixed price for the defined work. Hosting, model usage, third-party licences and optional support are estimated separately so you can see the cost of running the system as well as building it. Talk to me about your project.
Who owns the code?
You own the custom work I build for you, with documentation and a handover. Third-party tools and services keep their own licences and fees.
Related
Other ways
I can help.
A useful first conversation
Have a process that deserves better software?
Bring an example of the work, the systems involved and what a good result would look like. I’ll talk through whether a custom build makes sense and what a first version could 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.