Internal business tools
Project trackers, job scheduling, operational dashboards and approval tools. Bring the information and actions a team needs into one usable interface.
Custom software & AI.
Built around your business.
When your business has outgrown a workaround, build a tool that fits. Dragon AI develops bespoke software, business platforms and AI assistants for teams in Wales and across the UK.
Tell us what you want to build- A focused prototype
- A useful first release
- Documentation and handover
Your best process shouldn’t depend on a workaround.
A spreadsheet becomes a booking system. A shared inbox becomes a project tracker. A customer portal exists as a series of attachments. These workarounds can keep a business moving, until the volume, complexity or number of people makes them difficult to manage.
We turn the job your software needs to do into a focused product. That starts with understanding the users and testing a prototype. AI is included when it helps with a specific task, with its permissions, limitations and review steps designed alongside the rest of the application.
Software for the way you need to work.
Customer and partner portals
Give people an appropriate view of their requests, documents or project progress, with access designed around their role.
Company knowledge assistants
Make approved company knowledge easier to use. Retrieval-augmented generation (RAG) brings relevant documents into the answer, with source citations, permission checks and a clear response when information is missing.
Websites and digital platforms
Build a customer-facing experience connected to the work behind it. Plan content, performance, accessibility and operational requirements together.
Your company’s knowledge. Ready when it’s needed.
The starting point
Policies live in shared folders, project decisions sit in documents and experienced colleagues keep answering the same questions. Finding a reliable answer means knowing where to look and who to interrupt.
A better way forward
A company knowledge assistant retrieves relevant passages from approved sources and uses them to prepare a cited answer. People can inspect the original documents, and the assistant flags questions it cannot support. Retrieval respects the user’s access to the source material.
- 1
Prepare the knowledge
Choose trusted documents, identify their owners and agree access, updates and removal rules.
- 2
Retrieve and explain
Connect search and a language model, keep supporting passages visible and define when to decline an answer.
- 3
Test the difficult questions
Evaluate accuracy, stale documents, missing answers and permission boundaries with a representative question set.
An illustrative engagement. We agree the actual scope, access and success measures with your business.
Something useful.
Yours to put to work.
The engagement has a clear scope and a tangible handover. We agree the deliverables before work starts.
Discuss the scopeA product scope and working prototype
The users, essential tasks and boundaries of the first release, tested before the full build.
The agreed application
A working product with the required views, permissions and connections.
AI evaluation where applicable
Representative examples, known limitations and agreed review or fallback behaviour for AI features.
Deployment and operating documentation
How the application is configured, where it runs, its dependencies and the routine tasks it needs.
A practical handover
Team guidance and agreed arrangements for source access, ownership, maintenance and future releases.
How we work with you.
Define the job
Agree who the product serves and the essential work the first version must support.
Prototype
Make the important screens and interactions tangible, then refine them with your team.
Develop
Build in manageable stages with regular reviews, testing and visible decisions.
Introduce and improve
Launch against agreed acceptance criteria, train the team and prioritise the next improvements.
A few things
you might be wondering.
Questions about your own setup?
Let’s talk it through
When is bespoke software worth considering?
It is worth investigating when a core process is poorly served by existing tools, when the same workarounds recur, or when a distinctive customer experience matters. We compare custom development with adapting or integrating an existing product before recommending a build.
Does every custom application need AI?
No. A reliable form, database or workflow can be the right solution. AI is useful for suitable tasks such as finding information or preparing a draft, but it also needs evaluation and review. We decide feature by feature.
Can you work with our existing systems?
Yes, subject to the access and integration options those systems provide. We investigate APIs, data formats, permissions and vendor constraints during discovery, so a dependency does not become a surprise late in the build.
Who owns the software and hosting accounts?
Ownership, licences, source-code access and hosting arrangements are agreed in the contract. We make these explicit before building, including how the work would be handed over to your team or another supplier.
How are price and timescale agreed?
We first define the users, essential features, integrations and acceptance criteria. A prototype can help resolve uncertainty before a larger commitment. The proposal separates the build from third-party licences, hosting, AI usage and ongoing support.
What happens after launch?
The handover covers routine operation and support ownership. We can agree maintenance and further development separately. Monitoring, backups, updates and incident responsibilities should be understood before the application becomes business-critical.
A little clarity before the next step.
Let’s build the tool you keep wishing existed.
Tell us what you’re working on. We’ll help you find a practical next step.
Tell us what you want to build