Service
Custom AI and integrations for the systems you already run
We add AI features to the software you already use and build the connections between systems that no off-the-shelf connector covers. The work runs on your own AI provider account, so usage, data and billing stay under your control.
- Built for
- Businesses with their own website, app, portal or ERP
- Models
- OpenAI, Anthropic Claude or Google Gemini APIs
- Staff role
- Test outputs, approve what goes live
- Price
- Request a quote
For systems that need more than a ready-made connector
Some tools connect to each other in a few clicks. Others do not, or the connection leaves out the step you need. This service covers the gap: a feature added inside your own system, or a link between two systems built on their APIs and webhooks.
- Businesses with a system no off-the-shelf connector covers
- Teams with internal documents or data they want an assistant to use
- Companies with an IT lead who wants a build partner
What we build
Documents
Assistants grounded in your documents
An assistant that answers from your manuals, price lists or policies and shows which document an answer came from.
Workflows
AI steps inside existing workflows
Summaries on CRM records, a classifier that routes incoming email, or a product description drafted inside your store admin for approval.
APIs
API connections between systems
Records move from one system to another through official APIs and webhooks, with logging and error handling.
Interface
Internal tools with a simple interface
A small form or page for staff to use the feature without touching the technical parts.
From a record in one system to a reviewed result in another
- New record in your system (Webhook or API), Automated
- Data checked and prepared (Rules), Automated
- AI step reads or drafts (Model API), Automated
- Staff review the output (Approval), Your staff
- Result written back (Your system), Automated
- Usage and errors logged (Log), Automated
Systems we connect
The list depends on what your systems expose. Typical connections are CRMs, booking and point-of-sale tools, spreadsheets and databases, accounting exports and messaging platforms. Each platform has its own page on the integrations page, with the constraints stated plainly.
Security and data handling
Before any build, the scope names where data is processed, who has access, and what the AI model provider receives. API terms differ from consumer app terms, so each provider's official data-use policy is read and linked in the scope.
Personal data is handled under the Data Privacy Act of 2012 (RA 10173): only the fields the feature needs are sent, access is limited to named users, and nothing is kept for any purpose other than your project.
When an off-the-shelf tool is the better choice
A custom build is not always the right answer. We say so when:
- A ready-made connector already does the job and is maintained by its vendor
- The process changes often enough that a fixed build would be out of date quickly
- The volume is small enough that a manual step costs less than a build
The AI consulting service answers this question before any build is quoted.
What we need from you and what we handle
We need from you
- A technical contact who can answer questions about the systems involved
- API access or data exports for each system
- Sample data, with personal details removed where possible
- Acceptance criteria: what a correct result looks like
- An API account with the AI provider, in your name
We handle
- Technical review of your systems and their limits
- A prototype on sample data before the full build
- The build, usage limits, logging and error handling
- Testing against real examples and edge cases
- Handover with documentation written for a technical reader
What a first project includes and what it leaves out
Included
- One AI feature or one system-to-system integration
- Tests against real examples
- Usage controls and error alerts
- Documentation your own developer can follow
Left out
- AI provider usage fees, billed to your provider account
- Rebuilding the host application
- Mobile app development
Limits we state plainly
- Model output varies. Features are designed so that a wrong output is caught by a rule or a person, or is easy to correct.
- Providers change models and retire versions, so integrations need periodic maintenance.
- Rate limits and usage costs apply per provider account.
- Software with no API and no export cannot be connected reliably. This is checked in the technical review, before a build is quoted.
What it costs
Cost depends on the systems involved and the state of their APIs.
Custom AI and integrations
Request a quote
Quoted per project after a short review of your channels, volumes and systems.
Request a quoteWhat moves the price
- Number of systems connected
- Quality and limits of each API
- Data preparation needed
- Interface needs for staff
- Testing and acceptance work
- Ongoing support
Questions we get asked
Can AI be connected to software that has no public API?
Sometimes. Where a system offers exports, email notifications or file drops, those can be used. Where it offers none, a reliable connection may not be possible, and the technical review says so before you commit.
Which AI model do you build on, and can it be changed later?
The model is chosen on output quality, cost and data terms, and the build keeps it behind one setting so it can be switched. After a switch the outputs are tested again, because models behave differently.
Where is our data processed?
That depends on the AI provider and the hosting chosen for your project. The scope names both and links the provider's data-use policy, so you can read it before approving.
Do we own the code and workflows?
We build in accounts and repositories you own wherever the tool allows. What you own, and what stays with a vendor's platform, is written into the scope.
Can our own developer maintain it after handover?
That is the aim. The documentation is written for a technical reader, and the logging shows what each step did.
Next step
Describe the system and the feature you have in mind.
Tell us what the system is, what it exposes and what the AI step should do. We will say what a build would involve, or whether a ready-made tool is enough.