Custom software or off-the-shelf SaaS: how to choose
Custom software or an off-the-shelf SaaS product? Cost, timescale, fit, data and lock-in compared, plus the hybrid approaches that work well in practice.
5 min read
AI integration · Ajaccio, Corsica
AI integration into your business software
AI integration brings artificial intelligence to where your team already works: your CRM, ERP, website or business application. From Ajaccio, Corsica, Step By Step connects large language models to your software through APIs or an MCP server, with costs and data kept under control.
- LLM
- API
- MCP
- RAG
Features
AI integration: the features we add to software
The most useful AI features in business software are simple and targeted: summarise, draft, classify, extract, search, translate. They slot into existing screens without changing how your team works.
-
Summarise and rephrase
One button that summarises a customer file, an email thread or a report, right on the screen where people work.
-
Writing assistance
A first draft of an email, product description or meeting note, based on the data in the open record.
-
Classify and qualify
Automatically categorise the requests, tickets or documents that land in your application.
-
Extract data
Fill in a form from a PDF, an email or a photo, with a review step before anything is saved.
-
Search by meaning
Search that understands a question asked in plain language, not just exact keywords.
-
Translate
Content and conversations translated inside the interface, with an optional review before publishing.
Architecture
API, MCP or RAG: how do you connect a language model to your tools?
There are four main ways to connect a large language model to your tools. The right one depends on what the AI has to do: a one-off feature, standardised access, answers grounded in your content or a complete task.
| Approach | Principle | Typical use |
|---|---|---|
| Model provider API | Your application calls a language model for a specific feature | Summaries, drafting and classification built into a screen |
| MCP server | Your software exposes its data and actions to an AI assistant or agent in a standard way | Making your tool usable by an agent without agent-specific development |
| Retrieval-augmented generation (RAG) | Your content is indexed and fed to the model with every question | Document assistants, decision support grounded in your sources |
| Agent connected to your APIs | An agent chains calls to your software to complete a task | Preparing a quote, assembling a case file, updating several tools |
Good practice
An integration that lasts, with no lock-in
A successful integration survives the fast-moving AI market. We design it so you can change model, control costs, protect data and keep checking quality over time.
-
Swappable model
An abstraction layer lets you change provider or model without rewriting the application.
-
Controlled costs
Caching, the right model for each task and limits on what is sent keep the bill predictable.
-
Protected data
Minimal data sent, sensitive fields anonymised and providers that do not train on your data.
-
Measured quality
Test sets re-run with every change to check that answers stay correct and compliant.
Our software
AIKeep: already connected to your tools
Rather than integrating AI one application at a time, AIKeep — the software published by the studio — sits in front of them: it connects through APIs and MCP servers, drives a browser for services with no interface and becomes the single command centre of the business.
- Any REST API supported through configuration (OAuth 2.0, API key, token)
- Any MCP server connected the same way
- Installed on your Windows workstations, with no data migration
Method
The stages of an AI integration project
An AI integration project starts from a specific, frequently used feature whose output can be checked. Once in production, usage, costs and quality are tracked.
-
Choose the feature
Where does AI save time in your software? A specific, frequently used feature with an output you can check.
-
Design the integration
API, MCP or RAG, model choice, data sent, error handling and cost control.
-
Build and test
Integration into the interface, tests on real anonymised data, quality measured.
-
Roll out and monitor
Gradual release to production, usage and costs monitored, continuous improvement.
First call on us
Want to add AI to your software?
Tell us which software you use and what AI should do in it: we will reply within one working day with the options available.
Artificial intelligence
Our other artificial intelligence services
- AI agents AI agents that carry out multi-step tasks in your tools, under human control.
- AI chatbots and assistants An assistant that answers from your own content, on your website or internally.
- AI automation Emails, documents, data entry: repetitive tasks handed over to AI.
- Custom software ERP, CRM, HR, management tools: business software built around your processes.
- Web applications Portals, dashboards, SaaS: applications that run in the browser.
Blog
Further reading: our articles on the topic
-
-
AI, the GDPR and the EU AI Act: what a small business needs to know in 2026
Using AI in a small business in 2026: GDPR duties, the EU AI Act timeline, transparency, AI literacy for staff and choosing the right tools. An overview.
6 min read
-
The Model Context Protocol (MCP): connecting AI to your business software
The Model Context Protocol (MCP) standardises how AI models connect to your software and data. How it works, what it enables, security and how to adopt it.
5 min read
FAQ
Frequently asked questions about AI integration
What does AI integration into software mean?
AI integration means adding features powered by a large language model to existing or new software: summaries, writing assistance, classification, data extraction, search by meaning. Users get the benefit without switching tools.
Can AI be added to software we did not build ourselves?
Often, yes. If the software has an API, we can build an AI feature around it, or an MCP server that makes it usable by an assistant. If it does not, there are other routes: file imports, a browser extension or browser automation.
What is the MCP protocol?
The Model Context Protocol (MCP) is an open standard that lets an application expose its data and actions to AI assistants and agents in a uniform way. An MCP server built for your software makes it usable by several compatible AI tools.
Which language model should we choose?
It depends on the task, the volume, the cost and the contractual guarantees. We design the integration so the model can be swapped: you can start with one, try another and switch over without rewriting the application.
How much does day-to-day use of a language model cost?
Providers charge by the volume of text processed. The real cost depends on the number of users, the length of documents and the model chosen. We estimate it at design stage and put mechanisms in place to keep it under control: caching, the right model for each task, usage limits.
Is AI integration compatible with the GDPR?
Yes, provided it is planned for: minimal data sent, a data processing agreement with the provider, user information, a defined retention period and, where needed, anonymisation of personal data before it is sent to the model.