EN
ESEspañolENEnglishRUРусский
Talk to Digitup

AI integration for your business: ChatGPT, Claude, Gemini or other models

Digitup Studio is a technical studio in Madrid that integrates AI models such as ChatGPT, Claude or Gemini into the software your company already uses (CRM, ERP, email, website, app or WhatsApp) or into your own product. We choose the model by task, quality, cost, speed and privacy, without tying ourselves to any provider.

Talk about my integration See AI case studies

In short

  • Integrating AI means connecting, through an API, a model from OpenAI (GPT), Anthropic (Claude), Google (Gemini), DeepSeek, Perplexity or another provider to your systems to classify, summarise, extract data, draft or translate.
  • OpenAI and Anthropic do not train on data from their business plans or their API by default; Google does not either for Google Cloud data or the paid Gemini API (checked on 23 October 2026).
  • Projects start at €1,000 and delivery starts at 2 weeks; model usage is paid separately, by consumption.

What it is

What does it mean to integrate AI into a business (and what does it not)?

Integrating AI means connecting a language model to your company's systems and data so that it performs a specific task inside a process that already exists.

Digitup Studio has been designing custom software since 2019 and taking it to production. For us, integrating AI means the model works where your information already lives: classifying an email, extracting data from a document or summarising a customer's history in the CRM, with no copy and paste. What gets integrated is the model's API, not the chat application.

It is not giving ChatGPT accounts to the team, nor training your own model, nor putting a generic chat on your website: we design flows, context, data, prompts, rules, integrations and control panels. If the AI has to chain steps and use tools on its own, we are talking about AI agents; the full picture is in AI for business.

Comparison

Business plan, API integration or an assistant with your documents?

A business plan gives AI to each person, an API integration puts it inside your processes, and an assistant with RAG answers from your documents.

Aspect Business plan for the team API integration in your systems Assistant with your documents (RAG)
What it is Chat accounts for the team: ChatGPT Business or Enterprise, Claude Team or Enterprise, business editions of Gemini. The model works inside your CRM, ERP, email, website or app, with human approval where needed. An assistant that answers from your documents and cites the source.
Cost Subscription to the provider, usually per user. Development plus usage-based consumption. Development, consumption and indexing.
Data control Whatever the plan's admin console allows. You decide which data leaves and to which provider. Permissions per source; you update documents without retraining.
When it fits Individual productivity: drafting, summarising, researching. Repetitive tasks with data that already exists. Queries over extensive or changing documentation.

Plans checked on the OpenAI and Anthropic websites on 23 October 2026. More in RAG for business.

Which model should you choose: ChatGPT, Claude, Gemini or another?

No single model is best at everything: you choose by task, testing with your real cases. We do not resell licences from any provider.

Quality on your task

We compare models using your real emails, documents and questions.

Cost per use

For classification a small model is usually enough; larger ones are for drafting or reasoning.

Speed

In a chat or a call the wait is noticeable; in an overnight process it is not.

Privacy and location

Who receives the data, where it is processed and how long it is kept.

Formats

Long PDFs, images or audio: not every model accepts the same inputs.

Provider dependency

Models get retired: we design so they can be swapped without rebuilding the product.

Multi-model

Can you use several models in the same product?

Yes. In products with heavy usage it is the sensible approach: each task uses the model with the best balance between quality and cost.

In Rescrito, the AI SaaS for academic writing that we developed, the multi-LLM approach makes it possible to choose the provider by task, latency or price, with OpenAI, Gemini and other models, and each tool has its own prompts and rules. In a small integration it works the same way: a cheap model classifies the emails and a more capable one drafts the replies. Because the business logic does not depend on one specific provider, models can be swapped or combined according to cost and quality without rebuilding the product.

Where

Where can AI be integrated?

In any system with an API or access to its data. If your ERP works, it stays: we build the missing part.

Notes nobody reads?

CRM

Customer summaries and lead classification.

Custom CRM

Invoices keyed in by hand?

ERP and administration

Data extracted from invoices, ready to approve.

Automations

Overloaded inbox?

Email

Classify messages and prepare drafts in Gmail or Outlook.

AI Transformation

Nobody can find the document?

Documents

Answers from your contracts and manuals.

RAG for business

Your website does not help people find things?

Website and app

Natural-language search and assistants.

AI chatbots

Customers on WhatsApp?

WhatsApp and Telegram

Replies and bookings, with handover to a person.

WhatsApp chatbot

Scattered data?

Internal tools

A copilot that finds, summarises and prepares; you approve.

Custom software

Does your product need AI?

Your own SaaS

AI features with usage limits and spend control.

SaaS development

Data and privacy

What happens to your company's data?

It depends on the provider and the plan, and it is reviewed before any real data is sent.

According to their policies, checked on 23 October 2026:

  • OpenAI does not use data from ChatGPT Business, ChatGPT Enterprise or its API for training by default (business data at OpenAI). In the API it keeps logs to monitor abuse for up to 30 days, unless the law requires more (data in the OpenAI API).
  • Anthropic does not train by default on inputs or outputs of its commercial products, such as Claude for Work or its API, unless you submit feedback or authorise it (Anthropic privacy center).
  • Google does not use your data to train or fine-tune models without your permission in Gemini Enterprise Agent Platform, which has incorporated Vertex AI since April 2026 (Google Cloud data governance). In the Gemini API, the free tier may use what you send to improve its products; the paid tier does not (Gemini API terms).

If you send personal data, the provider processes it on behalf of your company: the GDPR requires choosing processors with sufficient guarantees and a contract (Article 28 of the GDPR), and transfers outside the European Economic Area must be reviewed. If a customer talks to an assistant, the European AI Act requires that they know they are talking to an AI, unless it is obvious (Article 50).

We limit what is sent to the model to what is necessary and add access control, action logs and human approval where needed. This is not legal advice: your own adviser validates the legal fit.

Costs

How much does it cost to use AI models?

There are two costs: building the integration, from €1,000, and using the model, which is paid by consumption and can be capped from the design stage.

Providers charge for the API by tokens, the fragments of text you send and receive. Prices vary between models and change often, so we do not publish third-party rates: we measure your real consumption in the first version and keep it under control like this:

  • The right model for each task, not the most powerful one for everything.
  • Only the context needed in each query.
  • Limits per user or per plan, as in Rescrito.
  • Monitoring of the cost and usage of each model.

With the Build & Run model, AI consumption has limits agreed in advance. More in ways of working.

Fit

When does an AI integration fit, and when does it not?

It fits when there is a concrete task with data that already exists. If it is not worth it, we will tell you.

When it fits

  • There is a repetitive task involving text, documents or conversations.
  • The data already lives in your CRM, ERP, email or database.
  • You need to control which data leaves and how much is spent.
  • You want to add AI to your product or combine several models.

When we are not the right option

  • A business plan from ChatGPT, Claude or Gemini is enough for you and nobody needs to write software.
  • You want to rent individual developers, without a product brief.
  • You need legal advice.
  • Your budget is below €1,000 or delivery has to happen in under 2 weeks.
  • You do not yet know which process to improve: integration comes after an AI audit.

Frequently asked questions

Can ChatGPT be integrated into my company's CRM or ERP?

Yes, if the system has an API or allows its data to be read and written. The integration takes the information from the record, calls the OpenAI model or another provider's, and returns to the CRM or ERP a summary, a classification, a draft or an extracted data point. If your ERP works, it stays: we build only the missing part, with human approval where needed.

Can Claude or Gemini be integrated instead of ChatGPT?

Yes. ChatGPT is OpenAI's application, whose GPT models are used through an API; Claude is Anthropic's family of models, and Gemini is Google's. All three are integrated through an API (Gemini also on Google Cloud) and differ in quality depending on the task, price, speed and data terms. We integrate OpenAI, Gemini, Anthropic, DeepSeek, Perplexity or others depending on the task, and we design the integration so that the provider is interchangeable.

Is buying ChatGPT Enterprise the same as integrating the OpenAI API?

No. ChatGPT Enterprise, like ChatGPT Business or the Team and Enterprise plans of Claude, gives each team member the chat application with business controls. The OpenAI API makes the model work inside your systems, with your rules, and is paid by usage. They can be combined: the plan for the team and the API for processes.

Do OpenAI, Anthropic or Google train their models on my company's data?

According to their public policies, checked on 23 October 2026, OpenAI and Anthropic do not train by default on data from their business plans or their API, and Google does not on Google Cloud data or on data from the paid Gemini API. Personal accounts and the free Gemini tier have different terms, and providers may keep logs for a period to monitor abuse.

Can you add AI to our own product or SaaS?

Yes. In Rescrito and SacaNota, AI is the core of the product, with models from OpenAI, Gemini and others to generate text, summarise and analyse PDFs. In Rescrito there are also plans, usage limits per plan and consumption control, so the product is sustainable. If your product already exists, we build the missing part without redoing what works.

How much does it cost to integrate AI into a company?

Our projects start at €1,000. Above that, the price depends on what the AI has to do, which systems it connects to (CRM, calendar, documents) and how much information it handles: something simple is closer to the minimum and integrations with several systems push it up. Model consumption is paid separately, by usage, and can be limited per user, per plan or per task.

How long does an AI integration take?

Delivery starts at 2 weeks. A first working version can usually be validated within weeks if the scope is clear; complex integrations with several systems may need more phases. That is why we start with one well-chosen case: a single case reaches production sooner than five at once, and it lets you measure responses, costs and real usefulness before expanding.

Do we need to train our own model on our data?

To start with, almost never. It is usually enough to give the model good context, with instructions, examples and rules, and to connect it to your documents and data with RAG. RAG lets you update the content without retraining models and control permissions and sources: if a rate or a procedure changes, you just update the document. Training your own model requires more data, time and maintenance.

Where is Digitup Studio based?

Digitup Studio is a technical studio in Madrid founded in 2019, with an office at Calle Alcalá 211. The work is mainly remote and, if the project calls for it, we meet in person. We work with companies in Madrid and elsewhere in Spain, as well as international projects, in Spanish, English or Russian depending on the project.

Updated on · Content reviewed by Anton Zverev, CTO of Digitup Studio

Want to integrate AI into your systems?

Tell us which process you want to improve and with which tools. In 30 minutes with Anton, no commitment, we will look at the approach.

Tell us about my case Talk to Digitup