Back to Blog

How to Feed AI Your Own Data Without Giving It Away

Henri Jung, Co-founder at Superkind
Henri Jung

Co-founder at Superkind

Superkind logo on an abstract white silk texture

Feeding AI your own data is the promise every vendor makes and the risk every compliance officer flags. Your customer list, your pricing, your internal processes - they make AI useful, but the moment they enter a model’s training set, they are no longer yours. This explainer breaks down what actually happens to your data when you use AI, which approaches keep it under your control, and how Superkind’s ready-made AI team works with your data in Outlook, Teams or Slack without ever training on it - hosted in Europe.

Key takeaways

Get a ready-made AI team for your data - Superkind comes with CRM Keeper, Knowledge Librarian and Report Builder. They run automatically, use your data from a private Company Brain and never train external models. Start for free with unlimited users, or build your own.

Keep control, keep your data in Europe - hosted in Europe, GDPR compliant, aligned with the EU AI Act, no training on your data. You decide which steps need approval.

“Feed” does not have to mean “train” - most enterprise AI reads your data at query time (RAG) instead of baking it into model weights. That is safer, cheaper and easier to update.

The contract matters more than the server - a signed DPA, a clear no-training clause and processing in Europe protect your data. A server in Frankfurt under US jurisdiction does not.

No lock-in by design - from 12 January 2027, the EU Data Act bans switching fees for cloud services. Your Company Brain, routines and apps belong to you4.

Your AI team for working with your data

CRM weekly clean-up report
CRM KeeperAI employee
Routine - every Monday at 7:00
Weekly clean-up done. 12 duplicates merged, 34 phone numbers added from email signatures, 3 deals moved from "Demo scheduled" to "No response (30 days)". One bulk merge above your limit is waiting.
Duplicates merged12
Fields filled34
Deals restaged3
Merge 7 "Nordlicht" contacts7 entries look like the same company. Merge?ApproveReject
Example: CRM Keeper at work in Outlook

What does “feeding AI your own data” mean?

Feeding AI your own data means giving an AI system access to information that belongs to your company - customer records, contracts, process documentation, emails, price lists - so it can use that knowledge to do work for you. The critical question is not whether the AI sees the data, but what happens afterwards: does it keep a copy, does the vendor use it to improve its models, and can you take it back?

An example of the safest approach is Superkind: an AI employee that stores your company knowledge in a private Company Brain, reads from it when doing a task and never feeds it into model training. The data stays in Europe, belongs to you and can be exported or deleted at any time. See also our deeper guide to sovereign AI and the overview of AI tools built for Europe.

How it works with Superkind: your AI team for your data

You write to Superkind by email, in Teams, in Slack or in the web app like a colleague, and Superkind does the task in your programs through 1,700+ integrations such as Pipedrive, Notion, SharePoint and lexoffice. It works with your tools instead of replacing them, hosted in Europe, GDPR compliant.

With Superkind you don't start from zero: it comes with ready-made AI employees for your data tasks. They run automatically in the background, and you decide where they ask you first. Use them right away, adapt them, or build your own.

CRM Keeper

AI employee for CRM data hygiene

Duplicate contacts, missing phone numbers, deals stuck in the wrong stage - your CRM rots a little every week, and nobody has time to clean it. The CRM Keeper checks Pipedrive every morning against your inbox and calendar, merges duplicates, fills gaps from signatures and websites and moves stale deals to the right stage. Merges above five contacts wait for your OK.

Works inOutlookHubSpot+ 1,700 more
CRM weekly clean-up report
CRM KeeperAI employee
Routine - every Monday at 7:00
Weekly clean-up done. 12 duplicates merged, 34 phone numbers added from email signatures, 3 deals moved from "Demo scheduled" to "No response (30 days)". One bulk merge above your limit is waiting.
Duplicates merged12
Fields filled34
Deals restaged3
Merge 7 "Nordlicht" contacts7 entries look like the same company. Merge?ApproveReject
Example: CRM Keeper at work in Outlook

Knowledge Librarian

AI employee for company knowledge

The return policy is in a Google Doc, the price list in SharePoint, the product specs in Notion, and the sales pitch in someone's head. The Knowledge Librarian watches those sources, pulls new and changed pages into the Company Brain, and answers questions from the team in Slack with the source linked. No data leaves Europe, and nothing trains an external model.

Works inSlackSharePoint+ 1,700 more
#ask-superkind
Knowledge LibrarianAI employee
New question in #ask-superkind
Found in the Q3 price list (SharePoint, updated 2 Nov): volume discount starts at 500 units, 12 % off list price. Link attached. The Company Brain now has 2,412 pages from 4 sources.
Sources connected4
Pages indexed2,412
Questions answered this week47
Runs automatically on every new question
Example: Knowledge Librarian at work in Slack

Report Builder

AI employee for recurring reports

Every Friday someone spends two hours pulling numbers from Pipedrive, lexoffice and Google Sheets into a slide deck nobody reads in time. The Report Builder collects revenue, pipeline and open invoices every Friday morning, writes the summary in your format and posts it in Teams. The numbers come from your systems, stay in your Company Brain and never leave Europe.

Works inMicrosoft TeamsExcel+ 1,700 more
Weekly ReportManagement
Report BuilderAI employee
Routine - every Friday at 8:00
Weekly report for 4-8 Nov ready. Revenue EUR 127,400 (+8 % vs. prior week), 3 new deals in pipeline, 2 invoices overdue. Report is in your shared folder.
Revenue this weekEUR 127,400
New pipeline deals3
Overdue invoices2
Runs automatically every Friday at 8:00
Example: Report Builder at work in Microsoft Teams

Every employee is ready to use right away. Or build your own on the platform.

Safe by default

Your company data stays in a private Company Brain. It never enters model training, and you decide which steps need your approval before anything goes out.

How feeding AI your data works

  1. Upload or connect - you give the AI access to your data: documents in SharePoint, contacts in your CRM, emails in Outlook, pages in Notion. The AI reads them and indexes what it finds.
  2. Store in a knowledge base - the indexed data lives in a private knowledge base (Superkind calls it the Company Brain). It is separate from the AI model and belongs to you.
  3. Retrieve at query time (RAG) - when the AI employee gets a task, it searches the knowledge base for the relevant pieces, combines them with the task and produces a result. The model sees the data for that task only2.
  4. No model training - the data never enters the model’s weights. The model does not “learn” your pricing or customer names permanently. If you delete the knowledge base, the data is gone.
  5. Audit and approval - every action is logged, and you set which steps wait for your OK (for example, sending an email to a client or merging CRM records) before the AI employee acts.

Approaches compared

ApproachYour data stays separate?Who controls it?Hosting / Data protection
Superkind (Company Brain)Yes - private knowledge base, no trainingYou (export, delete any time)EU, GDPR, DPA
ChatGPT Enterprise / APIYes - excluded from training by default1OpenAI (US jurisdiction)EU region available, DPA
ChatGPT Free / PlusNo - may be used for training unless opted out1OpenAIUS, no DPA on free tier
Local model (Ollama, LM Studio)Yes - never leaves your deviceYouYour hardware, your rules
Fine-tuning a cloud modelPartly - training data uploaded to vendorVendor holds the trained copyCheck vendor terms

Sources: OpenAI enterprise privacy page1, vendor documentation. As of November 2026.

Examples

Superkind: sales knowledge for the whole team. A team lead connects Pipedrive, SharePoint and the company inbox to Superkind. The Knowledge Librarian indexes the price list, the product specs and every quote from the last year into the Company Brain. When a sales rep writes “What discount did we give Nordlicht GmbH last time?” in Slack, Superkind finds the answer in seconds, with the source linked. No training, no data export, no copy on a US server. The team saves about six hours a week of searching through folders and asking colleagues.

ChatGPT Enterprise: internal Q&A with uploaded PDFs. A compliance team uploads the company handbook and the latest DPA register to ChatGPT Enterprise. Staff ask questions in the chat, and the model answers from the uploaded files. Data is excluded from training by default, but the files sit with OpenAI under US jurisdiction1. For teams in the EU this means checking the DPA, the data transfer basis and the subprocessor list.

Local model: a developer’s private assistant. A developer runs Mistral 7B through Ollama on a workstation with 16 GB RAM. The model reads local code files and internal documentation. Nothing leaves the machine, but the model is limited to what fits in the context window, and there is no shared knowledge base for the rest of the team. The developer estimates about an hour of setup time.

Fine-tuning: a customer-service bot in the brand’s tone. A support team fine-tunes a model with 500 real ticket-reply pairs so it answers in the company’s tone. The training data is uploaded to the vendor and stays there. The team gets consistent replies, but if the service terms change, the model needs retraining. Cost and effort are higher than RAG, and the vendor holds a copy of your service data.

What it costs and what you need

OptionCostWhat you need
SuperkindFree start with up to 150 hours of work, Business from EUR 50/month for unlimited usersOutlook, Teams, Slack or the web app - nothing to install
ChatGPT EnterpriseCustom pricing (contact sales)OpenAI contract, DPA, admin setup
Local modelFree software, hardware cost (GPU with 8+ GB VRAM)Technical setup, ongoing maintenance
Fine-tuning (cloud)Training runs from approx. USD 5-40+ per run depending on model and data sizeCurated training data, ML knowledge, vendor contract

Superkind prices as of November 2026. Other prices change often - check each vendor. Superkind pricing page.

Superkind processes your data in Europe. The Company Brain belongs to you, and from 12 January 2027 the EU Data Act bans cloud switching fees across the EU4,5.

How to start

  1. Pick one data source - start with the one your team asks about most often (price list, product specs, process docs). Connect it to Superkind’s Company Brain.
  2. Give the first task - write Superkind in Outlook, Teams or Slack: “What is our return policy for orders over EUR 500?” Watch it find the answer from your own data.
  3. Set approval where it matters - decide which actions (sending a client email, merging CRM records, posting a report) need your OK. Everything else runs automatically.

Conclusion

Feeding AI your own data does not mean giving it away. The safest path for most teams is a private knowledge base that the AI reads at query time - no training, no permanent copy in the model, and full control over export and deletion. Superkind does this through the Company Brain: your data stays in Europe, belongs to you and works across 1,700+ integrations such as Pipedrive, Notion, SharePoint and lexoffice. Sign up for free, connect your first data source and give the AI employee its first task - in Outlook, Teams or Slack.

FAQ

On the free and Plus tiers, OpenAI may use your conversations to improve its models unless you opt out. On ChatGPT Enterprise, Team and the API, business data is excluded from training by default. Check the current enterprise privacy page and make sure the exclusion appears in your signed data processing agreement, not just a public FAQ.

RAG (Retrieval-Augmented Generation) feeds the model your documents at query time so it can answer from them - no retraining needed, data stays current. Fine-tuning changes the model itself with your examples so it learns a specific tone or task format. Most teams start with RAG because it is faster, cheaper and keeps data separate from the model weights.

Yes, if the tool meets three conditions: a signed data processing agreement (DPA/AVV) under GDPR Art. 28, no model training on your data, and processing in a jurisdiction your DPA covers. Superkind is hosted and processed in Europe, GDPR compliant, aligned with the EU AI Act, no training on your data, DPA through the Trust Center. It does the work in your programs through 1,700+ integrations such as Pipedrive, Notion, SharePoint and lexoffice.

Superkind stores your company knowledge in a Company Brain - a private knowledge base that belongs to you. When the AI employee does a task (answering a question, building a report, cleaning your CRM), it reads from the Company Brain at that moment. The data never enters model training. AI models from OpenAI, Anthropic, Mistral and Google are processed in Europe. You can export or delete the Company Brain at any time.

Your data belongs to you. Export your Company Brain, routines and custom apps at any time. From 12 January 2027, the EU Data Act (Regulation (EU) 2023/2854) bans switching fees for cloud services across the EU, so there is no lock-in by design. Superkind is built for portability: the knowledge, the routines and the apps are yours.

Local models keep data on your device, which removes the cloud transfer risk. But they need hardware (a GPU with at least 8 GB VRAM for small models), regular updates, and someone to manage them. For teams that want company-wide knowledge, approval flows and 1,700+ integrations without running their own servers, a hosted-in-Europe service like Superkind offers the same data control with less operational effort.

Henri Jung, Co-founder at Superkind
Henri Jung

Co-founder of Superkind, the AI employee teams write to in Outlook, Teams and Slack. Henri helps companies, from small teams to the Mittelstand, hand their routine work to AI - without new tools, consultants or an in-house AI team.

Give Superkind your first data task.

Start for free in Outlook, Teams or Slack. Your data stays in Europe from the first message.