Use cases by function6 use cases
Internal AI tools your team will actually open
Your employees are already using AI with company data. KPMG and the University of Melbourne found that 48% of employees have uploaded sensitive company information to public AI tools, and that bans do not work: 67% of employees at companies that ban generative AI report doing it anyway. The useful question is not whether people use AI at work but whether they can use it with company data safely.
The builds here answer that in different ways: an assistant that answers from SharePoint and Drive and respects who can see what, an MCP server that gives Claude, ChatGPT or Copilot scoped access to your ERP and CRM, a private AI gateway with single sign-on, redaction and budgets, and reporting that explains itself, from a Monday brief to plain-English questions over your database.
What they share is an approach. They live where people already work, usually Teams or Slack. They cite their sources or say they do not know. They enforce permissions at the moment of each question. And they are measured against real questions from your staff, collected before the build starts.
Internal tools and knowledge systems I can build
A company assistant that answers from SharePoint and Drive, and respects who may see what
Answers staff questions in Teams and Slack from SharePoint, Drive and Confluence, cites every source, and only searches what the person asking may open.
SharePoint and Teams / Google Drive / Slack / Confluence or Notion / Microsoft Entra ID or Google groups
A Monday KPI brief that explains last week's numbers instead of just charting them
Computes last week's KPIs in SQL, checks the data is complete, finds what moved and why, and posts a short written brief to Slack at 07:00 every Monday.
Shopify / BigQuery or Google Sheets / GA4 and the ad platform APIs / HubSpot / Accounting and payments
Connect your ERP and CRM to Claude, ChatGPT or Copilot without handing over the keys
Gives the assistants your staff already use a few scoped tools into the ERP, CRM and databases: read-only by default, writes behind approval, every call logged.
SAP Business One / HubSpot / Postgres or SQL Server database / Claude, ChatGPT and Copilot Studio / Microsoft Entra ID or another OAuth provider
A private ChatGPT for every employee, behind single sign-on, with redaction and team budgets
A company chat app behind single sign-on that strips client data and secrets, routes to approved EU or self-hosted models, logs use, and gives each team a budget.
Microsoft Entra ID or Google Workspace / Slack or Teams / OpenAI, Anthropic and Mistral / Open models on EU infrastructure / Open WebUI, LibreChat or a custom interface
Agency client reports with commentary drafted from the data and approved by the account lead
Loads every client's ad and analytics data, measures it against their own targets, names only causes it can prove, and drafts commentary in the client's language.
Google Ads, Meta Ads and GA4 / Supermetrics or Funnel / BigQuery or Google Sheets / Looker Studio / HubSpot
Plain-English questions over your own database, with the SQL shown and checked
Answers plain-English questions from a read-only copy of your database, shows the SQL and definitions behind each answer, and asks when a question is ambiguous.
Postgres or BigQuery / dbt or plain SQL views / Metabase / Slack / Excel export
What makes an internal AI tool get used
It lives where people already work
A separate portal gets visited twice. An assistant in Teams or Slack gets used daily. Custom internal Slack apps also keep the full API rate limits that unlisted third-party apps lost in 2025.
It cites, or it says it does not know
Every answer links to the document it came from. When nothing relevant is found, the assistant says so, which is what earns trust for the answers it does give.
Permissions are enforced per question
The assistant searches only what the person asking is allowed to open, checked at query time against the same groups your file permissions use.
Numbers come from code
In reports and database answers, figures are calculated by queries you can inspect. The model explains them; it never computes them.
It is tested on real questions
A set of real questions with known good answers runs on every change, so improvements are measured rather than assumed.
Frequently asked questions
Can I connect our ERP or CRM to ChatGPT or Claude?
Yes, through the Model Context Protocol (MCP). Many systems now have official MCP servers, including HubSpot, Slack, Xero and QuickBooks Online. For systems without one, such as SAP Business One or an internal database, a custom MCP server exposes a small set of scoped, logged tools that Claude, ChatGPT and Copilot Studio can all use.
How do I stop employees pasting company data into ChatGPT?
Give them something better. Bans are ignored, as the KPMG research shows. A private gateway with company single sign-on, approved models, redaction of personal data and secrets, and assistants that already know your context removes the reason to use a personal account.
Do we need Microsoft 365 Copilot or a custom assistant?
If most staff already have Copilot licenses and your knowledge lives in Microsoft 365, start there. A custom assistant makes sense for staff without licenses, for knowledge spread across several systems, for strict EU hosting requirements, or when you need behavior Copilot does not offer, such as answering only from approved documents.
How do I build an internal AI assistant for company documents?
Connect the document sources, mirror their permissions, index the content with metadata about owner and freshness, retrieve only what the asking user may see, and require a citation for every answer. Then test it on a few hundred real questions before rollout. The permission model is the part to get right first.
Also browse: Outlook and Microsoft 365 · Slack · Gmail and Google Workspace · SaaS · all use cases