company-gpt
ai inside the company. compliant. adopted.
Your people already use LLMs - just not officially, not governed, not compliant. We bring it in-house: vendor-neutral selection, clean integration, GDPR-aligned, with adoption that still holds three months in.
most are already doing it. just not officially.
Three numbers you won't enjoy - and a fourth where we run against that picture.
share of knowledge workers using LLMs for work - mostly privately, on personal accounts.
average level of governance over that shadow usage. No audit logs, no permissions, no GDPR trail.
What happens when customer or personnel data lands in a personal ChatGPT tab? Exactly.
Governed, compliant, for everyone. A Company-GPT you actually have a grip on.
is this for you?
book a discovery callwhich platform fits you?
Neutrally framed, not sold. The right choice depends on your Microsoft footprint, compliance requirements and your use-case mix - not on whichever vendor has the loudest marketing.
microsoft 365 copilot
- Strengths
- Native integration with Outlook, Word, Excel, Teams, SharePoint. Data stays inside the M365 tenant - no extra platform for staff to learn.
- Ideal fit
- Organisations on a consistent M365 footprint (E3/E5 or Business Standard/Premium) with existing licences.
- Data / hosting
- Processing inside the Microsoft 365 tenant, EU Data Boundary configurable. Flex-routing caveat: at peak load, requests can be processed outside the EU by default - explicitly disablable.
- Integrations
- Deepest integration into the Microsoft stack. Graph connectors for external sources (Confluence, Salesforce, ServiceNow).
- Typical limitations
- Per-user add-on licence (around $30 on top of the M365 base), limited model choice, value strongly tied to M365 data quality.
langdock (made in germany)
- Strengths
- Model-agnostic (OpenAI, Anthropic, Mistral, Google), assistants and RAG out of the box, fast platform-wide rollout - even without M365.
- Ideal fit
- SMBs and mid-market on a mixed stack (Google Workspace, Slack, hybrid) or with strict EU-compliance demands.
- Data / hosting
- EU-only hosting, no US routing, ISO 27001 and SOC 2 Type II, listed on the German federal BeKI marketplace. On-premises available.
- Integrations
- SharePoint, Confluence, Drive, Slack, Teams, API. Model can be swapped per use case - no model-level vendor lock-in.
- Typical limitations
- Less tightly integrated into M365 than Copilot, separate UI layer (staff learn a new tool). Feature set evolves quickly.
claude enterprise
- Strengths
- Very strong reasoning, 1M-token context, MCP integrations, Claude Code for technical teams. Clear zero-data-retention policy.
- Ideal fit
- Tech-savvy teams with document- or code-heavy use cases - analysis, research, engineering.
- Data / hosting
- US processing by default. inference_geo parameter available for US-only. No training on customer data, BAA and audit logs from Enterprise tier.
- Integrations
- Projects, MCP servers, native connectors to Drive, Slack, GitHub. API-first - good for in-house extensions.
- Typical limitations
- UI primarily English, no EU-only guarantee without extra effort, less Office-native integration.
- Deeply embedded in Microsoft 365, licences already in place → Copilot.
- Model flexibility, mixed stack, EU-only mandate → Langdock.
- Highest reasoning quality for analysis or code, tech-savvy team → Claude Enterprise.
- Existing OpenAI setup, lots of tooling variety → ChatGPT Enterprise (for many SMBs more expensive and with weaker EU guarantees than the three above).
- Unclear? → We weigh compliance, use-case mix and existing systems in discovery.
concrete. not abstract.
platform selection
We match compliance demands, existing stack and use-case mix against the platforms - and justify why which one.
architecture & integration
RAG over your knowledge, connectors to SharePoint, Confluence, Drive, your database. Clean sources instead of hallucination risk.
governance & compliance
GDPR setup, roles and permissions, audit logging, DPA with the platform vendor. Auditable from day one.
use-case workshops
Prioritised pilot use cases with measurable ROI - not "now all 50 people draft emails with AI".
custom assistants & agents
Domain-specific GPTs, prompt libraries, workflow automation on top of the chosen platform.
adoption & enablement
Champions programme, training, impact measurement. So the system is still in use in month 4.
three examples. from the mid-market.
Mid-sized broker, several thousand active contracts in PDF/Word, manual clause lookups on every claim.
gebaut: RAG assistant over the contract portfolio, integrated into the claims workflow.
Clause lookup from hours to seconds.
Platform: Langdock with its own RAG layer. Live in a few weeks, EU-hosted.
Commercial law firm with grown-in internal case knowledge, client files spread across DMS and SharePoint.
gebaut: Research assistant over the in-house knowledge base, with source citations per answer.
Faster pre-research, less duplicated work.
Platform: M365 Copilot - data stays in the tenant. Custom assistant for the research workflow.
Custom-machine builder with a field service force, maintenance manuals scattered as PDFs on works laptops.
gebaut: Service bot over the maintenance documentation, accessible from the technician's phone.
Answers on site instead of callbacks to HQ.
Platform: Claude Enterprise - large context for long manuals, MCP connector into the doc store.
ready to get concrete?
book a discovery callvendor-neutral. not platform-besotted.
We pick the platform by three criteria - in exactly this order:
- 01use case - where is the pain today, which use cases show visible effect in 12 weeks.
- 02compliance & data residency - industry, regulator, EU mandate, DPA requirements, sensitive data classes.
- 03team readiness & estate - what's running today (M365 / Google / hybrid), who runs it after rollout, how tech-savvy is the workforce.
What doesn't feed into the decision: what's trending on LinkedIn, which vendor is our favourite partner, who put on the slickest demo.
Preach one platform regardless of fit. We're neither a Microsoft consultancy nor a Langdock reseller - we recommend the platform that fits your stack and your compliance.
Platform recommendation with pro/con per alternative in the concept doc after discovery. Including effort, licence band, and migration path.
discovery to rollout - no pilot graveyard.
discovery
Use-case workshop with the people who'll actually use it. Compliance inventory. Take-stock of your existing IT.
Discovery doc with platform shortlist, pilot use cases, risks.
platform & concept
Vendor decision, pilot use cases, architecture, migration plan. A clear recommendation, not a slide deck.
Fixed-price offer, architecture sketch, milestone plan.
pilot & build
Tenant setup, connectors, 2–3 use cases live in your environment. Pilot group onboarded, first impact measurable.
Runnable pilot in your tenant, docs, onboarding material.
rollout & enablement
Wave-by-wave rollout, champions programme, training, impact measurement. Iterative, not big-bang.
Adoption playbook, training material, impact dashboard.
On request we run the ongoing operation under retainer - maintenance, new use cases, licence and connector management. Separate contract, no minimum-term games.
fixed price per stage. licences separate.
We work on fixed prices per stage - pilot, rollout, optional managed operation. Range indication after discovery, final price before build.
Licence and token costs of the chosen platform run directly with the vendor (Microsoft, Langdock, Anthropic) on their terms - we help configure the right SKU, but we don't take a margin in between.
pilot
2–4 weeksOne department, 2–3 use cases, runnable in your tenant. Ideal for testing whether the platform choice holds.
rollout
2–3 monthsCompany-wide rollout. Connectors, governance, champions programme, training, impact measurement.
managed operation
retainerOngoing operation, new use cases, licence and connector management. Separate contract, no minimum term.
Platform licences (Copilot, Langdock, Claude - directly with the vendor on their terms), token costs of the models, M365 base licences, hardware. We disclose the ranges transparently in the concept doc.
more in the journal.
let's talk about company-gpt.
30 minutes, no pitch deck. We look at your concrete use-case mix, your compliance demands and your existing estate - and tell you honestly which platform fits. Even if the honest answer is: pilot first, decide later.