Data location
Lokalaise
Your data center. No document ever leaves the company.Cloud AI
the vendor's serversBuild in-house
own infrastructure, self-operatedLokalaise connects files, SharePoint, DMS, and and email into one knowledge layer that backs every answer with your documents and automates recurring work — run locally, GDPR-compliant, and and fully managed by us.
100% locally operatedGDPR-compliantproductive in 14 days
Key facts
Trusted by teams whose data must not leave the company.
100%locally operatedall data stays inside your company
14 daysto the first use casefrom kickoff to productive use
GDPRcompliantdata protection with no extra effort for you
1 : 1every answer with a sourceevery statement leads to the original document
01The situation
Right where cloud AI fails,
your most important data lives.
Your most valuable knowledge sits in files, contracts, emails, and and scans — scattered, sensitive, and and off-limits for public cloud AI.
1.1
Knowledge is scattered
PDFs, file servers, SharePoint, emails, and and scans — your knowledge is fragmented across teams and systems.
1.2
Cloud is not an option
Project, customer, and and contract data is too sensitive to entrust to public cloud AI.
1.3
No AI team in-house
Building, operating, and and rolling out AI takes data engineers and AI specialists — and the market barely offers them.
1.4
Tools without tangible impact
Copilot and the like are running — yet efficiency and workflows don't improve measurably.
The result: hours are lost to searching, processes stay manual — and risks remain hidden in the data.
02Lokalaise Knowledge Layer
One knowledge layer from all your sources.
The Knowledge Layer connects your systems into a searchable, permission-aware knowledge layer — inside your company network.
Internal company network
Sources — existing systems · no migration
- File server
- SharePoint
- DMS
- Scans
Knowledge Layer
indexedpermission-awareversioned
One index across everything — every match knows its source, permissions and version.
Answers with sources
Every statement backed by a source reference.
Automated workflows
Recurring processes run on their own.
You don't run an AI project. You get a finished, measurable result — as a managed local stack.
03Lokalaise platform
One platform for knowledge and work.
Four capabilities, one system. Pick one — on the right you'll see what it looks like in everyday work.
Lokalaise Chat · Project team
localpermission-aware
M. Petersen · 09:14
Which change orders are still open for Elbkai 12?
Lokalaise · 09:14
Three change orders are open: N-07 (foundation), N-11 (MEP) and N-14 (facade). N-11 has been awaiting the client's approval since May 21.
Change_orders_EK12.xlsxSharePoint · as of today, 07:40
Live example — ask conversationally, get an answer with evidence
04Agents
Digital coworkers — with owners, rules, and a duty to prove themselves.
Lokalaise agents take on defined tasks, ask when they get stuck — and improve night after night.
4.1
Clear ownership
Every agent has an Agent Owner — a person who is accountable, often IT or Lokalaise — and one or more Agent Feedback Givers the agent contacts when it has questions.
4.2
Starts in shadow mode
Agents are set up for defined tasks and initially run in parallel with the existing process — they have to prove themselves before they take over.
4.3
Asks instead of guessing
Agents notice when they can't do something and actively request feedback — for example via a Teams message to the people in charge.
4.4
Improves every night
Failed runs feed into a new version overnight, which is tested against the current one. If it improves, the Agent Owner is pinged automatically and decides on deployment based on evidence.
4.5
Strict boundaries
What an agent is not allowed to do is firmly defined — for example, approving invoices only up to defined amounts. Anything beyond that goes to a human.
4.6
Time for value creation
Agents bring transparency, traceability, and efficiency to recurring processes — and free up time for real value creation.
05Data sovereignty
Your data stays with you.
Completely.
The entire AI runs on your infrastructure. Every answer is traceable — down to the source.
Permissions
Lokalaise
Permission-aware down to the answerCloud AI
detached from your permission modelBuild in-house
build it yourselfTraceability
Lokalaise
Source, version, and and history for every answerCloud AI
black boxBuild in-house
depends on your buildAuditability
Lokalaise
Every answer and every agent run in the audit trailCloud AI
little insight into processingBuild in-house
build it yourselfTime to productive
Lokalaise
First productive answers from day 1, first use case live after 14 daysCloud AI
1 day — but with no connection to your company knowledgeBuild in-house
3+ months to first valueModel choice
Lokalaise
Models swappable, knowledge layer staysCloud AI
tied to the vendorBuild in-house
every migration is a projectOperations
Lokalaise
Managed by LokalaiseCloud AI
—
Build in-house
requires your own AI team| Criterion | Lokalaise | Cloud AI | Build in-house |
|---|---|---|---|
| Data location | Your data center. No document ever leaves the company. | the vendor's servers | own infrastructure, self-operated |
| Permissions | Permission-aware down to the answer | detached from your permission model | build it yourself |
| Traceability | Source, version, and and history for every answer | black box | depends on your build |
| Auditability | Every answer and every agent run in the audit trail | little insight into processing | build it yourself |
| Time to productive | First productive answers from day 1, first use case live after 14 days | 1 day — but with no connection to your company knowledge | 3+ months to first value |
| Model choice | Models swappable, knowledge layer stays | tied to the vendor | every migration is a project |
| Operations | Managed by Lokalaise | — | requires your own AI team |
§ 1
On your hardware
All processing — indexing, models, answers — runs in your data center. There is no outbound connection to external AI providers.
§ 2
Permissioned & auditable
Whoever asks a question only sees what they are allowed to see in the source systems. And if anyone wants to know: the audit trail shows who asked what, when — and which sources were used.
§ 3
No vendor lock-in
When a better model appears, we swap it in — your knowledge layer, agents, and and permissions stay unchanged. You commit to your data, not to a vendor.
06Gets better every night
An AI that adapts to your company.
The Lokalaise Knowledge Layer and Agent Layer keep adapting to your company, every night.
Every night
your AI keeps adapting to your company — entirely on its own.
6.1
Learns every night
Usage signals, feedback, and source quality automatically improve results, ranking, and answer precision.
6.2
Agents compound
Every implemented agent makes the system noticeably better and faster — and the next agent builds on everything the previous ones have learned.
6.3
Your knowledge grows with it
New documents, projects, and sources continuously expand the knowledge layer.
07Industries
For industries whose data must not leave the company.
Confidential documents, established processes, no in-house AI team — many industries know this pattern. That is exactly what Lokalaise is built for.
Construction & planning
Plans, expert reports, site diaries, contracts
Engineering & plant construction
Specifications, standards, test protocols
Industry & manufacturing
Work instructions, QA documentation, maintenance
Legal & consulting
Contracts, briefs, correspondence
Finance & insurance
Contracts, policies, audit reports
Research & public sector
Applications, official notices, tender documents
A fit if
- Your expertise lives in documents, plans, and emails
- Your data is too sensitive for the cloud
- You don't have an in-house AI team
- Approvals, reviews, and knowledge searches are done by hand
Your industry isn't listed? What matters is the pattern — not the industry.
08Rollout
Productive in 14 days, not months.
We start with an explicit use case, prove the value in real numbers, and grow from there.
Day 1–2
Goal & metrics
We define the first use case, the data sources, and the metrics the value must be measured against from day 14.
Day 2–12
Integration & workflow
Connect sources, index, inherit permissions — and the first use case takes shape on your real data.
Day 12–14
Go-live
The use case goes live in day-to-day operations. From day 14, the value is measured against the agreed metrics.
Day 1Day 7Day 14
Goal & metrics
Integration & indexing
Use case & permissions
Go-live on real data
measurement starts on day 14 — if the value shows in real numbers, we scale across teams, sources, and departments
09Model
As predictable as a subscription, with no surprises.
A clear, recurring model: one-time onboarding, the platform including hardware, and optionally managed agents. We share concrete pricing openly in a first conversation.
Setup & onboarding
one-time· per node
Setup, connectors, indexing, and permissions — all the way to go-live.
Platform incl. hardware
monthly· per node
Inference, models, agents, and the knowledge layer — with evidence and audit.
Managed agents
monthly· per agent
Effective agents, continuously managed: monitoring, tuning, and benchmarks.
We size the hardware to match your users, agents, and workload — you only pay for what you actually need.
10Questions
What decision-makers want to know first.
The most common questions before a first conversation — answered in detail.
No. All processing — indexing, models, answers, and agent runs — happens on hardware in your data center or your internal company network. There is no outbound connection to external AI providers, neither for requests nor for training or telemetry. Data sovereignty stays entirely with you, and the GDPR assessment becomes significantly simpler because no data is transferred to third parties.
Every statement in an answer is backed by the source it comes from — including document, version, and exact location. You can check any claim against the original with a single click. Anything that cannot be substantiated from your documents is not claimed; the system will say it found no reliable source instead of guessing.
No. Lokalaise is a managed service: we handle setup, operations, maintenance, model updates, and the ongoing tuning of the platform. Your IT stays involved where access, permissions, and infrastructure are concerned — but does not have to hire AI specialists or operate models. We also set up agents together with your subject-matter experts.
First productive answers on your company knowledge are possible from day 1 after the sources are connected. The first defined use case goes live after 14 days. Measurement starts on day 14: before the start, we agree on metrics the value must show up in — and only once the numbers prove the value do we scale across additional teams, sources, and departments.
Yes, completely. Lokalaise inherits the user roles and permissions from your source systems — such as SharePoint, file servers, or DMS. Each person only sees content in answers that they would also have access to in the source system. In addition, the audit trail transparently logs who asked what and when, and which sources were used.
Connected systems include SharePoint, file servers (SMB), email (Exchange), Confluence, as well as ERP and DMS systems; PDFs and scans are also made searchable via text recognition. There is no data migration: your source systems remain the system of record, Lokalaise indexes them and keeps the index up to date. If a connector is removed, your data stays unchanged exactly where it has always been.
Then you benefit from them without rebuilding anything. The models are an interchangeable component of the platform: we evaluate new models, test them against the current setup, and swap them in when they are demonstrably better. Your knowledge layer, your agents, and your permissions remain unchanged throughout — there is no lock-in to a specific model or provider.
Through clear ownership and strict rules. Every agent has an Agent Owner who decides on deployments, and Feedback Givers the agent contacts when it has questions. New agents initially run in shadow mode in parallel with the existing process and have to prove themselves there. What an agent is not allowed to do is firmly defined — for example, approvals only up to defined amounts — and every run is traceable in the audit trail.
11Insights
From our blog
Practice, regulation, and architecture around sovereign AI — documented by the team that builds the platform every day.
Local AI infrastructure
Your AI runs on your premises — not on someone else's servers.
Connected knowledge, grounded answers, and automated workflows — on your own infrastructure.
100% locally operated · GDPR-compliant · productive in 14 days
Contact sales