---
title: "AI platform: when it pays off and when it does not"
lang: en
type: knowledge
canonical: https://www.arocom.de/en/knowledge/ai/ai-platform
alternate: https://www.arocom.de/wissen/ki/ki-plattform
author: "Axel Roth"
published: 2026-08-07
last_updated: 2026-08-24
---

# AI platform: when it pays off and when it does not

An AI platform is one organisation's route to language models: one
contract, one central place where access is managed, one invoice,
plus shared assistants, access to your own documents and a log of
who used what. A single chatbot subscription brings none of that.
Very different products are sold under that same category label,
though, from plain chat access to an administrative layer across
several model vendors. Whether a platform beats individual licences
depends less on headcount than on whether the organisation has to be
able to prove how it uses AI.

## AI platform: building blocks, selection criteria and the trade-off behind them

"AI platform" has become the category label for very different
products: plain chat interfaces, add-on modules to an existing
office suite, and offerings that bundle several model vendors
under one shared administration. What ends up side by side in a
comparison is therefore often only alike on the surface.

This article first clarifies what the generic term actually
covers. Then come the building blocks, three use cases with a
named result, the selection criteria for smaller organisations,
and the question that decides between a platform and individual
licences. Details about Langdock come from the vendor's own pages,
retrieved on 7 August 2026, unless the sentence gives another
date. Everything else is attributed in the sentence.

## Chatbot or platform, where the difference lies

A chatbot subscription is access for one person. An AI platform is
access for an organisation. That sounds like semantics and in
practice decides three things: whether IT knows who has access,
whether there is one contract instead of many, and whether the
usage can be evidenced afterwards.

Technically, the difference is the one between an end product and
a layer that administers what sits on either side of it. A chatbot
wraps a model in an interface. A platform slots in between the
organisation and the models and governs who may use which model
with which data, where that data resides, and who receives the
invoice. The term operating layer is also common for this
intermediate layer; where it turns up in an offer, this is what
it means. Which is why the model name is the least important
selection criterion. It changes every few months, the operating
model around it stays.

## The six building blocks

At the start comes user management. New joiners get access,
leavers lose it,
and in larger organisations that runs through the existing staff
directory. Two standards govern it: signing in with the account
someone already holds rather than with a new password (the
technical term is SAML SSO), and creating and revoking access
automatically out of the directory (SCIM). Langdock documents
both, plus IP restrictions and session management.

Model access is where the products part company. Model-agnostic
platforms serve several vendors through one interface; Langdock
lists Google, OpenAI, Anthropic, Meta, Mistral and DeepSeek on its
models page and puts the number of models on its API page at over
40 (langdock.com, retrieved 11 August 2026). An outside count by
gewusst:KI on 4 August 2026 came to 35 listed models, seven of
them with no fixed EU region, among them GPT-5 Pro, o3 Pro,
DeepSeek v3.1 and Llama 4 Maverick. Vendor figure and outside
count thus diverge, and for a selection decision the split matters
more than the total anyway: a commitment to EU processing cannot
be held for every model. Which ones are available in your own
working area, which Langdock calls a workspace, is an
administrator's decision.

Assistants are called "Agents" at Langdock: a saved role with
instructions and attached files that you build once and then share
with a team or the whole organisation. Prompt know-how held by
individuals becomes a tool that works without those individuals.

Behind knowledge access sits a procedure that makes your own
documents searchable: they are split up and stored, the passages
matching a question are retrieved and handed to the model. The
technical term for this is
[RAG](https://www.arocom.de/en/knowledge/ai/rag). The appeal of a ready-made platform
is that you do not have to build this yourself. The price is a
lack of control over the details: Langdock's public documentation
says nothing about how documents are split, which search model is
used, or what the file size and document limits are.

Logging supplies the evidence. Without access logs and a usage and
cost overview, every statement about your own AI usage is a guess,
and that evidence is precisely the argument against
[shadow AI](https://www.arocom.de/en/knowledge/ai/shadow-ai).

That leaves billing: one invoice, one data processing agreement,
one notice period instead of twenty credit card subscriptions
buried in expense reports. The data protection side of this
bundling is covered in our article on
[GDPR-compliant AI](https://www.arocom.de/en/knowledge/ai/gdpr-compliant-ai).

## Three tasks that prove a platform

A rollout is decided by the first task somebody actually completes with it.
Three examples from organisations of this size.

In membership administration an email arrives asking about fees, deadlines
or evidence. The caseworker gets a draft reply that quotes the relevant
passage from the statutes and the fee schedule and supplies the reference,
and approves it in the same place. Federations of this size report 30 to 60
such enquiries a week in our initial conversations. That figure is not
measured; it is the order of magnitude we encounter there.

The project management of a funded project works from a grant decision with
its ancillary provisions, usually 20 to 40 pages. Out of it comes a list of
the obligations with deadline, form of evidence and owner, in a form you
can keep working on. That comes up once per decision, so ten times a year
with ten running projects.

The communications officer puts in a set of minutes or an annual report and
gets back three versions of the same text, for the website, the newsletter
and the noticeboard, in the organisation's agreed wording.

What all three examples have in common is that a particular document sits
at the start and a particular form of result at the end. How much time that
saves depends on the organisation, and no platform can promise it. Anyone
who cannot name such a task needs an assessment first and a tool second.

## Selection criteria for small and mid-sized organisations

The first question is not a technical one: what is the platform
supposed to make provable? Faster draft texts need less than an
organisation that must be able to explain to a supervisory
authority which data went where.

After that, six points matter. The table gives the question to put
to the vendor for each, and what marks an answer you can rely on.

**Six selection criteria, the matching question and the mark of an answer you can rely on.**

| Criterion | Question for the vendor | What makes the answer reliable |
| --- | --- | --- |
| Contract and jurisdiction | Who is the contracting party, under which law, where would a dispute be heard? | A signed data processing agreement naming company and registered office |
| Training | Are our inputs excluded from model training? | A clause in the contract, not a sentence on the product page |
| Prerequisites | Do we need a base plan we have to buy alongside? | The offer names the required licence explicitly |
| Nonprofit terms | Is there a programme for charitable bodies, and what does it contain? | A documented programme with a stated scope |
| Data processing | Does the EU commitment hold for every model we want to enable? | A commitment per model rather than one for the platform |
| Deployment model | From how many seats is a dedicated instance available? | A number rather than a conditional |

Three of these points are worth going into. A domestic contract
with a domestic place of jurisdiction is a different conversation
for many boards than a contract under foreign law. A discount for
charitable bodies is not automatically the cheaper offer, because
such programmes often carry different product variants from the
commercial range; at Langdock no such programme is documented,
neither on the pricing pages nor in the billing documentation, so
an association there pays the same list price as an industrial
company. And EU commitments often cover storage and not
necessarily processing. Langdock's EU commitment also has an
exception for models that are deployed globally, and the article
on [Langdock](https://www.arocom.de/en/knowledge/ai/langdock) breaks it down in detail.

The same principle holds for the deployment model. Dedicated
instances and self-hosting sit far above the size an organisation
of this class reaches with most vendors, at Langdock from 2,000
and 5,000 seats respectively. So "on-premise would also be
possible" is always followed by the question of the threshold that
applies from.

Download: Reviewing an AI platform (on the page: https://www.arocom.de/en/knowledge/ai/ai-platform)

24 review questions across six axes, with a findings sheet. Vendor-neutral.

## Platform versus individual licences: what the break-even depends on

The threshold decision-makers are really asking about separates
two cost types. A price per seat, meaning per person with access,
scales linearly with headcount: every additional person costs the
same amount again. A model with a
one-off setup and ongoing maintenance is independent of headcount.
The break-even sits where the other side's fixed cost meets the
seat cost over the same period, sensibly calculated across three
years.

Nobody should put that point value in a proposal, because two
items move it. On the seat side the rollout effort comes on top,
since configuration, assistants, usage rules and training occur
once; a fixed-price offer usually buries that item in the setup.
It lowers the break-even. On the fixed-price side, infrastructure
costs and the token costs billed by volume of text rise with
actual usage, which pulls the other way.
Your own figure therefore only emerges from two concrete quotes.
How arocom shapes a rollout is on the service page
[introducing Langdock](https://www.arocom.de/en/services/langdock).

Below about ten people, price does not decide anyway. There the
only thing that matters is whether the organisation has to be able
to evidence its AI usage. If it does, the cheapest route to that
evidence still runs through a platform, because a data processing
agreement and a log cost less than the first data protection
incident. There is no upper limit. From around a hundred seats the
task does change: more departments with their own requirements, an
IT lead at the table, and a staff directory that user management
is wired into. The Bitkom survey of 21 October 2025 shows the same
step in supply: 23 percent of companies with 20 to 99 employees
provide AI access of their own, and from 500 employees upwards it
is 43 percent.

## Two neighbours that are not AI platforms

A platform gets confused with workflow automation first. Langdock
has workflows in the product, and the temptation to equate that with
[n8n](https://www.arocom.de/en/knowledge/ai/n8n) is strong. The numbers say otherwise:
57 integrations with 754 actions at Langdock against 2,012 in the
n8n integration directory, five trigger types, 18 of the 57
integrations with an event-based trigger and the rest available
only as actions or through scheduled polling, a limit of 2,000
steps per run, and self-hosting only above the enterprise
threshold, while the n8n Community Edition can be self-hosted free
of charge indefinitely. The tools solve different problems.
Langdock workflows attach AI processing to data already connected
in your workspace. n8n connects line-of-business applications to
each other, including ones on nobody's integration list. The
details are in the article on [n8n](https://www.arocom.de/en/knowledge/ai/n8n).

AI inside your own website or product is a different case
altogether. A platform is a work tool for staff behind the login. A semantic
search in a customer portal, an editorial assistant in the CMS or
automatic data extraction from incoming documents belong in the
application itself, with model APIs wired in directly, your own
[vector database](https://www.arocom.de/en/knowledge/ai/vector-database) and full
control over data flow and cost. That is a development project.
The two do not exclude each other, and many organisations do both,
but the budgets and the people involved differ.

**Platform or your own integration?**

arocom establishes what your organisation actually needs, weighs
the options against each other and rolls out the platform you
choose. Write to us for a no-obligation conversation.

- [Request a first conversation](https://www.arocom.de/en/services/langdock#kontakt)

Which platform, which model, which legal framework: we answer
questions like these regularly on behalf of clients. For strategic
sparring at decision-maker level there is
[arocon](https://arocon.eu), the consulting brand in the arocom
family.

### What is an AI platform?

An administrative layer between an organisation and the language models. It bundles central user management, access to one or several models, shared assistants, access to your own documents, logging and joint billing. A single chatbot subscription offers none of that except the model access.

### Why would we need a platform, what do you actually do with it?

Whenever a recurring task has a named document going in and a named form of result coming out. Three examples from organisations of this size: a membership enquiry becomes a draft reply quoting the relevant passage of the statutes; a grant decision becomes a list of obligations with deadline and owner; an annual report becomes three versions of the text for website, newsletter and noticeboard. Without such a task, the survey comes first.

### What does the break-even between a seat price and a fixed price model depend on?

On two quantities. A seat model scales linearly with headcount, a fixed-price model is independent of it, and the break-even sits where the two meet over the same period. It is moved by the one-off rollout effort on the seat side, which lowers it, and by usage-dependent infrastructure and token costs on the fixed-price side, which raise it. A figure you can rely on therefore only emerges from two concrete quotes.

### How do I tell whether an offer is a platform or just a chatbot?

By three things that have to be in the data sheet. Central user management, ideally with sign-in through the existing company account and automatic maintenance of access, so that seats disappear when someone leaves. Logging that tells you afterwards who used what. And a contractual relationship with your organisation rather than with individual people. If one of those is missing, you are buying individual subscriptions in a bundle wrapper.

### Is there an upper limit above which a platform stops fitting?

No. From around a hundred seats the task does change: more departments with their own requirements, an IT lead who decides alongside you, and a connection to the existing staff directory. A dedicated instance instead of shared operation starts at Langdock only from 2,000 seats, self-hosting from 5,000.

### Are nonprofit terms for an AI platform automatically cheaper?

No, you have to do the maths. Nonprofit programmes often carry different product variants from the commercial range, and whether the cheaper variant is the one your organisation needs is decided by your own setup. Some vendors run no such programme at all; Langdock documents none, so an association there pays the same list price as an industrial company.

### Does an AI platform replace workflow automation like n8n?

No. Langdock workflows count 57 integrations with 754 actions, the n8n directory 2,012. Of the 57 integrations, 18 fire on events; the rest run as actions or through scheduled polling, a workflow is capped at 2,000 steps, and self-hosting starts only at the enterprise threshold. For automation between line-of-business applications you need n8n; for AI steps on data already connected, platform workflows are enough.

### Do we need a platform if we want AI in our website?

No, that is the other case. A platform serves staff behind the login. Semantic search, editorial assistants in the CMS or data extraction from incoming documents belong directly in your own application, with model APIs wired in and your own vector database. That is a development project and not a licence purchase.

## Read more

- [Langdock](https://www.arocom.de/en/knowledge/ai/langdock): The platform in detail
- [Shadow AI](https://www.arocom.de/en/knowledge/ai/shadow-ai): Why private accounts appear without an official offer
- [GDPR-compliant AI use](https://www.arocom.de/en/knowledge/ai/gdpr-compliant-ai): The four core questions before a rollout
- [RAG: Retrieval-Augmented Generation](https://www.arocom.de/en/knowledge/ai/rag): The technology behind knowledge access
- [n8n](https://www.arocom.de/en/knowledge/ai/n8n): Workflow automation, for contrast
- [Are we ready for AI?](https://www.arocom.de/en/ai-readiness): Ten questions as a printable checklist
- [Introducing Langdock](https://www.arocom.de/en/services/langdock): What arocom actually offers

## External resources

- [Langdock Security](https://langdock.com/security): Deployment models and their seat thresholds
- [gewusst:KI on Langdock](https://gewusst-ki.de/ki-tools/langdock/): Count of models and EU regions, as of 4 August 2026
- [Bitkom on shadow AI](https://www.bitkom.org/Presse/Presseinformation/Beschaeftigte-nutzen-Schatten-KI): Press release of 21 October 2025, source of the size breakdown