yonoo.
Investor brief · Antler · September 2026 · Confidential

You only
need one.

Yonoo is the orchestration layer above the AI models. One question goes in. The engine best suited to it answers — and when the stakes are high, several answer and Yonoo compares them.

02
The problem

AI is everywhere in the enterprise, and trusted nowhere that matters.

For the engineer

The answer exists. The evidence is in four systems.

Requirements sit in one tool, test results in another, design data in a third, the decision history in somebody's inbox. The work is not thinking — it is finding the right source, checking it is current, and proving it.

For the company

Dozens of models. No way to tell a good answer from a confident one.

Every team picks a different tool. Nobody knows which model is strongest for which task, whether the answer is grounded, or what left the building. So AI stays in the drafting corner and out of decisions that carry cost.

An engineering decision made on an unverified answer is not a productivity gain. It is a liability with a delivery date.

03
How we solve it

One gateway. The right intelligence for every task.

Engineering question or taskAsked once, in plain language
Yonoo intelligence layer
  1. Understand what kind of task this actually is
  2. Route to the model strongest at it, on benchmarks we maintain
  3. Convene a panel when the question carries risk — several models answer, Yonoo compares
  4. Activate the agent or workflow the task needs
A more reliable, traceable answerWith the route it took visible

One secure gateway

Many models through one enterprise environment, instead of many tools and many subscriptions.

Right model per task

Models differ in what they are good at. Yonoo picks, so the user does not have to.

Lower decision risk

On high-stakes questions a panel answers and disagreement is surfaced, not hidden.

Works in your context

Answers can be grounded in the documents the company already has.

04
Core mechanism

When one model is not enough, we use several.

Several models answer independentlyChosen for the task, not fixed
Yonoo compares

Where they agree, confidence is high. Where they disagree, that disagreement is the finding.

One synthesised answerStronger than any single model produced
Running today

Multi-model panels in production

Users already run several models on one question and receive a combined answer. This is live for every paying tier and is the capability partners are integrating.

Next build

Anchoring every claim to a source

Tying each statement back to the exact passage in a company document. This is what turns a good answer into evidence, and it is the single largest item the raise funds.

05
Why our approach is different

We do not sell a model. We decide between them.

Independent by design

No model to defend

Every model vendor has to argue theirs is best. We have no such incentive, so we can be honest about which one wins a given task — and when a better model ships, we adopt it instead of rebuilding around it.

The recipe is the asset

Routing intelligence, continuously maintained

Which model for which task, when a panel is worth the cost, how the answers combine. This is maintained internally against our own benchmarks and is not exposed to customers or partners.

Domain, not theory

Built by someone who lived the problem

Our engineering founder came from vehicle and aerospace programmes. The workflow we are automating is the one she ran by hand.

Above the stack, not against it

Infrastructure partners, not competitors

Secure environments, private deployment and connectivity come from infrastructure partners. We stay in the intelligence layer. That keeps us light and makes us easy to embed.

06
Who we solve it for

A horizontal product entering through one vertical.

Beachhead — engineering

Automotive, aerospace, energy and industrial

Chosen because the pain is sharpest and measurable: decisions are expensive, evidence is mandatory, and the documents are already structured. It is also where our founder is credible on first contact.

How they cope today: separate subscriptions per team, an internal wiki nobody trusts, and asking the one colleague who remembers why the decision was made.

Then outward

The layer is not industry-specific

Routing, panels and governance are the same problem in legal, finance and public sector. Engineering is the wedge, not the ceiling — and it is the hardest room to win, which is why we started there.

Buyer: the engineering or digital lead who is being asked to put AI to work without putting the programme at risk.

07
Where we were, where we are

Nine months from problem to partner integration.

Jan 2026

Problem identified

Too many models, no intelligent way to pick and trust one.

Feb — Apr

Product built and live

First version shipped and validated with real users.

Apr — Jul

Organic growth

Reached roughly 1,700 registered accounts with no paid acquisition.

Jul — Aug

Partner integration

Partner API designed and built for an enterprise AI platform.

Sep 2026

Pilot live

Partner engineers testing the live API against their own platform.

Now

Demand ahead of capacity

The bottleneck is engineering throughput, not interest.

1,698
Registered accounts
~3,700
Incl. anonymous trial users
1
Partner pilot live on our API
3
Enterprise conversations open
Where we stand

A product in daily use, roughly 1,700 accounts grown without paid acquisition, a partner testing our live API and enterprise conversations with Bugatti Rimac, FEV and KONČAR. Enterprise agreements are still in discussion and consumer subscriptions are early. We are raising to convert demand we can already see, not to go looking for it.

08
Revenue model

One core, sold two ways.

B2B — direct enterprise

We own the customer

Annual platform license
Access to the intelligence layer per business unit or deployment
Usage
Routing and orchestration above the included allowance
Private or on-premise
Priced at a premium
Integration
One-off, only where genuinely required
B2B2B — platform partner

The partner owns the customer

Yonoo embedded
Our routing and orchestration inside their enterprise platform
Paid on usage
Per routed request or per active user, on the agreed commercial structure
Distribution without headcount
Their sales team, their support, their end-customer relationship
Status
Commercial structure agreed with our first platform partner; pilot running

License the intelligence, not the customisation. Build once, let the customer configure their own models, data and policies, and keep the layer that decides.

09
Unit economics Draft — under validation

What a unit earns, and what it costs.

Smart Router
LineValue
Revenue, metered€4 / 1,000 requests
Revenue, per seat€2.50 / user / mo
Routing cost~€0.10 / 1,000
Answer generationNot yet included
Gross marginTo be confirmed
Orchestration
LineValue
Revenue, per seat€6 / user / mo
Included60 runs
AI cost, fully utilised€1.80 – €3.60
Gross profit€2.40 – €4.20
Gross margin40% – 70%
Known issue, being fixed

Metered orchestration at €35 / 1,000 runs sits inside an estimated cost range of €30–€60 per 1,000. It is loss-making at the top of that range and is being repriced with usage controls before it goes to any customer.

Why margin expands

Year one carries setup and integration. The connectors, workflows and configuration are then reusable, so each further deployment costs less to serve while the licence revenue recurs. The unit is a deployment, not a seat.

10
Where we are going

Build once. Configure by customer. License the layer.

01

Core intelligence layer

Understand the task, select the model, route, orchestrate, compare.

02

Customer-configurable

Customers connect their own models, data and tools, and set their own policies.

03

Plug and play licensing

Annual enterprise licence plus usage, with private and on-premise options.

04

Ecosystem scale

Embedded across enterprise AI stacks through partners and platform licensing.

The Yonoo intelligence layer stays constant across all four stages. Only what surrounds it changes.

We do not want to monetise access to another AI interface. We want to monetise the intelligence that makes an enterprise AI stack work better.

11
Team

One founder knows the problem. The other has sold the answer before.

Korina Mlinarević

CEO & co-founder · engineering and product

Worked across automotive, aerospace and defence at Rimac, Lilium and Airbus Defence & Space. She has seen what an unreliable engineering decision costs, and she knows how these organisations actually evaluate and adopt new technology.

  • Engineering domain expertise
  • Product strategy
  • Enterprise development

Casper van het Hof

Co-founder · commercial and AI product

C-level commercial leadership before Yonoo, then identified an AI opportunity, built the product and exited it. He has already turned an AI idea into realised commercial value once.

  • C-level commercial leadership
  • Built and exited an AI product
  • Entrepreneurial execution
And what we bought instead of building

For secure enterprise infrastructure, private deployment and connectivity we work with an established enterprise AI platform partner rather than hiring that team. It keeps us focused on the layer we actually own.

12
The ask

The bottleneck is build capacity, not demand.

Engineering

Source anchoring

Tie every claim to the exact passage in a customer document. This unlocks requirements review, change-impact analysis and compliance evidence — none of which we can sell before it exists.

Enterprise readiness

Controls and deployment

Single sign-on, roles, audit trail and private deployment: the checklist that decides whether an engineering enterprise can buy at all.

Commercial

Convert the pipeline

Turn the open enterprise conversations and the live partner pilot into signed, referenceable deployments.

Raising €500k pre-seed to reach our first paid enterprise pilots in Q4 2026 and Q1 2027.

casper@yonoo.ai · korina@yonoo.ai · yonoo.ai