Researchers, PhDs and postdocs
You have a research result that works, and you want to know whether it can become a product. Nine products have already come out of our own research. We know how to take a result from the lab to the market.
From your idea to an AI-native product for the market.
Call4Ideas is for researchers, PhDs and postdocs, industry experts and anyone who has studied a problem in depth through their work or studies. You have in mind a method, a process or a technology that lets AI change how organisations, professionals or financial institutions work.
Who this call is for
We are looking for people who know a problem better than anyone else. You have studied it or lived it for years, and you have a clear idea of how AI can solve it. We work out the rest together.
Industry experts
You have known a sector for years and you see a process that AI can change. You know who decides, what it costs and why today's suppliers do not solve it. You need someone to build it with you.
Managers and operators
You have run a function in a company and you know the real numbers. You have the right contacts in the sector. You want a partner who builds the product and shares the risk with you.
Obsessed with a problem
You have no specific title, but you have studied a problem for years. These people often bring the best ideas.
The areas where we are looking for ideas
We assess ideas on any problem. Our research and our projects focus on these six areas. That is where we can help you most.
AI agents for banks, insurers and professional firms
We are looking for ideas for agents that handle reviews, checks and paperwork in banks, insurers, law firms and accountancy practices. We are also interested in defences against fraud and against attacks that use AI.
Propose an idea in this areaAlready have a company in this area? Apply to one of these calls:
Some problems we are thinking about
- A credit review still takes days of manual work on accounts, company records and documents.
- At an insurer a claim passes through many hands before anyone decides on it.
- Law firms and accountants spend many hours on repetitive work. An agent can prepare it and leave the review to the professional.
- Fraudsters already use AI against payments and identities. Many bank controls do not use it yet.
The infrastructure that makes agents reliable
Many agents work in a demo and stall when they meet a real process. We are looking for ideas on how to design, coordinate, evaluate and secure agents that work with real data and real rules.
Propose an idea in this areaSome problems we are thinking about
- Several agents working on the same process have to coordinate without a person checking every step.
- A company needs to measure how often an agent gets things wrong before it hands the agent a job.
- An agent that acts on company systems needs protection from malicious instructions and from permissions that are too broad.
- Every call to a model has a cost. At high volumes the architecture has to stay affordable.
Organisational memory and knowledge models
A company's decisions are scattered across emails, meetings and documents. Agents cannot find them and people forget them. We are looking for ideas on long-term memory, domain ontologies and knowledge graphs that record what was decided and why.
Propose an idea in this areaSome problems we are thinking about
- An agent needs to know who took a decision and which version holds today.
- When two documents contradict each other, someone has to notice before an agent uses them.
- Building the ontology of an industry takes experts months. AI can draft one for the experts to correct.
- When an expert leaves a company, often nobody knows any more why a process works the way it does.
Analysis and decisions on company data
Companies hold a lot of data and few people who can query it. We are looking for ideas for agents that analyse company data, answer questions in plain language and show where every number comes from.
Propose an idea in this areaSome problems we are thinking about
- A manager waits days for a report that an agent can prepare in a few minutes.
- Two departments calculate the same metric in different ways and nobody knows which figure is right.
- Many companies still forecast sales and cash by hand in spreadsheets.
- Whoever decides must be able to trace a figure produced by AI back to its data.
AI for factories and the physical world
We are looking for ideas that bring AI into production. We are thinking of quality control, maintenance, the supply chain and robots that work alongside people.
Propose an idea in this areaAlready have a company in this area? Apply to one of these calls:
Some problems we are thinking about
- On many lines quality control still depends on the eye of an experienced operator.
- An unplanned machine stoppage halts production for hours. The signals that predict it are often already in the data.
- Industrial robots have to be reprogrammed for every product change. For small batches the cost does not pay off.
- Maintenance procedures live in the heads of the most experienced technicians and nobody has ever written them down.
AI for chemistry, electronics, materials and physics
Many discoveries start in a lab and never reach the market. We are looking for ideas that use AI in chemical and electronic engineering, materials science, biotech and pharmaceutical research, and physics applied to finance.
Propose an idea in this areaAlready have a company in this area? Apply to one of these calls:
Some problems we are thinking about
- Finding a new material or a new formulation still takes hundreds of lab trials.
- Designing and verifying an electronic circuit takes weeks of simulations and reviews.
- No researcher can read the whole literature of their field. An agent can read it, propose hypotheses and design experiments.
- Many risk models used in finance underestimate extreme events. Statistical physics has tools that describe them better.
Does your idea fit none of these areas? Send it anyway. We assess it with the same method.
How we select ideas
We assess every idea with the Innovation Potential Score. The score looks at the problem from five angles: labour mass, innovation void, automation leverage, crossable barriers and right to win.
IPSInnovation Potential ScoreIs it worth it?
If your proposal answers these five questions well, we can assess it accurately.
L
Labour mass
How many human hours sit inside the process today?
- What raises it
- Many people do the same work, and you can say how many they are, how often they do it and how many hours it takes.
- What lowers it
- The problem is annoying but rare. Or one person deals with it a couple of times a month.
V
Innovation void
How uncovered is the problem?
- What raises it
- The people who live with the problem make do with spreadsheets and email, because no supplier solves it.
- What lowers it
- Many companies already solve it well. A product that is only slightly better is not enough.
A
Automation leverage
How much of that work can AI genuinely take on today?
- What raises it
- You can say what has changed recently. For example, a model now reads documents it could not read before, new data exists or a rule has changed.
- What lowers it
- It would have worked five years ago too, or it needs a capability no model has.
B
Crossable barriers
What keeps competitors out, and can we get through?
- What raises it
- Real barriers exist and you know how to get past them. They can be regulation, integration with old systems, data nobody shares or multi-year contracts. Whoever gets past them first stays in the market.
- What lowers it
- There are no barriers, so anyone can copy. Or there is a barrier nobody can cross.
R
Right to win
Why you specifically, and not anyone reading the same report?
- What raises it
- You have access to customers, your own data, years of experience in the sector or a research result of your own.
- What lowers it
- You bring only enthusiasm, or a market analysis anyone could do with the same sources.
Two axes are decisive: automation leverage and right to win. If either is missing, the proposal does not pass. This holds even when the problem is large. If there is a barrier nobody can cross, we tell you straight away.
What we reply
We reply to everyone. Based on our analysis, we build a path that fits each idea we select.
Acceleration and company building
We build the company with you and take it towards the market.
Industrial research
If the idea is mainly technological, we start with an industrial research path with our team.
We use AI tools that help us analyse and validate ideas. But every decision is taken by a person on our team. Our decision-making is AI-powered, but it is always human-made.
Prepare before you send
Want to arrive better prepared? The assistant in the form does the short version of this work. For the long version, paste the prompt into ChatGPT, Claude or whatever you use, add your idea however messy, and let it question you.
You are a sceptical venture-building analyst. Your job is not to encourage me: it is to work out whether my idea holds, and to help me describe it so it can actually be assessed. I am about to propose an idea to KVA, a venture studio that builds AI-native companies and weighs every idea on five axes: 1. LABOUR MASS — how many human hours sit inside the process today. Little work, little value to free. 2. INNOVATION VOID — how uncovered the problem is. If ten companies already solve it well, an eleventh is not needed. 3. AUTOMATION LEVERAGE — how much of that work AI can genuinely take on today, not in theory. 4. BARRIERS — what keeps competitors out, and whether they can be crossed. 5. RIGHT TO WIN — why me specifically, and not anyone else reading the same market report. It is one of the two knockout axes, along with automation leverage: without it, the proposal does not pass. Here is my idea, messy as it is: <<< PASTE YOUR IDEA HERE >>> Work in this order, without skipping a phase. PHASE 1 — Interrogate me. Ask me at most eight questions, one at a time, starting with the ones you cannot answer yourself from what I gave you. Prioritise labour mass, who pays, and right to win. Do not move to phase 2 until I have answered. PHASE 2 — Assess me, no discounts. For each of the five axes tell me three things: what in my idea raises it, what lowers it, and what I should go and find out to answer better. If an axis is weak, say so plainly instead of softening it. If automation leverage or right to win are weak, tell me it does not pass as it stands, and why. PHASE 3 — Draft the answers. Write a draft for each of these fields, in my words, with no pitch jargon and no promotional adjectives. Use the concrete numbers where I gave them, and write "to be verified" where I did not. - The idea in one line (max 120 characters, descriptive, not a slogan) - The industry I want to change (my words, not a category) - The work I want to change (who does it, how many people, how many hours, what goes wrong) - Why now, and why with AI (what changed that was not there five years ago) - Who pays (the role that signs, not the company) - What I bring (customer access, data, years inside the industry, a research result of my own) - What I am looking for from a venture studio (AI research, product, go-to-market, capital, team, company setup) Then, only if I genuinely know them, two optional ones: how it is solved today, and what makes it hard to enter. If I do not know, leave them out: an invented answer weighs worse than a blank one. PHASE 4 — The last thing. Tell me the riskiest assumption in the whole idea, and the cheapest test that could prove it wrong.
What we offer innovators
When you join us, you work with our team from day one. This is what we make available to innovators.
Research
Our research team works with you. Our AI researchers publish papers and benchmarks. They help you turn the idea into proprietary technology that can compete.
Development and an enterprise-grade product
We help you build the product, not just design it. We work with banks and insurers every day and know what they ask of a supplier. So we get your product ready for their compliance, security and data residency requirements. With our AI sovereignty pipelines your AI runs on open models and European cloud, and the data stays in Europe.
Strategy, finance and investor readiness
You work with our strategists, product owners and investor strategists. They help you sharpen your strategy and lower the risk of failure that comes with every new idea. We also work with you on finance and management control, so that the startup's numbers make sense. Finally, we build the equity strategy with you and prepare you to talk to investors.
Tests with large companies
We advise large companies. When the product is ready and the chance arises, we test it with these enterprise clients.
Anthropic Ambassador
We are Anthropic Ambassadors and take part in the Claude for Organizations programme. All our venture builders are Claude-certified. Founders who work with us can get certified too.
Full-stack AI
We partner with the main AI vendors and have good agreements with them. Sometimes we supply the Claude, OpenAI and Grok subscriptions ourselves. Other times we help the startup develop its own open models and AI stack. Through our partners, such as Nebius, AWS and Google Cloud, startups can also buy compute.
When we decide to work with an innovator on their idea, we make our structure available to speed up its growth.
Applying takes ten minutes
You can tell us the idea however you like, in writing or by voice. You can also stop and come back later. The draft stays saved on your device until you send it.
The form asks you to answer these questions:
- Which work do you want to change, and how many people do it today?
- Why can AI solve the problem today, when it could not five years ago?
- Who would pay for the solution? Name the role of the person who decides the purchase.
- What do you bring that nobody else has?
- What do you need from us to get started?
We reply to everyone within 12 working days. When the answer is no, we explain why.
Propose your ideaThe idea stays yours and we always credit it to you. We do not pass it on to third parties and we do not use it to build without you. If it overlaps with a project we already run, we tell you straight away.
We do not keep your account of the idea. It only serves to prepare the draft, which you read and confirm.
Already have a company and an open round?
Then apply to Call4Ventures. There are seven themed calls and a general one, with a commitment of €50,000 to €100,000.