Feasibility · proof of concept · MVP
Evidence before the big budget
Feasibility, proof of concept and MVP work in a fixed number of weeks, at a starting price you can read before you call us. Fourteen U2 brands were launched from this ramp, and the ramp is now open to startups and investors, public institutions and large enterprises.
A product of U2 AI Studio Teknoloji A.S. The figures on this page are published starting prices for the scope described beside them. They are not an offer, and the binding figure is the one in the signed agreement.
Where this starts
Everyone wants to start with AI. Almost nobody can say where.
The usual first move is a large programme with a large budget, approved before anyone has checked whether the data is actually usable, whether the problem suits AI at all, and who inside the organisation would run the thing once it worked. Those questions are cheap to answer early and expensive to answer late. In a press release dated 29 July 2024 Gartner predicted that at least 30 percent of generative AI projects would be abandoned after proof of concept by the end of 2025, citing poor data quality, inadequate risk controls, rising costs and unclear business value. We quote that forecast as the reason the question is worth asking early. It is a 2024 prediction about 2025, not a measured outcome and not a result of ours. Every reason on its list is something a short, scoped engagement can surface before the budget is committed.
Do not open a programme. Ask one question, in a fixed number of weeks, for a price you can read before you call us.
Packages and definitions
Three priced packages, and the terms that come with them.
Fixed scope, fixed calendar, published starting price. The figures are stated in US dollars as a reference; where the client is resident in Turkiye the contract itself is concluded in Turkish lira. What we commit to is the calendar, the scope and the evidence we hand over. What that evidence turns out to say is not ours to promise.
Discovery
2 weeks · from $5,000
- For the organisation that has a shortlist of AI ideas and no cheap way to tell which one survives contact with its own data. Fixed scope, fixed calendar.
- A written feasibility answer on three axes: is the data ready, is the problem a fit for AI, and can your organisation operate the result. No code is produced at this stage.
- Output is a scoped brief plus a plainly written go, change or stop recommendation, tested against kill criteria we agree on day one.
Proof
6 weeks · from $18,000
- Six weeks to an answer, not six weeks to a guarantee. We build the part that could fail, and nothing else.
- A proof of concept answers one question: does the hard part work on your data, in your environment. We build the risky part only. It is not a product and it is not ready for production.
- You receive the build, the evidence log, a risk view structured on the four functions of the voluntary NIST AI RMF 1.0, and the preconditions that would have to be met before any production rollout.
MVP
12 weeks · from $45,000
- A working product placed in front of a first real user group. This is the step where a validated concept becomes something people can actually use.
- Scope covers architecture, security, data handling, documentation and handover, not only the model. What blocks the step from proof to production is usually governance, operations and team capability rather than the model itself.
- We commit to the calendar, the scope and the evidence. Adoption is something we test with you, never something we promise in advance.
What these words mean
Term discipline
- Feasibility study: can this be built at all, and at what cost. A written answer, no code. Proof of concept: does the technical approach work in practice on your data. A build of the risky part only, not a product.
- Prototype: will people use it. A realistic front end that answers a usability question and is then thrown away. Pilot: a limited, time-boxed deployment of a working build with real users in the real environment, run to test operations rather than to test adoption.
- MVP: a working product released to a first user group to test whether it is adopted. Five different things, five different budgets, five different risks. Three of them are the packages above. Prototype and pilot are scoped and quoted separately, because their price depends on your environment rather than on ours.
Data, IP and exit
Contract terms
- Controller and processor roles are fixed in writing before work starts. In the usual setup you are the data controller and U2 AI Studio is the processor acting on your instructions. Article 12 of Law No. 6698 makes the controller jointly responsible with the processor for data security, which is why the arrangement goes on paper before the first record moves.
- We apply data minimisation and prefer anonymised, masked or synthetic data wherever the question can still be answered with it. Your data is not used to train our models and not used to build our reusable components. Deletion or return of your data after the engagement is written into the contract, not left to goodwill.
- Background IP, meaning the frameworks, tools and reusable components each side brings in, stays with its owner and is licensed to you for use. For foreground IP the engagement contract sets out what transfers; because Turkish copyright law does not allow rights in a work that does not yet exist to be disposed of in advance, the assignment is signed as a separate written instrument once the work exists, listing each economic right transferred.
The launch ramp
Where the brands came from
- Fourteen U2 sub-brands were launched from this ramp: Carbon, Legal, Portal, FinTech, HR, Agro, Energy, Audit, Innovation, Readiness, Defence, Quantum, Robotics and Visibility.
- We built the ramp for our own ideas first. It works for yours because it had to work for ours, with our own money and our own name on the product.
- Across U2 AI Studio as a whole, 40 or more projects have been completed. That is a company-wide figure covering every brand, not a count of proof of concept engagements.
How it works
From an idea to a decision, in four steps
1. Frame the question
We write down the single question the engagement has to answer, and we agree what would count as a failure, before anyone is emotionally or financially invested in the answer. Those kill criteria go into the contract, not just into a slide. We also fix the controller and processor roles and the data handling rules at this point.
2. Check the ground
Feasibility comes before building. We assess data readiness as a question of accuracy, completeness, discoverability and governance rather than mere possession, we test whether the problem actually suits an AI approach, and we look at whether your organisation can run the result once we leave. Microsoft's Cloud Adoption Framework AI scenario asks along similar lines.
3. Build only the risky part
We build the part that could fail, on your data, in an environment close to yours, and we keep an evidence log as we go. We record what worked, what did not, and what remains unknown. Under the EU AI Act, Article 2(8) keeps research, testing and development activity outside the Regulation's scope until a system is placed on the market. That exemption does not extend to testing in real world conditions, so a pilot with live data and real users is scoped separately.
4. Decision session: go, change or stop
Every engagement ends with a written recommendation and a session with your team. There are three possible outcomes and all three are valid results: go, change direction, or stop. A stop is not a failed engagement. It is the engagement doing its job and saving you the cost of the build that would have followed.
WHO THIS IS FOR
Five kinds of buyer, five different questions
Most engagements have the same shape: a fixed number of weeks, a figure agreed in advance, and a written recommendation at the end. What changes is the question we are asked to answer. Investor technical opinions are the exception; they run in days and are quoted case by case.
Large enterprises
You have a shortlist of AI ideas and no cheap way to tell which one survives contact with your own data. We take one idea, run it against your environment under a fixed scope, and hand back a written go, change or stop recommendation. What we commit to is the calendar and the evidence, not the conclusion.
Public institutions
We publish scope, duration and price openly so the administration can make its own assessment. Choosing and documenting the procurement route is the administration's decision, not ours. Two things we say before anything is signed. A single requirement must not be split into separate orders so that each falls under a limit. And under Article 11 of Law No. 4734 a supplier who has prepared the documents of a tender cannot bid in that tender, so Discovery or Proof work that would shape a later technical specification may be the reason we are not eligible for it. Where a pilot runs on live data, hosting and data location are agreed before work starts, in line with the institution's own classification duties under the Presidential Information and Communication Security Guide.
Startups
Six weeks to an answer beats six months of building the wrong thing. We isolate the single riskiest technical assumption in your product, build only that part, and test it against criteria we agree with you in week one, before anyone is emotionally invested in the outcome.
Investors
A technical opinion for investors, not an audit. We look at architecture and scalability, code quality and delivery process, key person risk, security, open source licensing, and data and AI exposure. This work is measured in days rather than weeks and is quoted case by case, so it does not sit inside the three packages. U2 AI Studio is not an independent auditor, this is not an assurance engagement, and it is not investment advice within the meaning of Article 37 of Capital Markets Law No. 6362. Our findings are an engineering view formed on the material the company chose to show us, and the investment decision stays entirely yours.
U2 internal teams
We built this for our own ideas first, and internal proposals go through the same gate an external client gets: the same weeks, the same kill criteria, the same written recommendation. When an internal proof clears that gate, this is also where it gets an owner, a production checklist and the governance, operations and support questions answered before anything carries a brand name. It works for your idea because it had to work for ours.
WHY US
Four things we will put in writing
Fourteen brands were born here
Fourteen U2 sub-brands went through it: Carbon, Legal, Portal, FinTech, HR, Agro, Energy, Audit, Innovation, Readiness, Defence, Quantum, Robotics, Visibility. Across U2 AI Studio as a whole we have completed more than 40 projects. That is the company wide figure, not a proof of concept only count, and we make no separate performance claim for any individual brand.
Stop is one of the answers
Every engagement ends with a go, change or stop recommendation, written plainly. One possible outcome is: do not build this. That outcome is a result, not a failure, and it is the cheapest one on the list. The kill criteria are agreed in week one, before anyone is invested in the answer, and the test is held to those criteria rather than to a softer version of them.
The weeks and the starting prices are on this page
Discovery is 2 weeks from $5,000. Proof is 6 weeks from $18,000. MVP is 12 weeks from $45,000. These are starting figures in US dollars, excluding VAT, for the scope agreed in writing. They are not an offer: the binding figure is the one in the signed agreement, and it starts from these. No discounts, no campaigns, no limited time offers. What we commit to is the calendar and the evidence. What the evidence says is not ours to promise.
Five words that are not synonyms
A feasibility study answers whether a thing can be built at all, on paper, with no code. A proof of concept answers whether the hard part works on your data in your environment, building only the risky part and never a product. A prototype answers whether people would use it and is then thrown away. A pilot is a limited, time-boxed deployment of a working build with real users, run to test operations. An MVP is a working product released to a first user group to test adoption. Five separate kinds of work, so we scope and price them separately.
START HERE
Bring us the hard problem
Tell us the one assumption that would sink the project if it turned out to be false. That sentence is enough for a first call. We will tell you which package fits, or that none of them does.
Thanks. We'll get back to you within one business day.
FAQ
Frequently asked questions
What is included in the price, and when does it change?
Each package states a starting price in US dollars, a fixed number of calendar weeks, and a written deliverable. Discovery is 2 weeks from $5,000, Proof is 6 weeks from $18,000, MVP is 12 weeks from $45,000. Figures exclude VAT, and stamp duty on the contract is borne as the contract sets out. Where the client is resident in Turkiye the contract is concluded in Turkish lira; the dollar figure is the reference we quote from. The figure moves when the scope moves: more data sources, more environments, on-site work, extra integrations, or a longer calendar. Nothing on this page is an offer, we re-quote in writing before any of that begins, and the signed agreement carries the binding number.
What happens if the proof of concept does not work?
That is a result, not a failure. In week one, before anyone is invested in the answer, we agree the kill criteria with you in writing and we hold the test to those criteria rather than to a softer version of them. Every engagement ends with a plainly written go, change or stop recommendation. A stop is the cheapest outcome you can buy, because it saves the build that would have followed. We commit to the calendar and the evidence, never to what the evidence will say.
What happens to our data?
We settle roles before the first week. In the usual arrangement you are the data controller and U2 AI Studio is the data processor acting on your written instructions. Article 12 of Law No. 6698 makes the controller jointly responsible with the processor for the security of the data, and the Board's guidance on artificial intelligence asks for the roles to be determined at the outset, so we record the arrangement in writing before anything moves. We apply data minimisation and prefer anonymised, masked or synthetic data wherever the question allows it. Your data is not used to train our models or to build our reusable components. What is deleted, what is returned and on what date is written into the contract.
Who owns the intellectual property in the output?
We separate background and foreground intellectual property in the contract. Background is what each side already owned before the work started: our frameworks, tools, libraries and reusable components stay with us, and you receive a licence to use them in the delivered work. Foreground is what is created during the engagement: the source code, configuration and written report produced specifically for you. Turkish copyright law does not allow rights in a work that does not yet exist to be disposed of in advance, and it requires each economic right transferred to be stated separately, so the engagement contract records the undertaking and the assignment itself is signed as a separate written instrument once the work exists.
We are a public institution. How do we buy this?
The procurement method is the administration's decision, not ours, and we have no authority over it. What we do is publish scope and price openly so your procurement team can make its own assessment against its own budget line and its own legal basis. We will not advise you on which procedure or exemption to invoke, and we will never propose splitting a scope into smaller orders to fit under a limit. Article 5 of Law No. 4734 forbids dividing a procurement in order to fall below the threshold values, and the section of the Public Procurement General Communique dealing with Article 22 forbids dividing a single requirement so that the parts stay within the direct procurement limits. One more thing worth knowing before you buy: under Article 11 of Law No. 4734, those who prepare the tender documents of a work cannot take part in the tender for it, so if a Discovery or Proof report would shape the technical specification of a later tender, that may be the reason we are not eligible to bid in it. We would rather say so now than after you have paid us. Where the three packages answer three genuinely separate needs they are separate procurements; where they are stages of one requirement, that is one requirement and should be procured as one.
What is the difference between a proof of concept and an MVP?
They are different things and we scope them differently. A feasibility study answers whether something can be built at all, on paper, with no code. A proof of concept answers one technical question: does the hard part work on your data, in your environment. It is a build of the risky part only and it is not a product. A prototype answers a usability question and is then thrown away. A pilot is a limited, time-boxed deployment of a working build with real users, run to test operations. An MVP is a working product released to a first user group to test adoption. Our Proof package is a proof of concept. It will not arrive production ready, and we will not describe it that way. Prototype and pilot work is quoted separately, because its price depends on your environment.
Do you carry out technical assessments for investors?
Yes, as an engineering opinion, and it is not an audit. U2 AI Studio is not an independent auditor, this is not an assurance engagement within the meaning of Decree Law No. 660, and it is not investment advice within the meaning of Article 37 of Capital Markets Law No. 6362. We look at architecture and scalability, code quality and delivery process, team and key person risk, security, intellectual property and open source licensing, data and AI exposure, and the credibility of the roadmap. We tell you where the engineering risk sits and what removing it would cost. The opinion is formed in a fixed number of days on the material the company chose to show us, it is quoted case by case rather than from the packages above, and it is an input to your decision, never a substitute for it and never a recommendation to invest.
Did the 14 brands really come off this ramp?
Yes. Carbon, Legal, Portal, FinTech, HR, Agro, Energy, Audit, Innovation, Readiness, Defence, Quantum, Robotics and Visibility were all framed, tested and launched here. That is the evidence we offer: not a case study written after the fact, but the fact that we run our own ideas through the same gates, with our own money and our own name on the result. Across U2 AI Studio as a whole we have completed 40+ projects, a company-wide figure rather than a proof of concept count.