aws Digitalization | AI Adoption FAQ

AI-Adoption is the funding of innovative trustworthy AI projects in Austria with a focus on preparing for regulations, standards, norms and certifications.

In the AI-Adoption: Green, funding is also provided for projects that adress climate and environmental protection (mitigation) as well as adaptation to the consequences of climate change.

These FAQs will help you with questions before and during the application process as well as with the funding process. The FAQs provide a basic overview without any claim for completeness and are being updated for you on an ongoing basis. Only the guideline specifications for aws Digitalisation | AI-Adoption is legally binding. The FAQs will be updated if necessary.

If you know more tools or resources and would like to share them with the AI community as part of these FAQs, we look forward to your input!

Version: 01/010/2023

1. Overview

Are the FAQs legally binding?

No, they are not legally binding. They merely serve to summarise frequently asked questions.

2. Invitation for application

Who can submit a project for aws Digitalisation | AI-Adoption or AI-Adoption: Green?

Pre-foundation projects and existing companies (small and medium-sized enterprises/SMEs and large companies) based in Austria are eligible to apply.

Can research institutions, universities, associations or consortia apply?

No.

I am a company with a registered office outside of Austria, but I would like to implement an AI project at my place of business in Austria. Can I submit?

Companies with a permanent establishment in Austria are eligible to apply. In this case, the aws may impose separate requirements for proof of implementation of the project in Austria.

Can several companies submit an application together?

No, only one company can apply for funding at a time. Application as a consortium is not possible. 

What is an “innovative company”?

If you develop something new compared to the current state of technology and there is a corresponding risk associated with it. Or if you have invested at least 10 percent of your operating expenses in research and development in one of the last three years. The exact definition can be found in the guideline specifications. You can find this in the AI-Adoption download area on the website.

What is meant by “existing technical risks”?

That the concrete implementation of the project can technically fail. It is necessary to describe specifically what these risks are and how they are dealt with.

Should there be technical risks in the proposed project?

Yes. Innovation is risky. The funding is intended to help alleviate some of the risk. Projects without technical risk cannot be supported.

What is experimental development?

Experimental development is often characterised by the development of prototypes, demonstration or pilot plants. The focus is on the creation of new insights and skills. This does not include routine developments of existing products and services. The exact definition can be found in the guideline specifications.

What is a “young company”?

If the initial entry in the commercial register was no more than 5 years ago. The exact definition can be found under the following link in the legal basis, General Block Exemption Regulation (AGVO), Article 22.2

What are “small companies”?

Companies with fewer than 50 employees (full-time equivalent) and an annual turnover or balance sheet total of less than EUR 10 million. See also the EU’s definition of an SME (see Annex 1 of the General Block Exemption Regulation).

You can also find detailed information about small- and mediums-sized enterprises in the aws SME factsheet.

Can de-minimis and General Block Exemption Regulation funding be combined?

No. Before submitting your application for funding, you decide whether you want to be funded under de-minimis or General Block Exemption Regulation.

If a de-minimis framework is no longer available, can I be funded on the basis of the General Block Exemption Regulation?

Yes, if there is an unused General Block Exemption Regulation framework.

How do I find out if there is an unused de-minimis framework?

To do this, you need to sum up the de-minimis funding promised in your company or group of companies in the last three tax years (e.g. 2020-2022). You can find more details in the aws de minimis fact sheet.

Where do I enter the information on the EU legal basis which I would like to apply for in the aws Funding Manager?

After the point “requested funding in section 5 of the application for funding, select the desired EU legal basis: ”de-minimis”, “General Block Exemption Regulation Art. 22” or “General Block Exemption Regulation Art. 25“.
 

Which industries or applications are explicitly excluded?

Fisheries and aquaculture, primary production of agricultural products, banking and other financing, insurance and real estate companies, local authorities and legal persons in which local authorities directly or indirectly hold more than 50% of the shares. Furthermore, non-civilian applications are excluded.

What is the maximum amount of third-party costs in the project?

A maximum of 49% of eligible project costs. Since the projects sought are self-created innovations, the majority of the services must also be provided by the company itself. Therefore, the core know-how for the project must also lie or be built up in the company.

Can the development of the core know-how of the project be carried out by third parties if the total third party costs remain below 49%?

No. The core know-how of the promoted innovation has to lie in the company or has to be built up in it. This enables the best possible development and forms the basis for successful market development.

Are projects with external research partners preferred?

No.

Can the costs of implementing trustworthy AI (standard selection and implementation, communication, training, data analysis and cleansing, …) be funded?

Yes.

Can the costs of preparing for or implementing the EU AI Act, standards, norms, certifications be funded?

Yes.

How should the milestones planned?

The funding will be paid in three tranches. The first and second tranches amount to 40% each and the third tranche is 20% of the funded amount. Please plan this accordingly in the template which can be found on the aws Website of AI-Adoption under Downloads.

Are overhead expenses eligible?

No. Flat-rate compensation of overhead expenses is not possible with this funding.

Is there a key for how personnel and non-personnel costs must be distributed?

No. The distribution results from the requirements of the respective project.

What is meant by “own contribution”?

At least 20 percent of the total eligible costs are to be contributed to the project “in cash”.

Can the equity also be contributed by fictitious company wages or salary waivers?

No. The own contribution must be demonstrably provided in money and used in the project.

What can the “own contribution” be used for?

For eligible costs in accordance with the guideline specifications.

Can investments already made before the application, such as a computer optimised for AI calculations, be fully funded or aliquot funded to the extent of depreciation over the project term?

No.

How should the funding be treated for tax purposes?

See the aws information sheet www.aws.at/fileadmin/user_upload/Downloads/Beihilfenrechtliche_Grundlagen/2023_Informationsblatt_Steuerliche_Behandlung_von_Foerderungen.pdf (German only).

Which documents are mandatory for an application for funding?

For everyone: 

• Business plan (according to the template which can be found on the aws Website of AI-Adoption under Downloads.)
• Integral planning (according to the template which can be found on the aws Website of AI-Adoption under Downloads.)
• Copies of identity documents of the applicant, the authorised signatories and the beneficial owners

For existing companies, additionally:

• Financial statements of the last three financial years
• Full-time equivalent employees of the company or group of companies.

Do the “Business Plan” and “Integral Planning” templates have to be used?

Yes, the templates are mandatory. This is necessary to make it possible to compare all projects. 

How many pages should the business plan have?

Maximum 20 pages for AI-Adoption and max. 25 for AI-Adoption: Green - each including cover page and any table of contents. This is also stated in the introduction of the mandatory template.

The maximum length of the 20-page business plan respectively 25-pages für AI-Adoption: Green is too short, what can I do?

Briefly and concisely discuss the headings of the template. Try to avoid redundancies in the text and use references if necessary.

The maximum length of the 20-page business plan respectively 25-pages für AI-Adoption: Green is too long, can you alternatively upload a paper or an existing pitch deck?

No. Use of the template is mandatory.
You are welcome to upload additional documents, but these are not part of the set of documents, which is mandatory for assessment by aws.

Does the business plan have to be exactly 20 pages respectively 25-pages für AI-Adoption: Green long?

No, if you can discuss all the headings of the template and describe your project in fewer pages, the business plan can also be shorter. The length of the business plan – apart from the formal maximum length of 20 pages for AI-Adoption and 25 for for AI-Adoption: Green – is not an evaluation criterion according to the guideline specifications.

Is a general pitch deck sufficient instead of the business plan and/or integrated planning?

No. Please use the mandatory templates; this is the only way to compare the different projects in a call procedure.

What should be the distribution of the pages in the business plan on the different topics?

This is up to you to decide. However, be sure to cover all the topics given by the headlines.

Can I upload the plan balance sheet and plan profit and loss statement as an attachment?

Yes. In this case, a summary must be included in the business plan under 9.1.3 a) and b) reference must be made to the additionally uploaded attachment.

I have an existing company/am a member of a group of companies. Do I have to upload the annual plan for my entire company?

This is up to the companies. For SMEs & large companies, either 1) 3-year plan for the company + presentation of turnover share of new project or 2) 3-year plan for the submitted project + turnover share in the company can be submitted.

Is it possible to submit documents after the deadline for submissions?

No.

Can questions in the application be answered in English or can documents be made available in English?

Yes, the submission documents and the presentation to the jury can be made in English. The aws is also happy to provide advice in English. However, the processing of the funding - from the application to the funding agreement to the compilation of the invoice and the final report - always takes place in German.

Is there a checklist for submitting the application?

Yes, you can find it in the Download section of the aws Website für AI-Adoption.

Can the funding be combined with aws Seed Financing Innovative Solutions?

Yes, generally speaking, that is if the project content is different.
No, if the same project content is to be funded.

Can the funding be combined with aws Seed Financing Deeptech?

Yes, generally speaking, that is if the project content is different.
No, if the same project content is to be funded.

Can the funding be combined with AI Start-up, AI Knowledge or AI Marketplace?

Yes, generally speaking, that is if the project content is different.
No, if the same project content is to be funded.

3. Artificial intelligence

Do the innovations have to have been internally developed using trustworthy AI?

Yes.

Do new AI technologies need to be developed internally?

No. The use of state-of-the-art technologies is basically sufficient. If the specific project requires further developments, this increases the degree of innovation.

Can my project use ChatGPT or another Large Language Model already available on the market?

Yes, but note that with a) many of these models there are a number of open legal, ethical and robustness issues (see e.g. https://crfm.stanford.edu/2023/06/15/eu-ai-act.html) and b) you as a company can be held liable for putting the product or service on the market.

How important is the use of AI in the proposed project?

Very important. aws is looking for projects that are made possible for the first time through the use of AI or where this brings significant improvements to the status quo of previous solutions.

Can a research project on “Implementing a Trustworthy AI” be submitted?

No, the funding is aimed exclusively at the entrepreneurial development and implementation of concrete products, services or business models based on trustworthy AI.

Can a research project on the topic of “Implementation or preparation for regulations, standards, norms and certifications in the field of trustworthy AI” be submitted?

No, the funding is exclusively aimed at entrepreneurial implementation or preparation for legal regulations, standards, norms and certifications.

We already use classic algorithms or an existing software solution and would like to supplement them with an AI system. Should we submit the application?

Yes, if the AI project can solve problems for the first time and/or lay the foundation for new business models.

No, if only minor improvements are possible and/or there are no significant new business opportunities.

We already have an AI system in place and want to exclusively implement or prepare for regulations, standards, norms and certifications in the field of trustworthy artificial intelligence. Technical development is no longer happening, should we submit the application?

No. The experimental character of the project must continue to apply. The implementation or preparation for regulations, standards, norms and certifications are a necessary focus of the project, but not sufficient by themselves.

We already have an AI system in place and only want to do the creation of an impact chain, ESG report or sustainability report. Technical development is no longer happening, should we submit?

No. The experimental character of the project must still be given. The implementation or preparation for regulations, standards, norms and certifications, impact chains, ESG or sustainability reports are not sufficient.

What is a data strategy?

A long-term concept on how to deal with data. Within the framework of a data strategy, e.g. principles and rules are established in dealing with data in order to pursue defined goals. 

What can a data strategy look like?

There are many ways to build a data strategy. Please choose an approach that suits your project. Examples of procedural models for this can be found, for example, under https://dmponline.dcc.ac.uk/,  Data drivers or Kathrin-Mathis, etc.

What is meant by a trustworthy AI?

Trustworthy AIs are lawful, ethical, and respect the following elements (page 17f):

1. Priority of human action and human oversight, e.g. safeguarding the right to self-determination
2. Technical robustness and safety, e.g. resistance to attacks and security breaches, recovery plan and general safety, precision, reliability and reproducibility
3. Protection of privacy and data quality management, e.g. respect for privacy, quality and integrity of data, and data access
4. Transparency, e.g. traceability, explainability and communication
5. Diversity, non-discrimination and fairness, e.g. avoiding unfair distortions, accessibility and universal design and stakeholder involvement
6. Social and environmental well-being, e.g. sustainability and environmental protection, social impact, society and democracy
7. Accountability, e.g. verifiability, minimisation and reporting of adverse effects, trade-offs and remedies.

What is meant by performance goals of the AI system?

When developing and training AI systems, it is necessary to test their performance. Metrics such as Precision, Recall, F1 and many more serve this purpose. Select the relevant ones for your AI system and its application, make them known as well as the target values that are planned.

Where can I find relevant benchmarks for AI systems?

See e.g. aiindex.stanford.edu/wp-content/uploads/2023/04/HAI_AI-Index-Report_2023.pdf or https://paperswithcode.com/sota.

There is no benchmark for the AI system used in the project, can I then omit the topic in the business plan or in the integral planning?

No. If you have not found any directly or indirectly relevant benchmarks despite an intensive search in the broad field of benchmarks, then define your own benchmark - with everything that belongs to it, such as a test data set, define target values that are required for a successful deployment, etc..

Are there any templates on how to make an AI system trustworthy?

In addition to the ALTAI checklist, there is also the draft of the VDE SPEC 90012 V1.0 (en) as well as the VDE press release; which offers a practical approach.

The Fraunhofer Institute for Intelligent Analysis and Information Systems (IAIS) also offers a “Guideline for the design of trustworthy artificial intelligence” at (https://www.iais.fraunhofer.de/de/forschung/kuenstliche-intelligenz/ki-pruefkatalog.html#KI-Pruefkatalog-kostenfrei-erhalten).

Why is it important that AI systems be trustworthy?

Because trustworthy systems will find greater acceptance among users. Without acceptance, possible opportunities of AI systems for companies and society cannot be realized.

What is meant by “technical and social robustness”?

All measures taken in the project must not cause any harm, from either a technical or a social point of view, and must only make positive economic and social contributions.

What is meant by “technical robustness and safety”?

This refers to the technical robustness and security of the AI system in accordance with the guidelines for trustworthy AI.
This does not include general IT security topics such as the use of the latest software versions, backup strategies or protection against internal or external hacking.

What is meant by “ethical principles and values”?

The use of AI requires more than just compliance with the laws. It is also important to consider how socially critical questions and challenges that were identified during or before the start of the project are dealt with. The principles laid down in the project for dealing with sensitive issues such as fairness, racism, discrimination, etc., are explained here.

For this purpose, we recommend referring to the EU guidelines (page 16f). The four ethical principles

1) Respect for human autonomy;

2) Damage prevention,

3) Fairness and

4) Explainability

are outlined in detail there. See also, e.g. “Ethik in KI und Robotik” [Ethics in AI and Robotics], Hanser Verlag, 2019.

Should open standards and interfaces be used in the projects?

Yes. The use of open standards and interfaces can make a significant contribution to trustworthiness, interoperability and scalability.

What is the EU AI Act?

The EU AI Act is intended to regulate the use of AI systems in the EU. The current draft can be found here.
An overview presentation by the Commission is also available and an overview of developments.

If the EU AI Act is not yet in force, why is it already useful to prepare for it as a company?

Since essential principles such as the risk classification of AI applications, human oversight, protection against discrimination, prohibition of certain AI applications and others have already become established as robust elements in the discussion. Accompanying the EU AI Act, the EU is already preparing further regulations such as the EU AI Liability Directive in order to comprehensively regulate the use of AI.
The timely preparation for this ensures planning security in companies and can significantly strengthen competitiveness.

What other regulations are relevant for AI?

Different regulations may apply to the respective areas of application, please check your relevant legal framework in individual cases. In the field of medical technology, for example, this can be the Medical Devices Regulation or the planned new EU Machinery Regulation.

It is also worth taking a look at the planned AI Product Liability Directive and the General Data Protection Regulation.

What is the “risk pyramid” of the EU AI Act?

In the current draft of the EU AI Act, AI systems are divided into four categories:
1) Unacceptable risk, Article 5, e.g. social scoring). These are forbidden.

2) High risk, Article 6.2 and Annex III of the AI Act, e.g. medical technology, human resources, transport, justice. These are allowed, but compliance with AI regulation up to ex-ante assessments is required. See in particular Annex IV to the AI Act.

3) Medium risk, Article 52, e.g. chatbots, AI-generated video content, emotion recognition. These are allowed, but there are information and transparency obligations.

4) Small or low risk, article 69, e.g. spam filters, video games, offline heating control. These are allowed without restriction.

What EU requirements are planned for AI high-risk systems?

These include, among other things: (as of 12/2022) compliance, establishment of a risk management system, creation of data and data governance processes, required technical documentation, retention obligations, transparency and information obligations towards the users of the AI systems, effective control and supervision by humans as well as accuracy, robustness and cybersecurity rules.

Where can I make an initial self-assessment of the risk level of my AI system?

By classifying your project according to the classification of the EU AI Act. There are also first online tools, such as the EU AI Act Compliance Check, but note that this is only intended as support and the current state of the EU AI Act must be taken into account.

Another help can be the Risk Classification Template of the Initiative for Applied Artificial Intelligence: 

What standards and norms are there for AI

For this, the platform Industrie 4.0 (Industry 4.0), in cooperation with aws, offers an overview, incl. a list of selected AI-related forums (https://plattformindustrie40.at/services/#vertrauen).

At the EU level, AI Watch offers a good overview in its publications:

The AI Standards HUB also provides a good starting point for research.

What is the difference between a norm and a standard?

Norms are regulations that are published by standardisation organisations such as ISO and go through the corresponding processes of these organisations.
Standards are regulations that typically have a harmonising effect and are also recognised in an industry, but have not gone through a defined norming process of an organisation that deals with norms.

What certifications are available for AI?

This area is in continuous development. Please observe current trends in your area. Enclosed you will find some already known certification initiatives:

Currently, among others, “Trust you AI” from the Know-Center and partners, “Trusted AI” from TÜV Austria in cooperation with JKU, IEEE with CertifAIEd™ are active. The IEEE Authorized Assessor registry can be found here.

In the project “Certified AI”, the Fraunhofer Institute in Germany also offers an AI test catalogue.

Why are trustworthy AI and the preparation for or implementation of regulations, standards, norms and certifications more important in the evaluation than innovation and growth?

Because it is the core of the program to show the possibilities of a trusting use of AI. Furthermore, by supporting the preparation for the upcoming EU AI Act, the first standards that are being developed or also certifications are to be made possible and thus the competitiveness of the companies is to be strengthened.

What is meant by “legality” of AI?

That all relevant framework conditions, such as laws, norms, etc., are complied with.

Why are the weightings between AI-Adoption and AI-Adoption: Green different?

In order to evaluate the focus on Green, separate criteria are required. These change the set of criteria and the weights.

What is meant by "legality" of the AI?

That all relevant framework conditions such as laws, standards, etc. are complied with.

How do I “explain” e.g. a decision by a deep learning system?

The concrete explanation of a decision, e.g. in machine learning or deep learning systems, is not possible and is the subject of current research.

Depending on the possible consequences of decisions of such AI systems, alternative explanatory approaches are needed. How extensive these should be can be made dependent on how “critical” the decisions are. See, for example, the Criticality Pyramid and a risk-adapted regulatory system for the use of algorithmic systems (S177f, German Data Ethics Commission Export Opinion, 2019), which can also be found in the current draft of the EU AI Act.

Based on this, it is possible to e.g. determine the framework conditions, system boundaries, (human) supervision, etc. under which decisions are made. Accordingly, these must be communicated transparently.

How to implement “trustworthy AI” in companies?

The Rating List for trustworthy AI offers practical checklists for this purpose.

For example, the white paper “Building the organisation for scaling AI” (www.appliedai.de; 19/02/2020) shows how to get from a basic AI strategy to organisational embedding.

Do funded projects need to use guidelines to achieve trustworthy AI?

Yes.

Which trusted AI guidelines can be used?

The selection is in the responsibility of the project, as long as the core elements for a trustworthy AI are fulfilled according to the guideline specifications.

Where can you find guidelines for trustworthy AI?

The Ethics guidelines for trustworthy AI provide guidelines with checklists for concrete implementation. 
Furthermore, there is a comprehensive collection of guidelines under this link. When selecting guidelines, pay attention to their practical applicability for your project.

Can you also use your own guidelines for trustworthy AI?

Yes, as long as they meet the core elements of trustworthy AI according to the guideline specifications. This must be set out accordingly in the business plan.

Is it enough to simply classify the project in accordance with the EU risk pyramid in order to meet the project focus of preparation for or implementation of AI regulation, standards, norms or certifications?

No. In addition to the first assignment and explaining the reasons for the assignment, the program also deals with the concrete development and implementation of the resulting consequences. This must be reflected in the description of the project in the business plan and in the concrete implementation in the integrated planning.

Is it sufficient, for example, to cite a relevant standard, a norm or a certification in order to meet the project focus of preparation for or implementation of AI regulation, standards, norms or certifications?

No. In addition to the identification and explanation of the reasons for the selection, the program is about the concrete implementation. This must be reflected in the description of the project in the business plan and in the concrete implementation in the integrated planning.

Where can I find examples of AI applications and who works in this environment?

The 19 projects funded by the aws call “Digitalisation of Trustworthy Artificial Intelligence (KIplus)” (www.aws.at/ki) offer a wide range of applications. You can find them under www.aws.at/service/cases/gefoerderte-projekte-auswahl/. There you select “aws Digitalisation – Artificial Intelligence” in the drop-down field “Programme”.

Where can I find technology providers on the subject of AI?

Visit the aws AI Marketplace www.awsconnect.at/KI-Marktplatz for a wide range of vendors.

Can Asimov’s “3 Laws of Robotics” be used as trusted AI guides?

No.

 4. AI-Adoption: Green

What is the EU Green Deal?

With the European Green Deal, the EU plans to make the transition to a modern, resource-efficient, and competitive economy that

• emits no net greenhouse gases by 2050,
• decouples its growth from resource use,
• leaves no one, neither people nor regions, behind.

See also the European Commission's strategy paper. 

What are the priorities of the EU Green Deal?

1. clean energy

2. sustainable mobility

3. climate protection

4. biodiversity

5. from farm to table

6. buildings and renovation

7. sustainable industry

8. elimination of pollution

What is meant by "Green"?

In addition to the requirements of AI-Adoption, AI-Adoption: Green requires that the projects:

• make a significant, predominantly direct, measurable positive contribution to achieving at least one priority of the EU Green Deal ("positive impact") and
• present this contribution qualitatively and quantitatively - including an extrapolation of the impact and
• present the relevant impact chain and a "do no significant harm" analysis and
• provide a "do no significant harm" analysis.

What is meant by impact?

A positive contribution to the implementation of the EU Green Deal, e.g. the CO2 reduction targets. The contribution must always be presented "net", i.e. minus any negative effects of the project on e.g. the CO2 reduction target.

What are examples of quantitative contributions to the EU Green Deal?

These can be e.g. CO2 equivalents saved (tonnes/year), noise reduction (x% db SPL), particulate matter reduction (x%), habitat created (ha), pesticide/herbicide savings (tonnes/year), material recycled (x%), resource savings (tonnes/year), volume of water/soil/air purified (in m3 or tonnes).

It is important that the contribution is measurable.

How can the emissions of a project be presented?

To do this, select the main emissions in your project, e.g. on the basis of the climate-impacting gases CO2, methane, nitrous oxide or fluorinated gases (https://www.umweltbundesamt.at/klima/treibhausgase).

What are examples of AI-Green projects?

As an example from the 1st AI-Adoption Call, see the DrSOLAR project by Air6 Systems, which implements AI-assisted solar cell inspection.

What distinguishes AI-Adoption (AI-A) projects from AI-Adoption: Green (AI-A: Green) projects?

AI-A: Green projects have as their main project focus to contribute to the EU Green Deal. This is not required for AI-A projects. All other requirements are identical.

Can I submit the same project in AI Adoption as well as AI Adoption: Green?

No.

If the project does not make a direct and measurable contribution to the EU Green Deal but can make an indirect contribution, e.g. through long-term behavioural changes, is the project eligible for funding?

No. Every use of AI requires a critical look at the direct and measurable impact, e.g. on the carbon footprint. There is a high savings potential here through optimisations, e.g. https://ai.googleblog.com/2022/02/good-news-about-carbon-footprint-of.html.

What is meant by a "predominantly direct" impact?

The positive impact is as direct as possible as a consequence of the implementation of the project and is typically closely linked to the business model. In addition, most of the impact is immediate.

The use of "AI" does not have a direct positive impact on the Green Deal, since electricity is consumed in the process. How can a positive impact still be achieved?

AI can be a "tool" that makes innovation possible for the first time and is usually part of a larger product or service. The positive effects of using this product or service must then lead to a positive net contribution to the Green Deal. To do this, the energy expenditure of the AI system - over the entire life cycle - must be subtracted from the positive impacts in order to calculate the net impact.

Can the contribution to the EU Green Deal also be represented by predominantly indirect effects that first occur via several intermediate steps, e.g. through changed consumer behaviour or mobility behaviour?

No. A predominantly direct positive and also measurable contribution of the project to the EU Green Deal is required for AI-Adoption: Green. Long-term indirect effects such as behavioural changes have great potential for climate protection but are not the focus of the call.

If I save fossil fuels through an innovative AI solution, how can I convert that into CO2?

For this you can convert the fuel savings to CO2 equivalents, use e.g. https://secure.umweltbundesamt.at/co2mon/co2mon.html.

What is a CO2 equivalent?

The CO2 equivalent describes how much a greenhouse gas contributes to the greenhouse effect.  (https://secure.umweltbundesamt.at/co2mon/co2mon.html)

Does the project have to be mapped in the EU taxonomy to be eligible to apply?

No. The taxonomy is a classification for sustainable economic activities such as zero carbon activities, transitional activities and enabling activities to achieve the European sustainability and ESG (Environmental, Social, Governance) goals. See also www.bmk.gv.at/green-finance/finanzen/eu-strategie/eu-taxonomie-vo.html. However, many innovative AI Green projects will also find an equivalent in the EU taxonomy.

What should you do if you have no figures on the impact of the project on the Green Deal or cannot find external sources?

If you don't get any clues even through research via experts and the scientific literature in your field, make assumptions, plausibilise and explain them.

What is an impact chain?

An impact chain logically relates the resources used to achieve the desired impact goals. The focus is on targeted and consciously brought about change. Impact chains allow plausibility and feasibility to be checked.

Impact chains can be roughly divided into Input → Activity → Output → Outcome/Impact. There are also empirically proven impact chains that can be divided into inventory → conceptualisation → implementation → measurement → result.

Why is an extrapolation of the contribution - i.e. impact - to the EU Green Deal necessary?

To make the potential of the project for the growth phase visible. The net impact should always be used as the basis for calculation.

Is tokenisation of CO2 emissions, savings, avoidance, storage, ... (e.g. via blockchain technologies) eligible per se?

No, because typically neither AI as an innovation enabler nor tokenisation makes a direct contribution to the EU Green Deal.

What does "do no significant harm" mean?

A project in the call that makes a substantial positive contribution to at least one EU Green Deal priority shall not have a significant negative impact on the six environmental objectives of climate change, sustainable protection/use of water resources, pollution prevention/reduction, adaptation to climate change, transition to a circular economy and protection and restoration of biodiversity and ecosystems.
This is to be presented accordingly.

How can I find out if I fulfil "do no significant harm"?

By going through the six environmental goals of climate change mitigation, sustainable protection/use of water resources, pollution prevention/reduction, adaptation to climate change, transition to a circular economy, and protection and restoration of biodiversity and ecosystems and asking yourself if your project has or can have negative impacts on one or more of them.

What can concrete questions on "do no significant harm" look like?

If your project is active in one of the following Green Deal priority areas, you will find example questions below. It is important that you "check" all six environmental objectives for your specific project.

1. Clean energy:

E.g. Are non-renewable energy sources used? Does the economic activity lead to significantly increased energy demand or is energy used inefficiently?

1. Sustainable mobility

E.g. Does the economic activity lead to increased transport needs? By which means of transport (road, rail, water, air)?

1. Climate protection

E.g. Does the economic activity lead to more CO2-equivalent emissions compared to savings? Does it lead to an increase in pollutant emissions?

1. Biodiversity

E.g. Does the production/manufacturing/use of the economic activity lead to negative impacts on ecosystems? Does economic activity lead to the sealing of large areas of land?

1. From farm to table

E.g. Does the economic activity lead to increased use of fertilisers, pesticides or antibiotics in livestock farming and aquaculture? Is the ecological footprint of food or its packaging being increased?

1. buildings and renovation

E.g. Does economic activity lead to reduced energy efficiency in buildings? Are fossil fuel heating systems being used?

1. sustainable industry

E.g. Does economic activity result in significant inefficiencies in material use? Does the economic activity require critical raw materials?

1. elimination of environmental pollution

E.g. Are pesticides/herbicides used in the product/service? Are microplastics used in the product? How much waste is generated by the economic activity and how is it disposed of?

Is there anything at all in a software solution to show "do no significant harm"?

When training and operating AI systems, large amounts of energy are consumed and water is also used for cooling in data centres; this must be taken into account. Furthermore, there are possible negative consequences due to the specific products or services that are made possible by the AI systems used. For example, an AI logistics system may make 4-hour last mile delivery possible, but only through significantly more journeys with significantly higher CO2 emissions.

Is the use of a "scaled-down" AI model already a sufficient innovation for the Call?

If a) "scaled down" is understood to mean accesses such as those described here, b) this makes significant contributions to the EU Green Deal and c) this has not been technically possible so far and also entails corresponding risks (=innovation, experimental development), then yes. This is to be presented accordingly in the impact chain and expressed in figures.

Can I apply for an AI-Adoption: Green project as an association?

No.

5. Project selection and jury evaluation

How are the projects invited to the jury meeting selected?

All applications received on time will be checked for compliance with the formal requirements (see AI-Adoption Checklist in the download section of the AI-Adoption website).

All projects that meet these formal requirements are assessed and ranked according to four criteria: 1) innovation, 2) growth, 3) trustworthy AI, legal frameworks, standards, norms, certifications and 4) implementation. The best projects will be invited for presentation before the jury.

Does the amount applied for make a difference in the likelihood of receiving funding?

No. The selection is based on the selection criteria of the call, see in the guideline specifications under 7.3.

By when will you know whether you will be invited to the jury presentation?

The respective schedule will be published on the aws website for aws Digitalisation | AI-Adoption I.

How does the jury procedure work?

The selected project team will be invited to the jury presentation by the aws. It may be assumed that the jury is already familiar with the project plans. For each proposed project, 10 minutes are expected to be available for the presentation and 10 minutes for questions from the jury. The jury evaluates according to the assessment criteria, compiles a ranking and makes a jury recommendation.

Does the presentation take place in front of the jury online or on site?

The presentation is on-site.

Can the presentation of the team be done "hybrid" - by physical presence of part of the team and the rest online?

No. In order to present your project the best way possible as well as for organisational reasons, the physical presence of the team on site is required.

Who should be present at the jury presentation by the project team?

The essential members of the team, so that all jury questions can be answered well. This is typically 2–4 people.

When will the results of the jury presentation be announced?

The respective schedule will be published on the aws website under aws Digitalisation | AI-Adoption.

6. Settlement and payment

In what period of time can costs be recognised?

Costs incurred can only be accepted during the project period, i.e., between the date of project start and the date of project end. The duration of the project can be found in the funding agreement.

From what point in time can costs be supported?

Costs are eligible (see programme document 6.1) within the project duration. The project duration is defined in the grant agreement. Costs outside the project duration cannot be funded.. For example, reference date 01/06/20xx but rent will be billed from 01/05/20xx. à Rental costs before 01/06/20xx are not eligible.

Can a service be funded if the service is performed before the project start date from the contract, but the invoice is only issued after that date?

No, both the service and the invoice and payment must be within the project period.

What should be done if it turns out that a project cannot or should not be continued?

Possible options include:

• Termination of the project by a request for termination to the aws.
• Alternatively: Submit modified replacement project to the aws for approval within the framework of the objectives of AI-Adoption or AI-Adoption: Green (pivot) with justification.

Can the costs of implementing trustworthy AI (standard selection and its implementation, communication, training, data analysis and cleansing, ...) be funded?

Yes.

Can the costs of preparing for or implementing EU AI regulations, standards, norms or certifications be funded?

Yes.

Are overhead expenses eligible?

No. Flat-rate compensation of overhead expenses is not possible with this funding.

Are rents and other current payments eligible?

Yes, but only if these are clearly project-related costs and they are more than 150 euros net. The relationship to the project must be documented.

Only costs clearly attributable to the project are eligible. How do I prove this?

You document the costs attributed to the project, e.g. by setting up your own cost centre and making additional postings for the proportional share of costs.

On request, you submit the settlement for the cost centre during the billing check and explain the system, e.g. how are rental costs attributed to the project?

Can costs for patents be charged?

No. In the case of eligible costs, a reference to experimental development must be given, therefore patent costs are not eligible, as these are already attributable to the implementation of commercialisation.

Are start-up costs eligible?

No. In the case of eligible costs, there must be a link to experimental development, so start-up costs are not eligible as they are for implementation.

Are travel expenses eligible?

Basically yes, if there is a clear reference to experimental development, e.g. drive to the pilot project. However, the total travel costs must be at least EUR 150 (net) in order to be eligible. Attention: For grants under Article 25 of the GBER, travel costs are not eligible.

How should travel expenses be entered in the settlement file?

Travel expenses are billed as a collective item, which must amount to at least EUR 150 (net). Costs for transport, overnight stay etc. should not be listed separately. The documents should be submitted in a PDF (including per diems).

Is there a key for how personnel and non-personnel costs must be distributed?

No. The distribution results from the requirements of the respective project.

Within a project, the costs shift between personnel, non-personnel costs and third-party costs compared to the funding agreement. What can be done?

As long as the original objective of the project can continue to be achieved or better achieved and the approved funding amount is not exceeded, the costs applied for can be shifted between personnel, material and third party costs.

If the shifts are:

• less than 40% of the total project cost: the reports shall record and justify the postponements.
• greater than 40% of the total project costs: application with justification for postponement including updated cost structure must be submitted to the aws via the funding manager. The reports shall record any approved shifting.

Example: Total project costs 100k€, of which €70k personnel costs, €20k non-personnel costs and €10k third-party costs. 40% of the total project costs thus correspond to €40k.
Case 1: Personnel costs fall from €70k to €50k, in return, third-party costs rise to €40k and non-personnel costs remain the same at €10k. à Shift is less than 40% and shall be noted in the report with justification. Case 2: Personnel costs fall from €70k to €20k, in return the third-party costs rise to €40k and the non-personnel costs to €40k à shift is greater than 40% and to be approved by the aws.

It should be noted that third-party costs cannot be the focus of the changed use of funds (max. 49%) and the implementation of the project does not have to be carried out primarily by third parties, but predominantly by the beneficiary.

Must time sheets be kept for aws?

No. The programme management is impact-oriented and the payment of funds is based on the success-dependent milestones.

Can flat-rate overhead expenses be promoted?

No.

What personnel costs are eligible for funding?

Only the costs actually incurred and approved under the funding agreement can be funded. The personnel costs must be documented with the corresponding wage account statements and proof of payment.

Can personnel costs for the founders of a start-up also be promoted?

Yes, but here the maximum eligible costs are limited with the Austrian Science Fund (FWF) Senior Post Doc rate valid at the time of the funding commitment. The currently valid personnel cost rate can be found here: www.fwf.ac.at/de/forschungsfoerderung/personalkostensaetze.

Can employees who are also financed from other budgets but also work on the project be billed as personnel costs?

Yes, to the exact extent that they have worked for the project in question. The methodology used, e.g. cost centre accounting, must be explained and, if necessary, proven. For this purpose, the aws can also coordinate with other funding agencies.

How can the personnel costs be proven?

Personnel costs are to be documented by the payroll statements of all persons included in the settlement for the entire settlement period and the proof of payments to the employees and relevant bodies (tax office, social security, etc.). You can find specific examples in the template for cost accounting.

Do hourly rates have to be calculated when settling personnel expenses?  

No.

When should the settlements and milestone documentation be submitted?

In accordance with the contractually agreed periods: at the 1st and 2nd milestone immediately after falling due. At the 3rd and last milestone no later than 2 months after the end of the project period.

Are there any more detailed explanations concerning settlement?

Yes, in principle, the provisions of the Guideline apply mutatis mutandis, unless other provisions have been made in the contract, the guideline specifications, and in the directive.
Please also note the explanations in the template for proof of costs.

How should the personnel costs be settled?

In general, settlement is performed on a monthly basis (refer to the guidelines above). In addition to the name of the person, the settlement period must also be specified in the invoice compilation. The extract from the payroll account (https://www.wko.at/service/steuern/lohnkonto.html) of the accounting department must reflect the corresponding period of time.

When must different price quotes be obtained?

Whenever it is in line with the principle of economic and resource efficiency. In individual cases, the respective internal tendering regulations of the companies can be applied. They must be made available in the event of an audit together with the obtained quotes, upon request.

What is meant by ”... a detailed description of the working time spent” in the case of notes on fees?

That is a description that provides greater clarity and more detail of the specific services provided and the time spent on them. These also form the basis for a possible third-party comparison, i.e. the costs must be customary for the market and comprehensible.

Can costs for the catering at events be funded?

Yes, if the event is primarily attended by external persons and the content of the event serves for the further experimental development of the project. For example, service design or technical workshops with potential customers.

No, if it concerns internal events and/or general marketing, sales or information events.

Can internal services be billed?

Only if appropriate evidence is presented, these are in accordance with the arm’s length principle, and have been approved.

Can invoices issued to other companies, to partner companies or to companies in the same group of companies be recognised as part of the funding?

No. Only invoices made out to the beneficiary of the funding can be accepted.

Can costs for SAAS (Software as a Service) solutions be funded?

Yes, if these are required for development and implementation as part of the project and the costs per invoice are greater than EUR 150 (net). The distinction from general (non-project-related) company activities must be clearly presented.

Can invoices under EUR 150 be included in the settlement?

No. Exception: In the case of travel, individual documents below EUR 150.00, which in total must again be above EUR 150.00 (net), can be accounted for collectively in the travel expense reports.

How fixed are the payments according to the milestone plan?

The payment tranches are defined as part of the funding agreement. These can only be modified in justified cases, after application and approval in the context of an amendment to the agreement.

When is the payment made?

After implementation, documentation and positive examination of the requirements set per milestone in accordance with the funding agreement section “Disbursement of the funding”.

How is payment made?

Directly to the bank account specified by the company in the application.