Short answer
On 7 October 2026 Lithuania's Innovation Agency (Inovacijų agentūra) renewed call No. 02-126-K, "Deployment of artificial intelligence solutions". It is open to industrial SMEs: manufacturing and mining companies that have been operating for at least two years and had sales of at least EUR 145,000 in each of their last two financial years. Grants range from EUR 15,000 to EUR 70,000 per project and cover up to 50% of eligible costs. The call funds the purchase of an AI product that is already on the market, its integration with production equipment and IT systems, adapting the model to your data, the hardware it needs and staff training. The project implementation plan (PĮP, the Lithuanian name for the application) must be submitted by 8 December 2026 at 17:00 Vilnius time. The rule most often missed: make no purchase commitments before your application is registered, or the whole project becomes ineligible.
Below is a summary of the conditions, what counts as an eligible cost, a fictional budget example and a checklist for the roughly eight weeks that remain.
What makes this call different
This is one of the few EU-funded digitalisation calls for businesses open in Lithuania right now. We gave a wider overview in EU funding for business digitalisation in Lithuania 2026-2027. This article looks only at AI in manufacturing.
The Ministry of the Economy and Innovation doubled the budget for the renewed call to EUR 7.5 million. It also changed how money is paid out: eligible costs are reimbursed only once the final result is delivered and an expert has confirmed that the costs are eligible. The ministry's announcement stresses that the AI products and solutions installed must actually work. In other words, the grant pays for a working result on the shop floor, not for a purchase.
Key conditions
| Condition | What the rules say |
|---|---|
| Call | No. 02-126-K "Deployment of artificial intelligence solutions", renewed 7 October 2026 |
| Budget | Up to EUR 7,500,000 from the ERDF: up to EUR 2,500,000 for the Capital region and up to EUR 5,000,000 for Central and Western Lithuania |
| Grant per project | EUR 15,000 to EUR 70,000 |
| Aid intensity | Up to 50% of eligible costs, as a grant |
| Type of aid | De minimis aid |
| Selection | Competitive, with a minimum score of 40 out of 100 |
| Deadline | 8 December 2026, 17:00, via the DMS portal (dms.investis.lt) |
| Duration | Up to 12 months from signing the grant agreement; work must start within 3 months and finish by 1 September 2029 |
| Payment | For the final result, after a positive expert assessment |
| After the project | The investment must be maintained for 3 years after funding ends |
A quick rule of thumb from these numbers: to receive the maximum EUR 70,000 you need at least EUR 140,000 of eligible costs, and the minimum EUR 15,000 grant requires at least EUR 30,000. Because the grant is paid at the end, the company has to fund the whole project itself first, from working capital or a loan, and must show where that money comes from when it applies.
Who can apply
Applicants must meet two specific criteria.
1. An established, financially capable SME. The company has been registered and operating for at least two years before applying, and its annual sales were at least EUR 145,000 in each of the last two financial years.
2. Its main activity is industrial. Under the Lithuanian classification of economic activities (EVRK 2.1, aligned with NACE Rev. 2), the main activity must fall under:
- Section B, mining and quarrying, except extraction of crude petroleum and natural gas (division 06), peat extraction (class 08.92) and support activities for oil and gas extraction (group 09.1);
- Section C, manufacturing, except coke and refined petroleum products (division 19).
On top of that, industrial activity must have generated at least 51% of the company's total sales in each of the two years before applying. A company that manufactures but earns most of its revenue from wholesale trade may fail this test, so check it against your latest financial statements.
The call is for SMEs only. Large companies and public sector bodies cannot apply. If you belong to a group, SME status is assessed together with linked and partner enterprises, which is why an SME status declaration is required.
What the grant covers
Eligible costs fall into three groups.
Buying and deploying the AI product or solution
- Product and licence: purchase of the AI product or solution, a licence or a perpetual right of use.
- Implementation services: integration with your production equipment and/or information systems; building or configuring the interface through which the AI output triggers an action in production; configuration and calibration; training or adapting the model on your data; preparing and labelling training data; validating performance; a pilot run; training staff to use the solution.
- Essential hardware: sensors, cameras, lighting, edge computing or inference devices and actuators that the solution cannot work without and that are used only for it. These costs cannot exceed 50% of the eligible costs of the AI product or solution.
Software and licence rental
Subscription-based solutions also qualify where you pay to rent software or licences.
Indirect costs
These follow the general EU funds administration rules.
What integration means in practice
For most manufacturers the bulk of the work is not the AI model itself but connecting it to what already runs the plant: machines and controllers, the production management system, the ERP or accounting software. The call recognises this directly. Integration, and an interface through which the AI result triggers an action in production, are eligible costs. For example, a camera spots a defect, the system automatically diverts the part to the reject line and logs the event in production records.
We described which connections are realistic with Lithuania's most common accounting systems in Integrating with Rivilė, Directo and Agnum. For broader examples of AI automation, see Business process automation with AI: 5 real examples.
What probably won't qualify
- Building a new AI product from scratch. With the application you must prove that the AI product or solution is already available on the market on the day you apply: its technical specification and public information on how to buy it (pricing, a product page) or documents showing it is sold. Adapting an existing product is fine; developing a brand-new one is not what this call is for.
- General-purpose hardware. Computers, servers or cameras that will also be used for other things fail the "used only for this solution" test.
- Work started too early. More on that below.
Deadlines and pitfalls
Do not start the project before your application is registered. The project is considered started at the first legal commitment to buy the AI product or solution. If you sign a contract or confirm an order with a supplier before the Innovation Agency registers your application, the entire project becomes ineligible. Collecting quotes and negotiating is allowed; committing is not.
Costs after registration are at your own risk. You may start work between registration and the funding decision, but if the grant is not awarded, those costs stay with you.
Payment only for results. Plan cash flow as if the grant arrives only at the very end, and tie supplier payments to results that can be checked.
Time. There are about eight weeks until 8 December. In that time you need to choose a solution, get quotes, gather financial documents and complete every annex. An incomplete or poorly argued Annex 3 (information for the selection criteria) is grounds for rejecting the application without asking for more documents.
How projects are scored
Because the call is competitive, meeting the conditions is not enough; you also need points. There are three priority criteria:
| Criterion | Maximum points |
|---|---|
| The AI solution belongs to one of these fields: machine learning, computer vision, smart robotics and automation, natural language processing, digital twins | 40 |
| The applicant counts as a company with high digital intensity under the DESI methodology | 30 |
| The applicant's own contribution to the project | 30 |
Projects need at least 40 points out of 100. In practice the application has to state clearly which AI field the solution belongs to and which specific production problem it solves.
AI use cases that fit manufacturing
Some typical scenarios that match the AI fields named in the call:
- Quality inspection with computer vision: cameras and a model check parts, welds, packaging or labels on the line.
- Predictive maintenance: a machine learning model warns of an approaching failure from vibration, temperature or current data.
- Production planning and demand forecasting: a model helps plan shifts, raw material orders and machine load.
- Smart robotics: a vision-guided robot picks and sorts irregularly shaped parts.
- Digital twin: a model of a line or process lets you test changes without stopping production.
- Natural language processing: handling orders, drawing notes or quality complaints, with the result passed on to production systems.
A budget example
The company and all figures are fictional. The example shows how to structure a budget under the call's rules, not real market prices.
Medžio Detalė UAB (fictional) makes furniture board components near Kaunas. It has 60 employees and EUR 4 million in annual sales, almost all from manufacturing. Edge and surface defects are currently checked by hand, and some defects are only discovered by customers. The company wants to deploy an off-the-shelf computer vision quality inspection solution and connect it to the line controller and to its production records.
| Cost item | EUR |
|---|---|
| Licence for the AI quality inspection solution | 50,000 |
| Implementation: integration with the line controller and production records, reject-diverting interface, calibration, adapting the model to company data, data labelling, validation, pilot run, training | 40,000 |
| Hardware: industrial cameras, lighting, edge computing device (used only for this solution) | 20,000 |
| Total eligible costs | 110,000 |
| Grant requested (50%) | 55,000 |
| Company's own funds | 55,000 |
Hardware comes to EUR 20,000, less than half of the licence alone, so it stays within the 50% cap under even the strictest reading. The grant requested sits between EUR 15,000 and EUR 70,000, and the solution falls under computer vision. Until the grant is paid, the company funds the full EUR 110,000 itself, so it attaches a board decision allocating the funds and, if needed, a bank letter confirming a loan.
Checklist before 8 December
- Check eligibility: years in operation, EUR 145,000 in sales in each of the two years, the activity code and the 51% industrial revenue share, and SME status including linked companies.
- Check your de minimis headroom: under Regulation (EU) 2023/2831, de minimis aid to a single undertaking cannot exceed EUR 300,000 over any three-year period, and other de minimis aid you have received counts towards that ceiling.
- Describe the problem in numbers: current scrap, downtime or manual work, and the result you expect.
- Choose a solution that is already on the market and collect the evidence: technical specification, public pricing or a product page.
- Get quotes from the AI developer, vendor or official representative, with licence, implementation services and hardware listed separately.
- Plan the integration: which equipment and systems the solution exchanges data with, what action it triggers in production, and who prepares the data.
- Define acceptance criteria: how you and the expert will verify that the solution works. The approach in Technical specification for a software project works well here.
- Secure funding for the full project amount and prepare a board decision or bank letter.
- Collect declarations and annexes: the do-no-significant-harm declaration (Annex 2), information for the selection criteria (Annex 3), the trade commitments and national security declaration (Annex 4), the SME status and single undertaking declarations, and information on state aid received.
- Submit through DMS with time to spare, not on the last day, and do not sign purchase contracts until your application is registered.
Questions about completing and assessing the application go to the Innovation Agency's Project Evaluation Division: didiegimas@inovacijuagentura.lt, +370 614 53303. The call documents are in Lithuanian.
Next steps
If you have a clear production problem and are considering an AI solution, start by working out how it will connect to your equipment and systems, and turn that into a clear implementation plan with acceptance criteria. DESAmedia can help you define the integration scope and an estimate to sit alongside vendor quotes, and then build the connections to your production and accounting systems. See Integrations and automation, AI agent development and IT consulting and system design, or get in touch through our contact page.
This article is based on official sources as of 13 October 2026. The binding conditions are set out in the call's funding conditions description (PFSA), so check its current version before you apply.
Sources
- Innovation Agency. Deployment of artificial intelligence solutions (call page, in Lithuanian)
- esinvesticijos.lt. Deployment of artificial intelligence solutions (call, PFSA requirements, selection criteria, annexes; in Lithuanian)
- Ministry of the Economy and Innovation. EUR 7.5 million for businesses to deploy AI solutions (7 October 2026, in Lithuanian)
- Order No. 4-418 of 2 October 2026 amending the PFSA (e-tar.lt, in Lithuanian)
- Commission Regulation (EU) 2023/2831 on de minimis aid (EUR-Lex)
- INVESTIS data exchange portal (DMS)