A recorded job

Bid on a public AI tender

Use Datailo to run the job "Bid on a public AI tender": find an open AI-related public tender we could bid on, qualify it, classify the AI system we would offer under the AI Act, check the Lithuanian procurement-law points, and give me one bid pack with a compliance section.

For bid manager or founder at a software company that builds AI systems and sells to the Lithuanian public sector. When an AI-related public tender is open; the company must decide this week whether to bid and what the bid says about the AI Act and procurement law.

Recorded
2026-09-25
Tool calls
75
Services
3
  1. 1Task
  2. 2Tools
  3. 3Checks with you
  4. 4Document

Your AI assistant datailo connected

Use Datailo to run the job "Bid on a public AI tender": find an open AI-related public tender we could bid on, qualify it, classify the AI system we would offer under the AI Act, check the Lithuanian procurement-law points, and give me one bid pack with a compliance section.

  • Victano: Find and qualify the tender· 143 sfind-this-week · qualify · deep-read
  • You confirmwhich tender to bid on (or stop at a no-bid)
  • You confirmwhat the offered AI system does and whether the company is its provider
  • Rekvira: Classify the AI system and list the provider's AI Act duties· 9 shigh-risk-triage · deployer-obligations
  • Avokata: Check the Lithuanian procurement-law points· 44 smemo · procedure-deadlines
  • You acceptthe compliance section before it goes into the pack
  • Victano: Assemble the bid pack· 143 sresponse-plan

Prepared document: Bid pack with a compliance section. A conditional bid on Lot 2 of the Geology Fund tender, the AI Act classification with its evidence, and the Lithuanian procurement-law points, each with its source.

Read the document

Recorded on the live hub 25 Sep 2026 in Claude Code 2.1.282. The rows are the job’s stages, with the playbooks and time of each; the full log of 75 tool calls is below.
  • Duration

    5.5 min

    from the request to the document

    Claude Code 2.1.282

    2026-09-25

  • Tool calls

    75

    each one below

    datailo

    2026-09-25

  • Sources read

    22

    acts, rulings and notices

    avokata · rekvira · victano

    2026-09-25

  • Citations verified

    9

    quotes checked against the text

    rekvira

    2026-09-25

/ 01The document

Bid pack with a compliance section Status: As of 25 September 2026

The tender

Geology Fund documents: archiving, digitisation and e-services, Lot 2

Aplinkos projektų valdymo agentūra, Vilnius. Open procedure, published 23 Sep 2026.

Deadline 2026-10-02 11:00

ted:655738-2026. The time has no time zone in the record; check it on the portal.

Bid or no-bid

Conditional bid on Lot 2

It becomes a bid only if the specification allows automated recognition and classification of documents, and turnover and references meet its selection criteria.

Recorded: conditional

AI Act

Probably not high-risk. Yours to confirm.

No Annex III area covers archive-document recognition for a geology agency. The company would be the provider under Article 3(3); the agency the deployer.

Art. 6(1)Art. 6(2)Annex III

Read this first

Victano holds only the one-paragraph notice for this tender, not the specification. Selection criteria, award weightings, lot values and the contract are in the buyer’s documents, so the pack is built from the notice alone.

Where you decide

  • After stage 1: the user confirms which tender to bid on (or stops at a no-bid)
  • Before stage 2: the user confirms what the offered AI system does and whether the company is its provider
  • Before stage 4: the user accepts the compliance section before it goes into the pack

What it will not do

  • Submit the bid or contact the buyer
  • Decide the AI system's classification for the company: it lays out the test and the evidence, the user decides
  • Act as your lawyer: it assembles the provisions and the practice, the lawyer reads them

Checkpoint: I'm assuming you accept the compliance section (Stage 2: AI Act; Stage 3: Lithuanian procurement law) as drafted below, and I'm finishing the pack on that basis.


Bid pack: Geology Fund documents: archiving, digitisation and e-services (Lot 2: document-management system)

Buyer: Lietuvos Respublikos aplinkos ministerijos Aplinkos projektų valdymo agentūra (Vilnius, LT011) Notice: ted:655738-2026, https://ted.europa.eu/en/notice/-/detail/655738-2026 (open procedure, services, 2 lots, published 2026-09-23) Deadline: 2026-10-02 11:00 (time as recorded by Victano, which gives no timezone; check it on the portal). That leaves 5 working days from today. Prepared: 2026-09-25 Sources as of: Victano 2026-09-24 07:05 UTC (free tier, 24 h delay) · Rekvira (EUR-Lex, AI Act consolidated incl. Reg. (EU) 2026/1744) 2026-09-24 19:05 UTC · Avokata (e-TAR) 2026-09-24 21:39 UTC, court practice 2026-09-25 06:13 UTC

Read this first: Victano holds only the one-paragraph notice for this tender, not the specification. Selection criteria, award weightings, lot values and the contract are in the buyer's documents (criteria_source: in_documents). This pack is built from the notice alone. Sections 2 and 5 are skeletons until someone downloads the pack.

1. The tender and the bid/no-bid call

Conditional bid on Lot 2. It becomes a bid only if the specification allows automated recognition and classification of documents, and our turnover and references meet its selection criteria. Whoever reads the pack decides by Monday 28 Sep, noon.

  • Why this one. It was the best of 1,049 open Lithuanian notices ranked for "AI system" searches (Lithuanian and English). Only one open notice is explicitly AI: cvp:9687077 (AI-based cultural-heritage monitoring, EUR 151,250). It closes Mon 28 Sep with 1 working day left, and its clarification window closed on 21 Sep. Recorded as no-bid, deadline_too_tight.
  • Also rejected. LITGRID's renewable-generation forecasting tender (ted:504869-2026, 11 working days, quality and price) is ML work. We have no energy-forecasting references, so it's no-bid, wrong_scope.
  • Capability fit for Lot 2 ("Dokumentų valdymo sistemos sukūrimo paslaugos"). A system for digitised archival documents is document-AI territory: OCR, classification and metadata extraction. The notice doesn't say the buyer wants AI. The AI element is our proposal.
  • Scores: capability 3 · past performance 3 (two municipal projects; whether they count as document-management references is unknown) · geography 5 · value fit 3 (lot value not published) · deadline 2 · competition 3 (unknown) · strategic value 4 (national agency, e-services reference).
  • Recorded in Victano: conditional / pending_clarification.

2. What the specification requires, and what would sink the bid

RequirementWhere in the specificationOur positionRisk
Lot 2 scope: build a document-management system for the Geology FundNotice summary (TED)have (software, document AI)Low if it's a custom build; high if they want a specific commercial product
Lot 1 (archiving and digitisation services)Notice summarycannot (physical archiving) → bid Lot 2 onlyOnly a risk if the lots have to be bid together (notice says up to 2 lots per bidder)
Selection criteria: turnover, references, staffBuyer's documents, not readUnknownPossibly fatal. Under VPĮ 47(3)(1) a turnover floor may not exceed 2× the lot value unless justified. With a EUR 2 M turnover we're safe for any lot up to EUR 1 M.
Award criteria and weightsBuyer's documents, not readUnknownChanges the whole strategy if the award is price-only
Deadline 2 OctNoticetightPriceable. No room for a partner or new evidence
Clarification deadlineBuyer's documents, not readUnknownMay already be close; questions have to go through CVP IS

3. AI Act: classification and the provider's duties

The system, as confirmed at the checkpoint: a document-AI module inside the Geology Fund document-management system. It runs OCR on digitised geological archive documents, classifies them and extracts metadata for search and e-services, and people review the output. We develop it and put it into service for the buyer under our own name. Under Art. 3(3) (verified, quote held) that makes us the provider; the agency is the deployer.

Classification: probably not high-risk. This is for you to confirm once the specification is read.

  • Art. 6(1) doesn't apply. The system isn't a safety component of, or itself, a product under the Annex I harmonisation legislation. Art. 6(1a) adds that systems used solely for "automation or convenience" are not safety components (wording in force since 2026-07-27, from Reg. 2026/1744).
  • Art. 6(2) most likely doesn't apply. None of the eight Annex III areas covers archive-document recognition for a geology agency: biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration, justice/democracy. Annex III(2) needs a safety component in critical digital infrastructure, road traffic, water, gas, heating or electricity, and nothing here matches that.
  • Art. 6(3) last subparagraph. Profiling of natural persons would make an Annex III system always high-risk. Our design must not profile the people named in the archives.
  • Art. 6(4) (document the non-high-risk assessment, register under Art. 49(2)) applies only if the system is an Annex III system you claim is exempt. It shouldn't be triggered, but keep a one-page classification note on file anyway.
Provider dutyPinpointVerifiedWhat the bid should say
Classification rules (not high-risk)Art. 6(1), 6(1a), 6(2), Annex IIIheld (quotes found, 2026-09-25 text)"The system falls outside Art. 6(1) and Annex III; our classification note is attached."
No profiling of natural personsArt. 6(3), last subpara.held"The system performs no profiling of natural persons."
AI literacy for our staff and the buyer's operatorsArt. 4(1) (Omnibus wording, from 2026-07-27)heldOffer operator training and a short guide on how the model works and where it fails.
Disclose AI interaction, if the e-services include a chat or assistantArt. 50(1)held; applies only if the scope includes oneOnly if the specification asks for an assistant.
Machine-readable marking of synthetic text, if we generate summariesArt. 50(2) (exemption for output that doesn't substantially alter the input)held; applicability needs reviewOCR and extraction likely fall within the exemption. Generated summaries would need marking.

Recorded in Rekvira: 3(3) and 4(1) applicable; 6(1) not applicable; 6(2), 50(1) and 50(2) needs_review (awaiting the specification).

4. Lithuanian procurement law: the points that apply

All points below are from the Viešųjų pirkimų įstatymas (Public Procurement Law, "VPĮ"), No. I-1491. It is in force: consolidated version valid from 2025-02-01, checked with check_validity on 2026-09-25.

PointProvisionIn forceCourt practiceWhat the bid must do
Mandatory exclusion: final convictions (criminal organisation, bribery, fraud, money laundering, etc.) of the company or its manager or accountant within 5 yearsVPĮ 46(1), (2)yes, 2025-02-01 versionnone readDeclare in the ESPD; check the manager's and accountant's records.
Mandatory exclusion: unpaid taxes or social insurance over EUR 50VPĮ 46(3) and 46(3)(2)yesnone readCheck with VMI and Sodra before submitting.
Mandatory exclusion: false or withheld information, material breach of a public contract in the last 3 years, not meeting reliable-taxpayer criteriaVPĮ 46(4)(4), 46(4)(6), 46(4)(7)(b)yesSupreme Court eCIK-904/2025 and Court of Appeal e2-463-553/2026 ¶20 cite 46(4)(6) (not read)Answer the ESPD exactly. A false answer is itself an exclusion ground.
Exclusion at any stage; self-cleaningVPĮ 46(9), 46(10)yesSupreme Court 2026-01-28, proceeding No. 2-55-3-00814-2025-7, ¶4, cites 46(9) (not read)Update the buyer if anything changes before signature.
Qualification must be proportionate and tied to the subject matter; turnover floor ≤ 2× estimated valueVPĮ 47(1), 47(3)(1), 47(6), 47(7)yesLAT e3K-3-67-933/2026 ¶¶25, 34 (read): requirements follow the Public Procurement Office methodology (Metodika) and must fit the subject matter and value; excessive or specific requirements unlawfully restrict competitionIf a criterion is disproportionate, challenge it under Art. 101-102 before the deadline. Evidence references with client certificates (47(6)).
Challenge: a claim to the buyer (pretenzija) must come first, filed electronicallyVPĮ 101(3)yesnone readChallenge a specification term or a decision through CVP IS.
Time limits: 10 days from the buyer's electronic notice of a decision (5 working days in a simplified procurement); 15 days to court if the claim goes unanswered; 6 months to have the contract declared invalidVPĮ 102(1)(1), (2), 102(2), 102(3)yesLAT eCIK-260/2026 cites 102(4) (not read); Court of Appeal e2A-463-370/2025 ¶50 cites 102(1)(2)Put a watch on every decision notice: day 0 is the day it's sent, not the day it's read.

5. Compliance matrix

This is built from the notice. Every row needs re-checking against the specification.

Specification itemResponse sectionEvidence we attachOwner
ESPD: exclusion grounds (VPĮ 46)ESPD part IIIVMI/Sodra certificates, criminal-record extractsCEO / accountant
Right to carry on the activity (47(2))ESPD part IV.ARegistry extractAdmin
Financial capacity (47(3))ESPD part IV.BFinancial statements (~EUR 2 M turnover)Accountant
Technical capacity: references and staff (47(6))ESPD part IV.CCertificates from both municipalities, CVsBid manager
Information securityTechnical proposalISO/IEC 27001 certificateCISO
Lot 2 functional scopeTechnical proposalSolution design, document-AI demoTech lead
AI Act positionTechnical proposal annexClassification note (Art. 6 / Annex III), AI-literacy training plan (Art. 4(1))Tech lead
PricePrice formPrice build-upCEO

6. Questions to send the buyer before the deadline

These go through CVP IS only. I have not contacted the buyer.

  1. Lot 2: may the document-management system include automated text recognition, classification and metadata extraction (AI), or is a specific product or platform required? (notice summary, Lot 2)
  2. What are the award criteria and their weightings for Lot 2? (not in the TED notice)
  3. What is the estimated value of Lot 2, and what minimum turnover and reference contracts are required? (relevant under VPĮ 47(3)(1))
  4. What is the clarification-question deadline?
  5. Do the archive documents contain personal data, and where must the system be hosted?

7. What could not be verified

  • The whole specification. I have no web-fetch tool, and Victano holds only the notice. Selection criteria, award weightings, lot values, the draft contract and the clarification deadline are all unread.
  • Whether the buyer wants AI at all. "AI-related" is our proposal, not the notice's wording.
  • The deadline's timezone. It's recorded as 11:00 with no zone.
  • The AI Act classification is conditional on the scope described above. Your decision. I didn't check the AI Act's application dates (Art. 113) or GDPR/archive-law duties.
  • Court rulings marked "not read" are listed only because they cite the articles; I don't know what they held. Only e3K-3-67-933/2026 ¶¶25 and 34 was read.
  • Company facts: tax status, criminal records, and whether the two municipal projects count as document-management references.

Due today (Fri 25 Sep): the bid manager downloads the tender pack from CVP IS or TED. Mon 28 Sep, noon: confirm or drop the bid against sections 2 and 6, and send the questions. Wed 30 Sep: draft complete. Thu 1 Oct: review by someone other than the author, and upload. Fri 2 Oct, 11:00: deadline.

Bid pack with a compliance section, as the run returned it. The text is verbatim; only its layout is ours.

/ 02How you can check it

  • Every AI Act duty carries its article and paragraph and is checked against the regulation's own text with Rekvira's citation check
  • Every procurement-law point carries its article (and a case number where a court ruled on it)
  • Each source states its data-as-of date
  • A section lists what could not be verified

/ 03The recorded run

Recorded 25 Sep 2026 in Claude Code 2.1.282 (headless: claude -p): 5.5 min, 75 tool calls, 9 citations verified, 22 sources read.

Background: The EU's auditors found that bidders and buyers see public procurement as a heavy administrative burden. They also found that lengthy, resource-intensive procedures put companies off, especially small firms with few administrative staff. European Court of Auditors, Special report 28/2023: Public procurement in the EU (paragraphs 48 and 99)

Model claude-opus-5-5. The situation given to the agent is illustrative; the law, the notices and every citation came from the live services during the run.

Every tool call (75)
  1. datailo_start_here 0.1 s
  2. datailo_account 0.1 s
  3. datailo_workflow datailo bid-on-a-public-ai-tender 0.1 s
  4. datailo_enable victano 0.3 s
  5. datailo_enable rekvira 0.4 s
  6. datailo_enable avokata 0.7 s
  7. datailo_workflow victano find-this-week 0.9 s
  8. datailo_tool_schema victano build_profile 0 s
  9. datailo_tool_schema victano search_opportunities 0 s
  10. datailo_invoke victano build_profile 0.7 s
  11. datailo_invoke victano match_profile 29.3 s
  12. datailo_invoke victano search_opportunities 35.2 s
  13. datailo_invoke victano match_profile 26.9 s
  14. datailo_invoke victano search_opportunities 32.1 s
  15. datailo_invoke victano search_opportunities 6.6 s
  16. datailo_invoke victano get_opportunity 0.7 s
  17. datailo_invoke victano get_opportunity 0.7 s
  18. datailo_invoke victano get_opportunity 1.5 s
  19. datailo_workflow victano qualify 1.8 s
  20. datailo_invoke victano read_document 0.4 s
  21. datailo_invoke victano read_document 0.6 s
  22. datailo_invoke victano read_document 3 s
  23. datailo_tool_schema victano record_decision 0 s
  24. datailo_invoke victano record_decision 0.2 s
  25. datailo_invoke victano record_decision 0.4 s
  26. datailo_invoke victano record_decision 0.5 s
  27. datailo_workflow victano deep-read 0.6 s
  28. datailo_workflow rekvira high-risk-triage 0.3 s
  29. datailo_workflow rekvira deployer-obligations 0.4 s
  30. datailo_invoke rekvira list_regulations 0.3 s
  31. datailo_invoke rekvira search_regulation 0.5 s
  32. datailo_invoke rekvira read_unit 0.5 s
  33. datailo_invoke rekvira read_unit 0.4 s
  34. datailo_invoke rekvira lookup_obligations 0.4 s
  35. datailo_invoke rekvira read_unit 0.2 s
  36. datailo_invoke rekvira read_unit 0.3 s
  37. datailo_invoke rekvira read_unit 0.3 s
  38. datailo_tool_schema rekvira verify_citation 0 s
  39. datailo_invoke rekvira verify_citation 0.3 s
  40. datailo_invoke rekvira verify_citation 0.4 s
  41. datailo_invoke rekvira verify_citation 0.4 s
  42. datailo_invoke rekvira verify_citation 0.3 s
  43. datailo_invoke rekvira verify_citation 0.2 s
  44. datailo_invoke rekvira verify_citation 0.3 s
  45. datailo_invoke rekvira verify_citation 0.4 s
  46. datailo_invoke rekvira verify_citation 0.2 s
  47. datailo_invoke rekvira verify_citation 0.4 s
  48. datailo_tool_schema rekvira record_assessment 0 s
  49. datailo_invoke rekvira record_assessment 0.3 s
  50. datailo_invoke rekvira record_assessment 0.3 s
  51. datailo_invoke rekvira record_assessment 0.4 s
  52. datailo_invoke rekvira record_assessment 0.3 s
  53. datailo_invoke rekvira record_assessment 0.4 s
  54. datailo_invoke rekvira record_assessment 0.4 s
  55. datailo_workflow avokata memo 0.4 s
  56. datailo_workflow avokata procedure-deadlines 0.6 s
  57. datailo_tool_schema avokata search_law 0.1 s
  58. datailo_tool_schema avokata read_paragraphs 0 s
  59. datailo_invoke avokata search_law 2.9 s
  60. datailo_invoke avokata search_law 3.7 s
  61. datailo_invoke avokata search_law 5.4 s
  62. datailo_invoke avokata read_paragraphs 1.5 s
  63. datailo_invoke avokata read_paragraphs 1.5 s
  64. datailo_invoke avokata read_paragraphs 3 s
  65. datailo_invoke avokata check_validity 0.5 s
  66. datailo_tool_schema avokata cited_by 0.6 s
  67. datailo_tool_schema avokata search_practice 0.3 s
  68. datailo_invoke avokata cited_by 2.8 s
  69. datailo_invoke avokata cited_by 4.9 s
  70. datailo_invoke avokata cited_by 7 s
  71. datailo_invoke avokata read_paragraphs 0.9 s
  72. datailo_invoke avokata search_practice 2.9 s
  73. datailo_invoke avokata search_practice 3.5 s
  74. datailo_invoke avokata read_paragraphs 1.3 s
  75. datailo_workflow victano response-plan 0.3 s

/ 04Start it yourself

Connect Datailo, then paste this request with your own facts.

Use Datailo to run the job "Bid on a public AI tender": find an open AI-related public tender we could bid on, qualify it, classify the AI system we would offer under the AI Act, check the Lithuanian procurement-law points, and give me one bid pack with a compliance section.

Run it on your own case.

Sign in once with your email: one free account for every service, no card.

https://mcp.datailo.com/mcp