SINTEF Partnership — Akershus REV4 Scope Narrowing (28.04.2026)

Pregled / Overview

On 28 April 2026, following feedback from Brian Elvesæter (Senior Research Scientist, SINTEF Digital), ALAI narrowed the scope of the Akershus fylkeskommune research funding application from a multi-theme FoU program to a focused pre-project centred on one primary research question. The decision was prompted by Brian's observation that the NOK 750K / 12-month budget could not realistically support the breadth of research themes originally included in REV3 (regulatory ontology, multi-agent compliance orchestration, explainable audit logging, AML/transfer learning, prototype integration, and publication ambitions). REV4 was finalized the same day, with scope reductions agreed and a follow-up email sent to Brian confirming the changes. The Akershus application is scheduled for submission on 4 May 2026 via regionalforvaltning.no. SINTEF's Letter of Intent (LOI) is in process through their Contract Advisor → manager → signature path; a signed copy will follow once internal approvals complete.

---

Brian Elvesæter Feedback / Brian's Feedback (verbatim)

From Brian's email, 28 April 2026 11:27 UTC:

> "Regarding the Akershus application, I think the main issue is that the scope is still too broad for a 750 KNOK / 12-month pre-project. As it stands, it covers several substantial research topics at once, including regulatory ontology, multi-agent compliance orchestration, explainable audit logging, AML/transfer learning, prototype integration, and publication ambitions. > > I also think SINTEF's role should be defined more narrowly. A senior researcher at SINTEF is typically around 3 000 NOK per hour, so to make good use of the budget, our contribution needs to be focused on the parts of the project where research input is most valuable. > > My recommendation would therefore be to use the Akershus project as a more focused pre-project, centred on the question or questions that are most important or most uncertain for your company. I think that would make the application more realistic and also provide a stronger basis for a later IPN application."

Regarding the LOI:

> "I have now sent the LOI to our Contract Advisor for review. After that, it will need to go to my manager for approval and signature on our side. Since 1 May is a public holiday, I cannot guarantee that we will be able to complete the process before then, but I will do my best to follow up as quickly as possible."

---

Donesene Odluke / Decisions Made

ALAI accepted Brian's feedback in full and made the following scope reductions in REV4:

AML / transfer learning — Entirely removed from the Akershus phase. Deferred to IPN Work Package 4 (diaspora-specific pattern recognition, full 36-month research program). Publication ambitions — Removed from Akershus deliverables. Deferred to IPN, where 36-month timeline and multi-partner structure better support publication output. Ontology and explainable audit logging — Reframed as instrumental tools serving the primary research question, not co-equal research themes. Section 1.4 now explicitly labels these as "Metode 1" (minimal NO-HR regulatory ontology) and "Metode 2" (agent dialog as explainable audit trail) — both subordinate to the primary question. SINTEF role — Narrowed to approximately 250 hours over 12 months (using Brian's 3 000 NOK/hour benchmark = NOK 750K), focused on two concrete deliverables: 1. Multi-agent orchestration design (architecture contribution + research-based evaluation methodology for the negotiation protocol) 2. Minimal NO-HR regulatory ontology (machine-readable representation of payment requirements from Finanstilsynet + HNB with conflict-detection semantics) Broad IPN themes — Deferred to the planned 36-month Forskningsrådet IPN application (NORBAL-PAY), where full multi-jurisdictional coverage (NO-HR-RS-BA), extended research questions, and larger SINTEF FTE allocation (1.0 FTE / 36 months) are appropriate.

---

Glavno Forskningspitanje / Primary Research Question

REV4 Section 1.2 establishes a single primary research question:

Norwegian: "Hvordan kan et multi-agent AI-system levere verifiserbare kryssjurisdiksjonelle compliance-beslutninger i NO-HR betalingsstrømmer?"

English gloss: "How can a multi-agent AI system deliver verifiable cross-jurisdictional compliance decisions in NO-HR payment flows?"

Two instrumental sub-questions serve this primary question (Section 1.4): 1. Which minimal regulatory ontology (NO + HR) is required for agents to share common understanding? 2. How does the multi-agent system generate an audit trail that is explainable to regulatory authorities?

These sub-questions are explicitly framed as tools (methods) for answering the primary question, not as separate research contributions of equal weight.

---

Što je Ostalo / What Remains

Sections 5, 6, and 7 of the Akershus application (Regional Verdiskaping / Regional Value Creation, Bærekraft og Samfunnsnytte / Sustainability and Social Benefit, Om Søker / About the Applicant) were kept verbatim from REV3. These sections are required by the Akershus application format and are not subject to the research scope narrowing. They document:

Regional impact: 1–2 new positions in Kløfta/Ullensaker within 18 months, tech competence building in Øvre Romerike, export potential from Akershus headquarters. Sustainability: Addresses UN Sustainable Development Goal 10.c (reduce remittance transaction costs to <3%, Drop targets <1%), digital alternatives to physical cash transfer offices. Applicant background: ALAI Holding AS (org.nr 932 516 136), founder Alem Bašić, 25+ years software development, Entur Tech Lead, Norwegian-Bosnian background, no prior Akershus funding, Innovasjon Norge rejection addressed (innovation height + Norwegian research anchor via SINTEF Digital).

---

LOI Status — 28.04.2026

Brian Elvesæter confirmed on 28 April that the revised LOI (REV2, with pre-commitments removed per SINTEF internal policy) has been forwarded to SINTEF's Contract Advisor for review. After Contract Advisor review, the LOI will proceed to Brian's manager for approval and signature.

Timeline constraint: 1 May 2026 is a public holiday in Norway. Brian stated: "I cannot guarantee that we will be able to complete the process before then, but I will do my best to follow up as quickly as possible."

ALAI action: Akershus application will be submitted on 4 May 2026 with a note that the signed LOI will follow once SINTEF's internal approval process completes. The signed LOI will be attached to the subsequent Forskningsrådet IPN application (NORBAL-PAY, deadline TBD).

---

Files / Datoteke

| Path | Purpose | |------|---------| | ~/Public/sintef-outreach/akershus-tilskudd-draft-REV4.md | Final Norwegian application source (markdown), post linguistic review | | ~/Public/sintef-outreach/akershus-tilskudd-draft-REV4.pdf | Rendered PDF (12 pages, 80KB, AI-NATIVE INFRASTRUCTURE branded), ready for 4 May submission | | ~/Public/sintef-outreach/follow-up-brian-2026-04-28.md | Email draft Alem sent to Brian Elvesæter confirming REV4 scope narrowing | | ~/Public/sintef-outreach/preview/akershus-tilskudd-draft-REV4-page-1.png | Page 1 preview (if generated) | | MC #9985 | BookStack documentation task (this page) | | MC #8720 | SINTEF partnership parent task | | MC #7412 | Funding tracker master task |

---

Sljedeći Koraci / Next Steps

1. Akershus submission — Submit final PDF on 4 May 2026 via regionalforvaltning.no before 23:59 deadline. Application includes note that signed LOI will follow once SINTEF internal approvals complete. 2. Monitor Brian's response — Await Brian's confirmation of: - LOI comfort level with REV4 scope narrowing - Any Contract Advisor flags requiring LOI text adjustment - Signature completion timeline 3. Signed LOI ettersendes — Once Brian confirms the LOI has been signed, attach the signed copy to the Akershus submission portal (if portal allows post-submission attachments) and retain for the IPN application. 4. IPN application preparation — Use Akershus pre-project as basis for full 36-month IPN application (NORBAL-PAY) to Norges Forskningsråd. IPN application will include broader themes deferred from Akershus (AML/transfer learning, publication ambitions, full multi-jurisdictional coverage NO-HR-RS-BA).

---

Lessons / Pouke

Brand tagline source of truth: The AI-NATIVE INFRASTRUCTURE tagline used in REV4 PDF header was sourced from ~/system/templates/proposals/standard-proposal.html, not from CLAUDE.md memory. CEO caught an earlier draft that used a different tagline from LLM memory. Lesson: Always verify brand assets from canonical template files before rendering final PDFs. Pitch scope discipline: The original email pitch to Brian (27 April 2026) already had tight scope (multi-agent compliance for NO-HR payment corridor). The REV3 application document broadened scope to include AML, publications, and multi-theme research. Lesson: Keep grant application scope as tight as the initial pitch — scope creep between pitch and formal application dilutes focus and reduces credibility with research partners. Pre-project framing: Brian's recommendation to use Akershus as a "focused pre-project, centred on the question or questions that are most important or most uncertain for your company" provided clear structural guidance. REV4 adopted this framing explicitly in Section 1.2 (primary question) and Section 3.3 (sequenced pre-phase positioning before IPN). Lesson: When a senior research partner offers structural framing advice, adopt it verbatim — they know what evaluators expect.

---

Reference / Reference

MC tasks: #9985 (this BookStack documentation task), #8720 (SINTEF partnership parent), #7412 (funding tracker master) Email thread: [email protected], [email protected] SINTEF contact: Brian Elvesæter (Senior Research Scientist, SINTEF Digital, Software Engineering Safety and Security), mobile +47 909 74 377 Akershus portal: regionalforvaltning.no (submission deadline 4 May 2026, 23:59) IPN future application: Norges Forskningsrådet (Forskningsrådet.no), Innovasjonsprosjekt i næringslivet (IPN) call (NORBAL-PAY, 36 months, NOK 8M budget)

---

Author: ALAI, 2026 Last updated: 28 April 2026 Document ID: MC #9985