Make Your Real Estate & Construction Documents Talk to Your AI.
Cue turns the documents trapped across construction and asset projects into reliable, connected, AI-usable knowledge — and runs domain-specific agents on top of it.
The data already exists. It just isn't usable: locked in PDFs, CAD and BIM files that modern AI can't read. Cue consolidates this information into a single, organised index. It then interacts with the agents and tools that teams are already using, providing each user with access to comprehensive knowledge.
8-year federal CDE
Won the Swiss federal CDE tender (Astra, Armasuisse, BBL) in consortium with Zühlke and Vrame.
Helvetia
PoC completed; licence under negotiation. The decisive near-term opportunity.
~15 FTE of output
A small team using modern AI tooling to ship platform output equivalent to roughly 15 engineers.
A category is forming — and we're already standing in the middle of it
AEC/O intelligence is consolidating into one layer, and the position is still open. We hold what nobody else does: an 8-year federal mandate, an in-production ontology the industry already runs on, and a data flywheel that compounds with every client correction. The industry's move to managed agents — Claude, ChatGPT, the MCP standard — handed us distribution we had budgeted to build. So we're closing pre-seed early, pivoting to ARR now, and raising seed to build the capture machinery while it's still cheap to own. In two years this layer will have a default. We intend to be it.
Three lessons from the trade-fair floor
Not assumptions — what every prospect told us, repeatedly, across the industry's three biggest fairs.
Every company struggles to feed, manage and access its own data
The value is trapped in PDFs, CAD and BIM files that modern AI simply can't read. The data exists — it just isn't usable.
Every company already runs systems you have to work with
There's no appetite for rip-and-replace. Anything new has to connect to what's already there, not ask the organisation to start over.
Every company consumes its information differently
Different teams need different things from the same data, so one rigid, one-size product never fits.
One governed index. Three ways to use it.
Cue brings the client's whole world into a single, connected index — then meets users wherever they already work.
Cue Index
An AI-native database, built ground-up to speak the language of AI agents. Auto-populated from the client's full world — documents, BIM/CAD, ERP, GIS — and easily curated for full control and steerability.
Open ontology + Tuner + Studio
An open, standards-aligned ontology shapes the data to each client's domain. The Tuner lets clients adapt their schema without us in the loop. The studio collects Dashboards, Skills and Tools ready to use.
Three surfaces on the index
External Agents
Cue inside Claude and ChatGPT via MCP — the bridge from mountains of documents to the AI tools teams already trust.
Cue Studio
An agent-native toolkit to build dashboards, browsers and analytics on the index in minutes — instead of chasing one use-case after another.
Cue Agent
The genuinely industry-specific work: compliance, quality checks, due diligence — the tasks a generalist agent can't do well.
The people who wrote the standards this industry runs on
Founded August 2024 as an Amberg Group spin-off. Not outsiders disrupting AEC/O — the authors of its data standards, with the board seats and regulatory access to match.
7 years on the boards that set Swiss AEC policy
SIA KIN president and buildingSMART Switzerland board — direct decision-maker access at every major Swiss owner-operator. Credibility earned, not bought.
The ontology is a founder's PhD
Mads Rasmussen authored BOT, now an open W3C standard. The information model Cue runs on is his doctoral work — not a schema licensed from someone else.
CHF 770K closed by one person
All booked revenue to date driven solely by Philipp through network selling. Zero paid marketing, zero SDR team, CHF 4,100 CAC.
Philipp Dohmen
7 years CDO at Amberg Group. 7 years president of SIA KIN (Swiss Society of Engineers and Architects, informatics commission) and buildingSMART Switzerland board member. Architect, Dipl. Ing. ETH. Led the CDE Bund consortium bid and closed every franc of the CHF 770K booked to date. Leads company, commercial and fundraising.
LinkedIn
Mads Rasmussen
Author of BOT (Building Topology Ontology) — his PhD, now an open W3C standard and the foundational linked-data schema for the AEC industry worldwide. Civil Engineer, Copenhagen. Leads QAECY's information modelling and the Enterprise Knowledge Graph — and is the Nordic bridgehead.
LinkedIn
Manos Argyris
5 years Head of Data Science at Amberg Group, with MIT data science credentials. Architect (Dipl. Ing.) turned data scientist — ML and knowledge-graph specialist. Architect of the Cue intelligence layer; leads platform engineering.
LinkedIn
Christian Frausing
Architectural Engineering + Computer Science. Pipelines, ingestion and data quality — the ETL and AI infrastructure connecting models to client systems. Joined December 2024.
LinkedIn
Already in production — and compounding from here
The platform is live in federal and enterprise environments today: pipelines running, documents resolved, an ontology in production. That is the asset a competitor would have to catch up to. What this round buys is the capture machinery that turns every client's use of it into a lead that widens — while it's still early and cheap to own.
Two loops, by design. For paying clients, results improve inside their own account — isolated, on-prem where needed, never pooled into a shared model. On the free tier, opt-in usage lets us observe and improve Cue itself. Competitors can rebuild the machinery; they can't rebuild the verified, provenance-stamped corrections our users generate — and in this industry, keeping each client's data in its own walls is the point, not a limitation.
In production today — and what each layer accrues
Connecting information across documents — the "holy grail" of connectivity — runs live today at ~100M nodes across federal and enterprise deployments. It becomes progressively harder to catch as the corpus of resolution corrections trains it on signal no competitor has. Defensibility grows with every correction.
Auto-populated and, crucially, easily curated. Curation is the capture surface: it stamps every fact with provenance and confidence and turns use into the corpus. In a liability-averse industry, that verified, traceable property is what actually gets bought.
Each client's data stays in its own walls — isolated, and on-prem for clients with high security needs — never pooled into a shared model. The trust model insurers, federal owners and infrastructure operators require, and that generic LLM apps can't credibly offer.
Head start & amplifiers — on top of the moat, not the moat itself
A mature, in-production information model, published openly (dev.qaecy.com/ont/), so there's no lock-in. A genuine head start and the scaffolding the corpus refines over time — but the durable asset is the upkeep and the corrections, not the static schema.
A premium expression built on accumulated trajectories, with frontier reasoning kept inside an AEC/O harness. It competes on one axis — who uses the Cue Index best to serve AEC/O — but as an amplifier on the asset, never its source.
IP & defensibility position
| Asset | Type | Note |
|---|---|---|
| Data flywheel (verified corrections + provenance) | Compounding asset | The moat — improves each client's results in isolation (paid) and Cue itself from opt-in free usage. Richer capture (trajectories) is what this round builds. |
| Cue entity resolution | Proprietary | Highest-value capability; defensibility scales with the corpus behind it. |
| Index capture & provenance layer | Proprietary | The curation and trust surface that turns usage into the corpus. |
| Cue ontology | Open / standards-aligned | Battle-proven head start, published openly — value is the proven model and domain expertise, not ownership. |
| Founder & employee IP assignments | In place | Standard assignment of all work product to QAECY AG. |
We're building the category others will be measured against
AEC/O intelligence is consolidating into a single layer, and Cue is the only platform holding all three pieces it requires: a battle-proven domain ontology, entity resolution across a client's whole estate, and a trust architecture regulated owners can actually sign. Point solutions solve fragments. We own the layer they'll all have to plug into.
We set the domain standard
A battle-proven AEC/O ontology already running in federal production, plus entity resolution across a client's whole estate. Horizontal document-AI has no domain model; generalists never connect documents into something usable.
We're the only one signable
Per-client isolation, on-prem for high-security clients, provenance on every fact. Insurers, federal owners and infrastructure operators require this. Generic LLM apps cannot credibly offer it — which removes them from the deal entirely.
We're first to the flywheel
Paid data compounds in-account; opt-in free usage improves Cue. Every correction widens a lead competitors can't buy back. This round funds that capture machinery — while it's still early and cheap to own.
Competitive landscape
Named competitors — what they do well, and where they stop.
| Company | What they do | Strengths | What they miss vs. QAECY |
|---|---|---|---|
| Structured AI | AI QA/QC on engineering drawing sets — automated compliance & clash detection | Best-in-class AEC-Bench score (75%); strong on 2D PDFs and Revit; YC-backed; SOC 2 + on-prem option; integrates with Procore / ACC | Drawing-review point solution — no governed knowledge graph, no cross-portfolio entity resolution, no asset operations intelligence layer |
| Nomic AI | Horizontal embedding + vector-search platform; Atlas for data map visualisation | General-purpose, fast to deploy; strong visualisation of unstructured data; open-source Atlas | No AEC/O domain model or ontology; no entity resolution; no provenance or audit trail; not built for the built environment's trust requirements |
| AEC Foundry | AEC-native AI platform consultancy — data pipelines, agentic workflows, FDE | Deep AEC domain expertise; platform-first methodology; publishes AECV-bench; strong on BIM/CAD/PDF ingestion pipelines | Services / FDE model rather than a product — no persistent governed index, no compounding knowledge graph, no scalable recurring licence model |
| Howie Systems | AI knowledge management & enterprise search for AEC / real estate teams | Natural-language search over project data; Vienna / Europe footprint (Pi Labs, Dar Ventures); architect-founded; customer data not used for retraining | Document search layer — no entity resolution or knowledge graph; no BIM/CAD-native pipelines; early-stage (<20 clients); no infrastructure / regulated-sector positioning |
| Constructable | All-in-one AI construction management — RFIs, submittals, drawings, financials | Broad project-lifecycle coverage; AI-powered search across project docs; single system from bid to closeout; BuiltWorlds Top 40 2026 | Construction project management tool — not an asset intelligence layer; no knowledge graph or entity resolution; no infrastructure / owner-operator / FM positioning |
Where others sit — and the gap
Adjacent players each solve part of the picture. None close the loop from raw documents to connected, AEC/O-native knowledge a client can trust and keep as its own.
| Category | What they do | What they miss |
|---|---|---|
| Horizontal document-AI / IDP | Extraction and RAG over PDFs | No AEC/O domain model; no cross-document connectivity |
| BIM / CDE platforms | Store and coordinate project files | Don't turn documents into AI-usable, connected knowledge |
| Generic embedding / RAG | Vector search over text | No domain ontology, no entity resolution, no provenance |
| Frontier-LLM apps / generic agents | General reasoning over your documents | No domain depth, no isolation or trust guarantees, no AEC/O workflow |
Larger players
Own project management and BIM authoring — not the AI intelligence layer. They control the file formats and project workflows QAECY plugs into. Potential acquirers, not competitors.
No governed graph, no entity resolution, no audit trail. Fast to deploy, fast to fail in a liability-averse industry. QAECY's Tuner and provenance layer exist precisely to replace these.
The market is validating the direction
A federal anchor, a strategic insurer, and a deep, qualified pipeline — on lean capital.
8-year CDE tender
Won with Zühlke and Vrame for Astra, Armasuisse and BBL. A multi-year, recurring federal mandate.
Helvetia
PoC completed; licence under negotiation, with an implementation scope well into seven figures. Decisive near-term.
Amberg Group
Pre-seed investor and anchor customer — the spin-off's first proving ground.
| Selected wins & priority pipeline | Type | Value |
|---|---|---|
| CDE Bund (federal CDE, recurring) | Won WTO tender· (Zühlke+vrame cooperatio) | CHF 740K |
| Airport Zurich (Requirement management) | Won public tender (Pilot planned) | CHF 145K |
| tennet (Integration cue as enterprise knowledge grapH9 | Direct contract (Pilot planned) | CHF 90K/m |
| Helvetia — Pilot and Implementation | Offered (Q3 2026) | 813K+ |
| Astra Rheintunnel (05.2025-05.2026) | AI Agent as archive assistant | CHF 3K/m |
| ETH Zürich · Swiss Post · tend | Priority pipeline | ~CHF 100K ea. |
| Qualified pipeline (total) | CHF 4.47M |
The built environment is the last frontier for AI
Real estate, infrastructure, energy, and construction together represent ~13% of global GDP — transitioning from analogue to digital at an accelerating rate driven by regulatory pressure and AI capability.
| Driver | Timeline | QAECY impact |
|---|---|---|
| EU AI Act enforcement | 2025–2026 phased rollout | Compliance urgency → inbound |
| Swiss FADP / DSGVO tightening | Ongoing | On-prem + privacy moat |
| CDE mandates (ISO 19650) | EU public sector rolling out | CDE Bund replicable across EU |
| Foundation-model commoditisation | GPT-4 → Claude → open models | Data layer becomes the moat |
| MCP protocol adoption | Anthropic standard, 2024–2025 | Distribution via existing AI tools |
CHF 709K from one person, zero paid marketing
All revenue so far was generated by Philipp through pure network and relationship selling — no SDR team, no channel budget. This is the baseline. The seed round funds the transition to a repeatable, scalable motion.
A 135× LTV:CAC at licence stage
A CHF 4,100 CAC generating a 135× LTV:CAC ratio on enterprise licences is rare at any stage. The 9-month sales cycle is the main friction point — and the one problem a dedicated sales hire plus VC network access directly addresses.
LTV based on CHF 222K/yr enterprise licence × 3-year average × 83% gross margin. The CHF 4,100 CAC is fully loaded — Philipp's time allocation, trade-fair costs, proposal preparation.
As the revenue mix shifts from FDE-heavy (50% GM) toward licence and subscription (83%+ GM), blended gross margin improves materially — a key driver of the Y3–Y5 financial profile.
The ask is lean because the engine is efficient
Year 1 returned roughly a franc of booked revenue for every franc deployed — rare for a platform at this stage. CHF 3M isn't a small ambition; it's what a team producing ~15 FTE of output actually needs to reach a CHF 10M ARR run-rate. A conventionally-staffed competitor would need three times the capital to attempt the same thing.
CHF 709K paid + CHF 61K invoiced
Booked revenue is realised cash plus invoiced work — not pipeline. The franc-for-franc headline is on a Year 1 basis.
~15 FTE of output at ~CHF 500K/yr
Modern AI tooling lets three engineers ship what conventionally takes fifteen. A traditionally-staffed equivalent costs ~CHF 3.5M/yr — a 7× structural cost advantage that compounds as the tooling improves.
Efficiency is the strategy — not a constraint on the ambition
The category opportunity is a CHF 10M+ ARR platform business sitting on a USD 3.2B European market, and the federal anchor plus the flywheel are what make that reachable. But raising CHF 10M today would buy headcount this team does not need and dilution the founders would rather not take. CHF 3M funds three years to a CHF 10M ARR run-rate — team from 4 to 8, on-prem for the regulated segment, and the capture machinery. If the pipeline converts faster than modelled, the right response is a larger Series A at a materially higher mark, not a bigger seed now.
CHF 3M to convert pipeline into recurring revenue
CHF 3M at CHF 27M pre-money / CHF 30M post-money. Deliberately sized to the team's efficiency, not to a headcount plan — three years of runway to a CHF 10M ARR run-rate.
| Post-round cap table | Ownership |
|---|---|
| Founders | 78% |
| New investor (seed) | 10% |
| ESOP pool | 9% |
| Amberg Group | 3% |
| Total | 100% |
Amberg Group loan
CHF 725K drawn of CHF 1.5M committed — the remaining CHF 775K is undrawn. Pre-Seed round is closed earlier to go into scaling. and focus on ARR.
Convert to roughly 3% post-round equity.
New investor buys Amberg's position as secondary.
Repay over three years from operating cash.
ESOP
An 9-10% pool with a 3-year cliff. The cliff is deliberate — not the market-standard one year — and chosen for long-term team commitment over the company's build phase.
A major Swiss institutional investor is in active discussion at CHF 47.5M pre-money — a 35× F-ARR valuation reflecting the specific strategic value QAECY delivers to their portfolio. We are not pursuing that route for this round. A VC partner at CHF 27M pre-money is a deliberately lower entry price in exchange for the international scaling capability, network, and visibility that capital alone cannot provide.
CHF 3M — three years of runway
Doubling the team, building the on-prem and open-source core, and standing up the free-tier MRR engine — without breaking the cost advantage.
If Amberg's loan is repaid rather than converted, repayment is spread over three years from operating cash — it does not come out of this raise.
8 people with AI tooling = ~40 FTE equivalent output. Team cost stays ~CHF 500K/yr — vs CHF 3.5M for a traditional equivalent.
What we want from a VC — beyond capital
The product is ready and the pipeline exists — Orsted (DK), Novo Nordisk (DK), VIE Airport (AT), Tennet (DE), Happold (UK). What QAECY needs is a partner who can open doors faster than cold outreach can.
A deliberately lower entry price (CHF 27M pre-money, vs a strategic investor's CHF 47.5M reference) in exchange for the international scaling capability, network, and visibility a hands-on VC partner brings. We are choosing the partner, not just the cheque.
Document index
Status of the materials supporting this round.
01 Corporate & legal
02 Financials
03 IP & technology
04 Commercial & pipeline
05 Team & governance
Let's talk
We are running a focused process and meeting a small number of aligned partners. If the fit is right, here is how it goes.
Figures as of June 2026. Forward-looking statements are illustrative and not a guarantee of future performance.