AI-EraTech Due Diligence

Performing Due Diligence for portfolio companies of VCs and funds - not just technology, but organizational readiness and potential as well.

The question we answer

How does AI affect the portfolio - does it erode the moat or increase the potential?

import { verify } from "@/auth";export async function sync(req) { const token = req.headers; if (!token) return 401; const claim = await parse(req); const tx = await db.find(id); return res.json({ ok: tx });git commit -m "fix: billing sync"41 async settleClaim(doc) {42 const s = await ai(doc);43 if (s.score < 0.4) {44 const q = `SELECT * FROM45 await db.raw(q + doc.id);46 // Co-authored-by:47 // Claude <noreply@...>48 return pay(doc.iban);49 } // TODO: verify source50 }

AI-era Due Diligence

In response to the exponential growth of AI, we have redesigned our due diligence service. When analyzing a startup risk, it is no longer enough to simply look at technological and organizational factors. The application of complex AI models and agentic development methodologies can replicate decades of development experience with measurable resources in days. However, AI-developed products can be very misleading - with vulnerabilities, compliance, legal, technical and functional problems layered on top, which may not be apparent to the untrained eyes.

What verticals we screen?

Due diligence report, risk profile

Risk register, top items

Constant firefighting due to low regression test coverageLo/Hi

Knowledge silos: dependency on one single lead engineerHi/Hi

Unstructured task prioritization leading to overloadingHi/Hi

Human under-performance due to missing KPIsLo/Lo

Data leakage from collaboration platform misuseLo/Hi

AI-ready data assets not used in AI-supported critical processesHi/Lo

High AI-repliability of critical value chain componentsHi/Hi

Score per area from 0 to 5 compared with a typical Series A benchmark, and the top items from the risk register.

Our methodology

Personal interviews

We conduct in-depth interviews with the organization's leaders, key personnel and random employees to map the organizational culture, health, processes, obstacles and cross-validate.

IT Due Diligence

We examine the software systems, codebase, development methodology, and QA processes, identify vulnerabilities, scalability and operational obstacles, and apply AI where appropriate.

Process Analysis

We analyse the strategic planning, collaboration, project management, internal communication and external partner relationships, and carry out a detailed risk assessment.

AI Replication Study

We model the identified moat, replicate it from scratch with AI, and compare it across multiple dimensions with the source. This reveals where the "real moat" lies.

Use case · Insurtech Startup

Example analysis · Document fraud detection, Series A target2026

Series A, June 2026: €8.2M
clients: 70+
countries: ~15
revenue growth in 2025: 3×

Situation

The startup flags forged documents (IDs, invoices, payslips, bank details) through a real-time API during onboarding, claims and credit decisions. Its clients include MetLife and Allianz Direct.

What we examined

We mapped the moat, then modelled how an AI-native competitor would replicate it. From that we derived what a due diligence has to test.

Key finding

The value sits in the claim document fingerprint network; generative AI will commoditise per-document forensics.

Moat and USPs

Cross-client fingerprint databaseFlags a document that was already submitted to another insurer. A real network effect, but only for exact or near-exact reuse.Strong
Integrations at large clientsEmbedded in claims and onboarding flows, so switching is costly.Medium
Privacy-preserving designStores fingerprints, never document content. A sales advantage under GDPR.Medium
Forensic expertiseFile-structure and metadata analysis. Preventive defence against generative-AI forgeries and against documents that are printed and then photographed.Weak

How would an AI-native challenger compete?

The startup's model asks whether a claim document looks forged? An AI-native challenger tries to validate the claim facts by confirming them at various sources (banks, invoices, communications) and keeps docuemnt forensics as a fallback.

Product · Claims evidence verification

01
Integrations
Beneficiary IBAN verification, open banking for bank statements, e-invoicing networks for B2B invoices, and the EU Digital Identity Wallet for identity.
02
Evidence graph
Claimants, documents, bank accounts and claims-handling locations linked in an annotated graph, so fraud rings become visible.
03
Similarity search
AI-based similarity search using optimised embeddings and perceptual hashes, which also identifies near-identical and regenerated documents.
04
AI investigation agent
Compares the claim description with weather, on-site and claims-adjuster data, generates an evidence checklist and prepares a sourced case file for decision support (GDPR Article 22).
05
Fallback document analysis
Where integration points are scarce, we carry out the fallback check by categorising the documents and scoring them with AI.
06
Sales
Marketplace apps on claims platforms (Guidewire, Duck Creek) and third-party claims administrators, without a long integration project.

* Due to confidentiality, we cannot publish client data.

Next step

Talk to us about due diligence

Book a call

Other services

01AI-Ready IT 02ISO preparedness 04Interim leadership & consulting