# Evaluate the Evidence Behind the AI

> Evaluate AI capabilities, model dependencies, data assets, security, and operating costs before an investment or acquisition with Auditsuisse.

- Canonical: https://auditsuisse.com/technical-due-diligence/ai-data
- Language: en
- Published: 2026-11-18
- Updated: 2026-11-18
- Alternates: fr → https://auditsuisse.com/fr/due-diligence-technologique/ia-donnees, es → https://auditsuisse.com/es/due-diligence-tecnologica/ia-datos

AI Products and Data Assets

Understand what an AI product does, how it is built, and what it depends on. Auditsuisse assesses AI capabilities, data pipelines, security, and operating economics to help investors and acquirers evaluate the technical foundations of the business.

At a glance

For

Investors and acquirers evaluating AI products and data assets.

We assess

AI capabilities, data pipelines, security, and operating economics.

Deliverables

Can include an AI capability map, model and data dependency analysis, a review of evaluation evidence, security findings, and an assessment of operating-cost assumptions.

## Connect the AI claim to the operating reality

An AI product may combine proprietary software, external models, customer data, retrieval systems, and human review. The value and risk depend on how those elements work together and whether they perform reliably for the intended use.

We examine the supporting architecture, evaluation evidence, workflows, and costs. The assessment distinguishes capabilities demonstrated in operation from prototypes, management assumptions, and planned development.

## What we assess

### Product capability and differentiation

The AI functions delivered to customers, the workflow they support, and the technical features claimed to distinguish the product. We identify which capabilities are developed internally and which rely on external providers.

### Models and provider dependencies

Model selection, fine-tuning where applicable, retrieval architecture, provider concentration, version changes, portability, and constraints affecting continued operation.

### Data assets and pipelines

Data sources, quality, lineage, access controls, processing, and retention. We identify technical dependencies and unresolved questions about permissions for legal review.

### Performance and evaluation

Test sets, evaluation methods, production monitoring, failure handling, and evidence of performance on the tasks customers actually perform. We consider whether claimed results are repeatable and relevant to intended use.

### AI security and control

Sensitive data exposure, prompt injection, access to connected systems, tool permissions, isolation, and human approval for consequential actions.

[Cybersecurity and IT due diligence](https://auditsuisse.com/technical-due-diligence/cybersecurity-it)

### Operating economics

Inference and hosting costs, usage assumptions, human review effort, latency, and sensitivity to changes in provider pricing or customer demand.

### Team and roadmap

Engineering and data capabilities, evaluation ownership, specialist dependencies, and the work required to move planned features into reliable production use.

[Software and code due diligence](https://auditsuisse.com/technical-due-diligence/software)

## Assess the technical basis for defensibility

An external model dependency does not by itself determine the strength of a business. Differentiation may come from data access, workflow design, integrations, performance, distribution, or operational knowledge.

We assess the technical evidence behind claimed advantages and identify dependencies that could weaken them. These findings support the broader commercial diligence undertaken by the transaction team.

## A global perspective on data-dependent businesses

International AI businesses may move data among customers, cloud regions, model providers, and development teams. We map the relevant technical flows and controls, highlight unresolved questions, and coordinate with counsel on matters requiring legal interpretation.

## Clear findings for your investment decision

An assessment can include an AI capability map, model and data dependency analysis, review of evaluation evidence, security findings, and an assessment of operating-cost assumptions. The report explains what was examined, what the evidence supports, and the priorities for further work.

## Common questions

### Can you assess a product built on third-party models?

Yes. The review can examine the application, workflows, data architecture, evaluation, security, and economics surrounding those models.

### Does the review include model training or specialist scientific validation?

The scope defines the models, technical methods, and validation work included. Specialized research, retraining, and extensive benchmarking require an expressly agreed workstream.

### Do you provide a legal opinion on training data or intellectual property?

We assess technical evidence and identify questions about data sources, dependencies, and permissions. Legal counsel determines rights and legal consequences.

Technical due diligence services are provided under an agreed scope and engagement terms. Coverage, evidence access, report use, and any assurance services are defined in that engagement. Where Auditsuisse also provides audit or attestation services, engagement acceptance and scope are subject to applicable independence requirements.

---
Source: https://auditsuisse.com/technical-due-diligence/ai-data · This Markdown edition is generated from the same content as the HTML page.
