Skip to content
CodeDD Logo

Case study · Big 4 advisory · Technology & AI due diligence

AI Claims, Verified in Code.

Code-level evidence behind an AI valuation premium before IC approval.

Client
Big 4 advisory firm
Target
Mature B2B software platform
Use case
AI-native technology & IP diligence
Feature
AI-Native assessment
  • AI-Native score

    45/100

    AI as Product tier, evidenced repository by repository

  • Codebase reviewed

    2.5MLoC

    across 12 repositories, 5 carrying genuine AI signal

  • AI integrated

    27%

    of code with real model integration; the rest carries no AI signal

  • Evidenced

    100%

    of AI claims mapped to source code, not management interviews alone

The challenge

Positive interviews. No code-level evidence.

The advisory team was supporting a private equity client on a significant minority investment in a mature B2B software platform. Newly launched AI capabilities were a core part of the equity story, and the investment committee needed those claims verified before the deal closed.

Management interviews supported the AI narrative, but the advisory team declined to rest the IC's comfort on conversations alone. The claims needed evidence across the software stack and the codebase itself.

  1. 01

    Where does AI actually sit in the platform, and is it genuine integration or a thin wrapper?

  2. 02

    Does the depth of AI investment match what management described to the IC?

  3. 03

    Which parts of the platform carry no AI signal at all, and does that match the stated rollout scope?

The solution

Every AI claim mapped against the codebase.

Working alongside the advisory team, CodeDD's AI-Native assessment located where AI sits in the platform, classified genuine integration versus wrapper use, and assessed platform support for the roadmap.

  1. 01

    Scored AI-Native maturity

    Every repository scored across product embedding, model integration, retrieval and data plane, adoption velocity, MLOps and evaluation, and platform readiness.

  2. 02

    Classified integration depth

    AI technologies fingerprinted per repository, from LLM frameworks down to ownership of models versus wrapper use, tiered by IP depth.

  3. 03

    Isolated the AI-integrated code

    Machine-learning lines of code measured against total lines of code, repository by repository, to size the real AI footprint.

  4. 04

    Matched claims to rollout scope

    AI signal mapped against management's stated rollout, confirming where the platform matched the story and where it did not.

codedd.ai / AI-native assessment

Product screen

2.5M LoC · 12 repos analysed · 5 with AI signals

45/100

AI as Product
Product Embedding100/100
Model Integration72/100
Retrieval / Data Plane58/100
Adoption Velocity46/100
MLOps & Evaluation18/100
Platform Readiness12/100

codedd.ai / repository signals

Repository signals

12 repositories analysed · sorted by AI-Native score
RepositoryLoCML LoCAI technologiesTierUsageIP depthScore
Repo 01341K74K · 22%LLM Framework · langchainT1AI as ProductOwn models52/100
Repo 02198K31K · 16%Vector DB · pgvectorT1Agentic systemIntegrator44/100
Repo 03126K9K · 7%LLM API · OpenAIT2WrapperWrapper21/100

+9 additional repositories

The results

AI claims substantiated, and bounded, in time for the IC.

  • AI claims substantiated own models and real integration confirmed in the core modules of the platform.

  • Rest of platform: no AI signal matching the rollout scope management described, no gap between story and code.

  • 27% of the codebase carries genuine AI integration; the remainder shows no AI signal at all.

  • Findings evidenced at code level not interviews alone, mapped repository by repository against the equity story.

  • IC comfortable the minority investment completed on code-evidenced findings.

“Interviews gave us a story. CodeDD gave us the evidence to put that story in front of the investment committee.”

Director, Technology & AI Due Diligence, Big 4 advisory firm

Put your target's AI claims to the test

See what your next diligence could tell you.