We took part in this year’s Microsoft SharePoint Hackathon 2026 with our submission SharePoint Change Explorer. We entered into two categories at once: Agents and SharePoint and SharePoint Framework.
- Official submission: SharePoint Change Explorer (Issue #209)
- Project video: YouTube – SharePoint Change Explorer
For us, the hackathon was less about the competition and more about an opportunity to sharpen an idea we had been chewing on for a while: how do you finally make SharePoint’s version history actually review-ready?
What it is about
For more than two decades, SharePoint has reliably stored document versions. But in many business processes, that is not enough. Reviewers still have to open versions one by one, read them manually, identify the differences, and decide what needs attention before approval. That is exactly the gap SharePoint Change Explorer closes.
The solution compares two versions of a document directly inside the SharePoint library, extracts structured content from the real file versions, computes a deterministic diff, and uses Azure OpenAI for one job only: explaining the changes in the language of the document.
The guiding principle is intentionally simple: Deterministic comparison proves, AI explains. Every sentence the AI emits is traceable back to a concrete diff entry – the AI never invents changes.
What reviewers actually get
The output is shaped to fit a SharePoint review flow, so approval decisions become faster and easier to document:
- Deterministic Diff Engine – character- and structure-level comparison across Word, PDF and structured documents, driven by Azure Document Intelligence.
- Business-Language Explanations – Azure OpenAI rewrites each grounded diff into plain business language, grouped by topic instead of by paragraph.
- Risk & Impact Indicators – each change group is scored (legal / financial / operational) so reviewers triage what actually matters first.
- Multilingual, SharePoint-Native – runs as an SPFx extension in the version panel; summaries in German, English and customer-language – no context switch.
Grounded AI, not generative guesswork
A generic AI tool that simply reads two documents and “summarises” them is unusable in enterprise review: it can hallucinate diffs, lives outside SharePoint, and leaves no auditable trail. SharePoint Change Explorer flips that around.
We deliberately separate truth from explanation. SharePoint versions are the source of truth. The diff is computed deterministically. Only after that does the AI layer step in and produce an explanation that is grounded exclusively in this diff. Every sentence links back to the exact diff – which is what makes the output signable for legal, compliance, and regulated environments in the first place.
Architecture
The MVP stack:
- SPFx 1.22.2 – library command set / extension inside the SharePoint version panel
- Azure Functions v4 on .NET 9 isolated worker for the diff and enrichment pipeline
- Azure Document Intelligence for OCR, table and layout extraction
- Azure OpenAI for the explanation layer (GPT-4o-class, EU region)
- PnP Core as the default SharePoint access path, with Microsoft Graph used only where it is technically required (such as change notifications or delta queries)
- Terraform as the infrastructure-as-code baseline – the entire stack is deployed into the customer’s own Azure tenant; no document leaves the tenant
In the MVP deployment, the compare endpoint ran live in our hackathon environment against a real SharePoint site and compared real PDF versions end to end.
Demo
The demo video walks through a typical review case: Compare is started straight from the library, the deterministic diff highlights the structural differences, and the AI layer explains the business meaning of the change – matching the language of the document.
You can watch the full video here: SharePoint Change Explorer on YouTube.
What’s next
The hackathon is over – the product is not. We keep building SharePoint Change Explorer based on what we see in real SharePoint libraries at our customers. If version compare, review briefings or audit trails for regulated document processes are on your radar, or you’d like to see the Change Explorer applied to your own libraries, get in touch – we are happy to walk through it in detail.
