It's a fair question. Here's an honest answer.
We recently asked an AI assistant (Claude) to build a tool that exports HCL Notes/Domino data to a standalone HTML website. Within minutes, it produced working code: a LotusScript agent to extract documents, a Python script to convert the output to JSON, and a polished HTML generator with navigation, search, and a clean responsive layout.
It was genuinely impressive. And it raises an obvious question for anyone planning a Notes migration: if AI can write this code in minutes, why would you pay for a tool like Teamstudio Export?
The answer says a lot about where AI is useful — and where it quietly falls apart.
The easy 20% looks like the whole job
The code that Claude produced works. Run it against a simple test database with clean text fields and you'll get a reasonable HTML site. It's a great demo. The problem is that real Notes databases aren't simple, and the script has no way to tell you what it's getting wrong.
Here's what the AI-generated script silently skips or breaks:
Rich text. Notes rich text is a proprietary binary format. It contains formatted content, inline images, embedded tables, doclinks, and attachments — and it's present in the vast majority of real Notes databases. The script simply ignores it, although Claude did offer to write a rudimentary rich-text-to-text converter. Your migrated documents look complete until someone notices the body of every memo is blank, or missing key parts.
Attachments. Files embedded in Notes documents don't exist in the exported output. They're gone. This is often a compliance or legal problem, not just an inconvenience.
Multi-value fields. Notes fields can hold a list of values — multiple authors, multiple categories, multiple dates. The script collapses them unpredictably. You may not notice until someone searches for a document that used to be there.
Doclinks. Notes documents link to other Notes documents, sometimes in other databases. Those links need to be resolved and rewritten as HTML anchors pointing to the correct exported page. The script has no concept of this. Every doclink in your data becomes a dead end.
Views and folders. Notes databases are navigated through views — sorted, categorized, filtered collections of documents. The script produces a flat dump. The organisational structure that made the data usable is gone.
Scale. A LotusScript agent running against a 500,000-document production database will time out, run out of memory, or fail silently when it hits a malformed value in a single field. There is no retry logic, no checkpointing, no progress reporting, and no way to know whether the output is complete.
The real cost is the edge cases you don't know about
An experienced Notes developer will read that list and recognize most of it. Someone who isn't a Notes specialist — which describes most of the people responsible for Notes migrations today — will run the script, see it produce output, and assume the job is done.
That's the deeper problem. AI-generated code is confident. It doesn't caveat its own limitations. It doesn't warn you that it skipped the rich text, or that 12% of your documents failed silently, or that the three databases in your estate that use a non-standard attachment format will produce corrupted output. You find out later, often after the original Notes environment has been decommissioned.
Teamstudio Export has been built and refined against the full range of what Notes databases actually contain — not a representative sample or a clean demo database, but the messy, inconsistent, years-old production data that real organizations need to migrate. The edge cases that would break a handwritten script are exactly what the product was built to handle.
Migration is not a code problem
There's a more fundamental point here. Notes migration isn't primarily a technical challenge — it's a governance challenge. Organizations migrating away from Notes need to be able to demonstrate that the process was complete, accurate, and auditable.
An AI-generated script cannot be audited. It cannot be supported. It cannot be validated by a compliance team or signed off by a regulator. If something goes wrong six months after the migration — a document that can't be found, an attachment that wasn't preserved, a legal hold that wasn't honored — there is no vendor, no support contract, and no documented methodology to point to.
Teamstudio Export provides all of that. The output format is documented. The process is repeatable. There is a support organization behind it. When a customer needs to demonstrate to their legal team that the migration was done properly, they have something concrete to show.
So what is AI actually good for here?
Quite a lot, as it turns out — just not as a replacement for the migration tool.
AI is useful for understanding the problem space: exploring what a Notes database contains, prototyping approaches, writing the glue code that connects systems. It's useful for building the downstream tooling — the ingest pipelines, the search interfaces, the reporting dashboards — that consume the output Teamstudio Export produces. And as the exchange that prompted this post shows, it's useful for demonstrating the complexity of the problem to people who might otherwise assume it's simple.
That last point is worth sitting with. Someone who asks "can't AI just do this?" and then watches an AI-generated script fail on rich text, attachments, and doclinks has just received the clearest possible demonstration of why a production-grade tool exists. The question answers itself.
Postscript: a confession
As if to underscore the previous section of this post, after we’d finished experimenting with Claude, we asked it to write a blog post about its findings. Everything in this post prior to this final section was generated by Claude itself. Something else that AI is pretty good at!
Teamstudio Export is built for organizations migrating away from HCL Notes and Domino. It handles the full complexity of Notes data — rich text, attachments, doclinks, views, and scale — and produces output that meets compliance and legal requirements.