Key takeaways
- Most AI features sit next to your existing system and talk to it through an API. The core system barely changes.
- Start with AI search or an editor assistant: visible value, low risk.
- Rebuild only the parts that block growth, in small releases.
Why a rebuild is usually the wrong first move
Systems built before AI still run the business: the CMS with ten years of content, the custom PHP app with every edge case baked in. Rewriting them to "add AI" puts all of that at risk for a feature you could ship in weeks.
The better approach is to leave the core in place, add a thin layer that lets AI read from and act on it, and modernize the pieces that actually hold you back.
Five patterns that work
- AI search over your content. Index pages, documents and products so visitors can ask questions in plain language and get answers with links. This uses RAG and works with WordPress, Drupal and most custom sites.
- Editor assistants. Help editors draft summaries, meta descriptions, alt text and product descriptions inside the admin screens they already use, always with a human approving.
- Translation workflows. Generate first-draft translations, such as English to Spanish, that editors review, instead of starting from scratch.
- An API layer for old systems. Wrap a legacy database or app in a small, secure API so modern apps, AI agents and MCP servers can use it safely.
- Automation on top. Once the API exists, agents can handle intake forms, update records or route requests without anyone touching the old UI.
Platform notes
- WordPress: add features through a custom plugin that calls your AI service; avoid putting API keys in the browser.
- Drupal: use custom modules and Drupal's entity and permission system so AI respects existing access rules.
- Shopify and e-commerce: use the platform's APIs for product data and keep AI generation in a separate service with review steps.
- Custom PHP, .NET or Java apps: add a read-only API first, then carefully chosen write endpoints.
A safe, step-by-step plan
- Assess code, data, hosting and integrations. Find what's fragile before adding load.
- Connect AI through APIs without disrupting daily operations.
- Modernize the parts that limit growth, one small release at a time.
- Operate with monitoring and a clear roadmap.
This is exactly how our legacy modernization + AI service works. If a larger rebuild does make sense, our app development team can take it from there.