Compare
Merlean vs chatbots
A chatbot retrieves. Merlean executes. They look similar because both present a text box, and they solve different halves of the same user problem.
What chatbots do well
A well-built retrieval assistant over good documentation is genuinely valuable. It answers "what does this setting do", "is this included in our plan", "why did this calculation come out that way" - questions where the user wants to understand something, and where a correct paragraph is exactly the right output.
For a product with substantial documentation and users who read it, that is a real reduction in support load and a real improvement in the experience.
The ceiling
The limit shows up on the other kind of question. "How do I create a policy from this quote" does not want a paragraph, it wants a policy. A chatbot answers with steps, and the user still has to perform them, and some proportion of them get stuck partway and open the ticket the chatbot was meant to prevent.
Instructions also age badly. They describe an interface, and the interface changes, so a confident answer pointing at a button that moved two releases ago is worse than no answer. Every documentation-based assistant inherits that decay.
This is why deflection rates plateau. The questions that can be answered with text are being answered. The rest were never text problems.
Executing instead of describing
Merlean turns the request into the outcome. It resolves which quote is meant, calls the endpoints in the order the application requires, applies the permissions attached to that user, and reports what changed. There are no steps to follow, so there is no step to fail on.
And because the model is derived from current source rather than from written documentation, the action reflects how the application works now. The staleness problem does not arise in the same form.
Chatbot
Retrieves and explains. Output is text. Best for understanding. Decays as documentation ages behind the product.
Merlean
Executes. Output is a change in your application. Best for doing. Follows the source, so it tracks the product.
Together
A common and sensible arrangement: retrieval for the questions that want an explanation, execution for the ones that want a result.
Frequently asked questions
Should we replace our chatbot with Merlean?
Usually not. If yours answers conceptual questions well, keep it. Merlean addresses the requests it structurally cannot serve, which are the ones asking for something to be done.
Can Merlean answer questions too?
It can read from the application to answer questions about state - what is in an account, what happened to a record. Explaining concepts is what a documentation assistant is for.
Is this just a chatbot with tools attached?
The difference is where the tools come from. Hand-written tool definitions drift from the application and cover a fraction of it. Merlean derives the executable surface from the source and regenerates it when the source changes.
See it against your own application
Merlean reads your repository and shows you what it understood before anything is exposed to a user. The fastest way to judge the fit is to point it at a real codebase.
Related
Cut support tickets by doing the task, not explaining it
Most support tickets are not bugs, they are "how do I". Merlean completes the task inside your product instead of documenting it, removing the reason to open a ticket.
Add an AI assistant to your web app that actually does the work
Embed an AI assistant in your SaaS product that completes tasks instead of explaining them. One script tag, no per-feature integration code, works on any web stack.
Merlean vs in-product copilots
Most product copilots suggest, summarise and guide. Merlean completes the workflow. What separates a copilot that assists from a layer that executes.