Rasa Conversational AI Assistants for Business
Rasa is an open-source framework for building conversational AI assistants. See what businesses use it for and how MILabs sets it up.
Updated 2026-03-01
What Rasa is
Rasa is an open-source framework for building conversational assistants. These are programs that take part in a chat or a phone conversation, understand what a person is asking, and reply in a useful way. A Rasa assistant is built from a few parts. A language understanding component works out what the person wants. A dialogue component decides what the assistant should do next, based on examples and rules the team writes. Custom actions, usually written in Python, let the assistant look things up or pass a request to another system, then report back. Because Rasa is open source, it can be run on infrastructure the business controls, which some teams prefer when conversations contain private client details. It is a framework rather than a packaged chatbot, so the assistant is shaped around the questions a particular business receives and is maintained as those questions change. The team decides how the assistant reasons and where its conversation data is stored.
Rasa in plain language
Conversational AI assistants are software that answers people in plain language through a chat window, a messaging app or a phone line. Most of them fall into one of two groups. Some are packaged products where the vendor hosts everything and the business configures a few settings. Others are frameworks, which give a development team the building blocks to design the conversation themselves. Rasa belongs to the second group. Its appeal is control: the team decides how the assistant reasons about a request, what it does when a request falls outside its training, and where conversation data is stored.
For a non-technical owner, the practical question is less about the code and more about the shape of the work. Someone has to decide which questions the assistant should answer, which ones belong with a person, and what counts as a good answer. The software then turns those decisions into a working assistant that can be tested, corrected and extended over time.
Frameworks suit businesses whose questions change often. New services or policy changes can be reflected by updating the relevant answers rather than rebuilding the whole assistant. Conversation logs help show which topics need better answers and which should reach a person sooner.
What businesses use it for
- Answering routine questions about opening hours, services and prices on a website.
- Collecting details from new enquiries before a staff member replies.
- Taking booking requests and checking them against a calendar the business already uses.
- Helping staff find answers in internal policy documents and procedures.
- Handling messages that arrive when staff are away and passing urgent ones to the person on call.
Who it suits
- Clinics that receive the same questions about hours, referrals and forms every day
- Service businesses with an enquiry inbox that is slow to answer at busy times
- Teams with private client information that must stay on infrastructure they manage
How MILabs sets it up
The easiest first step is to say hello through the enquiry form. From there the owner or team lead decides which conversations the assistant handles and which go to a person. Together we choose how the assistant is hosted and agree what it may say for the business. We then connect the systems the business already uses, such as the website chat, enquiry inbox, booking calendar or customer database, so the assistant answers from real information. Permissions limit the assistant to the information and fields it needs to touch, and one client's details are never shown to another. Before going live, the team tests the assistant on the questions customers ask. We agree who looks after it, reviews logs and updates answers when prices, services or policies change.
Builds it fits
Industries it fits
Rasa in practice
In a typical business, a Rasa assistant sits between the public and the team. A visitor opens the website chat and asks whether a service is available on Thursday. The assistant checks the booking calendar through a custom action and either offers a slot or collects the details a staff member needs to follow up. When a question falls outside its design, the assistant says so and passes the conversation on rather than guessing.
Inside the business, the work is mostly about the handover. Staff define which topics the assistant owns, which require a person, and which details to capture before a handover. Those decisions go into the assistant's flows and examples, then get tested with real customer messages, including badly phrased ones and questions that mix two topics.
After launch, the rhythm is steady and largely routine. Someone reviews a sample of conversations, flags answers that were wrong or unclear, and the team updates the relevant part. A new price list or a changed policy means editing the specific answers it touches, then retesting those before release. Internal use follows a similar pattern, with staff querying a knowledge assistant built on procedures the business already keeps. Named people stay responsible, so the assistant reflects decisions the business has made.
Frequently asked questions
- Does Rasa need technical skills to run?
- Yes, for the build. Rasa is a developer-oriented framework, and its assistants use configuration files and Python code for custom actions, so a team member or outside developer needs to be comfortable with that. MILabs handles the build, the connections to existing systems and the testing, and agrees with the owner who makes routine edits afterwards. Staff who answer enquiries need no code to review conversations or flag answers.
- Can Rasa connect to the software I already use?
- Yes. A Rasa assistant can reach other systems through its custom actions, which are written in Python and can call an API, a database or a booking calendar. Which channels it reaches, such as a website chat or a phone line, depends on how the assistant is deployed. MILabs reviews the systems the business uses first, and each connection is confirmed before it is built.
- How does Rasa compare with other tools for the same job?
- Rasa is a framework for building an assistant that the business hosts and shapes itself. Botpress and Typebot are visual builders that suit simpler flows, while Dify and LangGraph suit teams building retrieval or agent logic. The right choice depends on who will maintain the assistant, how much hosting control the business needs, and how often the questions change. MILabs weighs these points with the owner before recommending one, as no tool fits every business.
- How do I get my team to use Rasa?
- Staff adoption depends on whether the assistant saves time they already spend. MILabs starts with the questions the team answers most often and builds around them, so staff see fewer repeated messages rather than a new screen to learn. We agree who reviews conversation logs and how often, set up a way for staff to flag a wrong answer, and hand conversations to a person when the assistant is unsure.
- Can I self-host Rasa?
- Yes. Rasa is open source and can be installed and run on servers the business controls, which some teams choose for conversations that contain private details. Self-hosting means the business is responsible for the server, its security updates and its backups, and MILabs can set this up and agree those duties before launch. If a business would rather not manage servers, we discuss other hosting options during planning.
