Routines: How-to Guide
Here are tips to make sure that Routines are integrated in the best way possible.
For initial estimate of the effort necessary for the routine development and guidance you can also review Routine preparation document.
1. Decide for solutions
Decide for the solutions that you want to automate.
It’s a good case for automation if
- the backend system already holds the information that should be displayed to the user
- an agent can trigger the action with 1 click (no extra review required)
- the functionality already exists in a self-service portal/app
To ensure correct and long lasting performance of the implemented routine it is not recommend to hold the business logic in the routine. E.g. retrieving information from 2 different API endpoints and determining final state in the routine.
It is however possible to connect to one or multiple external systems like backends, databases or third-party software within same routine.
Instead it is recommended that the third-party system already has the final correct information. Some data manipulation is possible (formatting date, time, adjusting texts).
2. Consider the flow from user perspective
This consists of 2 parts:
User enters information in the bot.
This is the information collected from users in the bot before executing routine. Depending on the use case it be one or multiple fields that are required to fetch information and/or perform customer authentication.
User receives information
Decide how users will receive the answer and how many details to display. Depending on the sensitivity of the information and the customer authentication performed. The response can either be displayed directly in the bot and/or trigger out via other channels (e.g. send an invoice via email).
3. Verify internally
Internal verification (usually requires assistance from the tech department) is an important factor in insuring that desired result can be achieved.
Depending on the design of the systems and authorization/authentication required the solutions may vary vastly.
The more of this information that can be gathered the faster and more efficient routines can be put live. Do not hesitate to approach us for advice.
4. Develop Code
Develop the right code, that will allow the bot to interact with third-party system. Code can be written in Python or Node.js.
If Solvemate is developing your code - check out Let Solvemate build your routine.
Code can be developed internally or simply approach your Success Manager at Solvemate and let us help you.
Format the routine message
Same formatting rules apply for routine texts as for the solutions - you can use bold text, emphasis, lists, links and emoji.
5. Add this code to the Solvemate interface
Create a new component in the Workspace - Routine, add the code, set up input fields (and their validation), add the secrets, customize texts, test routine with test data and save it.
6. Link the right solution to the right Routine
Connect routine with the solution.
7. Publish and monitor
Publish your routine and follow up with the performance to iterate on improvements.
Let Solvemate build your routine
Solvemate is happy to help in creating routines, developing custom code and integrating it.
Depending on the contract type this service might already be included. Get in touch with your Customer Success Manager to find out more.
The routine development process is iterative and handled in several stages:
- You will receive a Routine preparation document that needs to be filled in.
This will require input from Customer support (e.g. define use cases, prepare texts) and IT (e.g. share access and information about systems and their setup, prepare test data) departments.
2. Development and Testing
After reviewing the preparation document Solvemate will arrange an alignment call with you to define responsibilities, timeline, and answer open questions.
Once open questions are clarified the code is being developed by Solvemate and testing is performed by the responsible party as agreed.
Depending on your backend system and the business logic the routine(s) will process additional alignement calls might be needed.
Final testing is performed by you to confirm that the routine is ready to Go Live.
3. Go Live
Solvemate will set up the routine components in your account within the Solvemate WebApp.
Add routine to the right solution, polish final texts and publish.
Improve routine based on performance/user feedback together with your Success Manager.
Cover these questions
Once it is decided which solutions can be automated these are the recommended questions to cover in order to implement a routine.
For initial estimate of the effort necessary for routine development and guidance review Routine preparation document.
In the next tabs (Access, User Input, Result, Sharing, Security) you can find overview of recommended questions that should be reviewed before moving forward with the routine development.
- Is the necessary information available in your own backend or outsourced third-party software?
- In case of your own backend - are developer resources available for chatbot integration project?
- In case of third-party software - is API with documentation available?
- What are the authorization/authentication methods used to access information in the systems?
- Is it possible for third-party (Solvemate) to request and consume information?
- Is the authentication method suitable to be used with chatbot? Each executed routine uses its own authorization/authentication.
- Which information will be required from the user? Mandatory and optional.
- Which information will be used to retrieve results and which for customer authentication (where applicable)?
- Is the required information always in a specific format (e.g. OrderID starting with M followed by 6 digits M123456)?
- This is important to know how to use the right type input fields and/or add RegEx to prevent users from entering invalid information.
- What information and details should be displayed? Ideally, you draft sample text of a standard case.
- Will the result be displayed in bot and/or communicated via other channels (e.g. email)?
- What language should information be displayed in?
- is the third-party system already providing texts in the desired language?
- Are there any edge cases within the system for the topic? How should the bot handle those?
- Texts for errors and edge cases.
In case Solvemate is developing the routine, access to backend systems with data is required.
- Can you create and/or share account access with Solvemate for the system? For most cases read-only access is enough.
- Do you have test data that can be used for developing routine?
- Do you have test data to cover most cases, ideally also edge cases?
Review our Security page. It is sufficient for most use cases.
- Do you have any specific security requirements that are not covered?
Approach your manager at Solvemate to clarify any concerns.