8 tips for developing AI assistants

AI assistants for businesses are hype, and many teams were already eagerly and enthusiastically working on their implementation. Unfortunately, however, we have seen that many teams we have observed in Europe and the US have failed at the task. You can read why they failed here:

→ How AI assistants are often developed, what they fail at and what they can learn from it

8 practical tips for implementing AI assistants

Business analysis and prioritization

Start by asking business-related questions. Identify steps in your processes that can potentially be automated, taking inspiration from projects that have already been successfully implemented. Focus on the easiest and most obvious areas for quick improvement and success (the so-called "low hanging fruit").

Stakeholder involvement

Talk to those involved in the process to gain insight into how things really work. Look for specific, repeatable patterns. ChatGPT/LLMs are good at working with patterns.

Definition of quality criteria

Define a quality metric before you write the first line of code. For example, develop a list of 40 questions that the system should be able to answer correctly. Identify these correct answers in advance. This will be your first assessment dataset!

Prototype-
Development

Give the developers a list of 20 questions with answers and ask them to create a prototype. Don't ask for a polished user interface at this stage; that would just be a waste of time.

Pattern recognition and pipeline development

Ideally, the developers bundle the questions into categories. They then develop a pipeline that prepares documents in such a way that a ChatGPT prompt can generate a correct answer for this question category.

Quality inspection and customization

Once the developers are satisfied with the quality, test the system prototype with the remaining 20 questions. This will be a wake-up call for many. Count the number of errors.

Iterative
Improvement

Give the developers some time to adapt the system so that it can handle these 20 questions. Measure the time they needed for the customization.

Extension of the test phase

Give the developers 20 more questions that they have not yet seen. Measure the time they now need for the corrections and also count the number of errors after the iteration.

Repeat the last step a few times without investing in a complex user interface and user experience.

Interested in using AI assistants for your company?

We look forward to hearing from you!

mail: christoph.hasenzagl@trustbit.tech

tel: +43 664 88454881