Project: AMO, AI Mortgage consultant and loan processing platform

SolutionsProject

Ongoing project, details will be updated when appropriate.

In this blog we will discuss how ASSV helps to solve challenging technical problems while saving time and money for the client along the way.

The problems

In order to solve problems, it's crucial to identify and agree what the problems are. ASSV team takes time to collect data, explore the business on premise and come up with agreed problems:

  1. Customer consultancy is a repeatable task and it's hard to train a good salesperson.
  2. Existing workflows can be optimized with custom processing tools.

The solutions

Problem #1

After analyzing a week worth of the business best salesperson script, we discover three prominent patterns on every sale talk:

  1. Questions & answers about common business knowledge. Example: What is A? Why is B high, C is low? Etc. These kinds of questions are frequently asked questions (FAQs) and can be answered directly.
  2. Analytical questions requires user's specific data. These questions require salesperson to collect a set of data from user and respond with computed metrics so user can make a more economically decision.
  3. Colloquials. These are human's casual everyday communication like: How are you? How's the weather today? Etc.

Solution: We conduct research on existing technologies and select Google Diaglogflow as a natural language understanding platform. A chat bot framework is designed and implemented on top of this platform so we can quickly address #2 whenever we add a new capability to the bot. We utilize Diaglogflow "knowledge base" feature to address #1 and "small talk" feature to address #3.

Result: Bot can answer popular 80% of the cases compared to a salesperson. Bot can be deployed on various platforms (chat platforms, phone, voice platforms) quickly. With iterations to cover the long tail conversations, bot can replace salespersons to consult clients in a foreseeable future.

Problem #2

Legacy software used within the business no longer supports the fast pace of productivity. We rewrite the software stack cloud-based style. We also automate a larger portion of the workflow, cutting processing time down to 70% compared with legacy workflow.

Some technologies used

  • Google Cloud Services: Diaglogflow, Firebase, Identity.
  • Amazon AWS: EKS, S3, SES, SMS.
  • Facebook: Messenger, Identity.
  • Apple: Identity.
  • Github: CI, CD, Container Registry.
  • Backend: Node.js, NATS, MongoB, Redis, Microservices, Docker, Compose.
  • Web: React.
  • Mobile: React Native.
  • Deployment: Kubernetes.

Conclusion

We achieve some important objectives:

  • Bot will gradually and eventually replace the best salesperson.
  • (Dramatically) cut cost in recruiting, training, providing salesperson & consultancy.
  • Cut down 70% business processing time compared to legacy systems.
  • Much more affordable software development.
  • Business owners only need to spend minimal time on the software project.