Saving $162,800 for the client
with AI-driven development

Custom web app

Business need

Electricians, plumbers, and other tradespeople can do a quarter of  work using video calls. Yet, they still do those jobs in person, as there was no way to charge customers for such calls reliably.

Result

A single Solution Architect created the app from scratch using our in-house AI framework. The LLMs handled over 90% of the project tasks. However, getting production-ready code still required expertise in cloud and full-stack development.

Industry
Professional services
Location
USA
USA
Working together since
2023

Our client is a Silicon Valley startup that helps technicians earn more using a remote service call app.

The challenge

The app required a unique combination of API and custom logic so that technicians could charge for video consultations. Users also needed to recreate calls without duplicating payments. The estimated budget grew to $200,000 with two full-stack developers, a QA engineer, a Product Manager, and a UI/UX designer. We had to find another way.

The solution

Service Call app allows preliminary estimations and easy fixes over the phone. Technicians can log in with SMS, share video calls, chat with customers, and view analytics. The entire app was built by one engineer and our proprietary AI framework. They designed the UI/UX and database architecture, core backend functionality, Docker Compose CI/CD, and automated testing.

Unique set of integrations

Service Call is the only known app to use Daily.co SDK for WebRTC video chat with an on-brand design; Stripe API for customer authentication and secure payments; and Twilio API for SMS messaging.

This was the first major challenge we had to overcome using AI-driven development.

Correct sequencing of events and code output

The second challenge for the AI was understanding complex, iterative user flows. For example, technicians should be able to easily create service calls for customers to join using credit card details. However, customers should only be charged when their problem is fully resolved. And this might take multiple calls with different technicians.

Correct sequencing of events and code output

The second challenge for the AI was understanding complex, iterative user flows. For example, technicians should be able to easily create service calls for customers to join using credit card details. However, customers should only be charged when their problem is fully resolved. And this might take multiple calls with different technicians.

AI doesn't do well with ambiguity. Engineers have to describe user flows and edge cases in great detail. This means creating the right set of prompts, database structures, and sequence of transactions.

Getting the database right

Database architecture was another challenge for the AI. Here's an example. Technicians are a part of their company. All clients are unique within the company. However, clients can have multiple meetings with different statuses within one payment.

Getting the database right

Database architecture was another challenge for the AI. Here's an example. Technicians are a part of their company. All clients are unique within the company. However, clients can have multiple meetings with different statuses within one payment.

It's especially challenging to use a chat interface to validate database structure and connections (one to many, many to many, uniqueness) as well as work with schema (text presentation of database relations).

Distinct user type flows

Technicians and their employers needed two distinct login mechanisms and permission sets. Twilio integration allows technicians to authorize using a mobile phone. The app sends a one-time password via SMS and locks it on the application window where the technician wants to authorize.

Distinct user type flows

Technicians and their employers needed two distinct login mechanisms and permission sets. Twilio integration allows technicians to authorize using a mobile phone. The app sends a one-time password via SMS and locks it on the application window where the technician wants to authorize.

Employers, on the other hand, sign up with a corporate email. They have permissions to change the company info, add/remove technicians, and view company-level analytics.

Call status tracking

Each service call has a different status, including in-progress, successful, charged, and failed. An invitation link expires within 10 minutes. You can recreate failed calls as many times as needed without the client having to re-enter credit card details.

This is just one of the edge cases our engineer had to consider before letting the AI write the code.

LLMs ar great for debugging and error resolving (suggestions, summaries, insights).
AI allows you to describe the business logic once and get output with any framework or approach.
You can create code from almost nothing. Ask AI to describe business logic, then ask for adjustments, and re-generate the code.
To avoid generic output, understand the desired result enough to describe it. Provide details, set rules, and guide AI to correct answers.
Code quality is generally lower than what's expected from skilled developers.
AI is generally poor at working with file and folder structure of the project.

Analytics dashboard

The app tracks every call and transaction to analyze a variety of metrics. They include the number of calls per technician, their time, total amount of fees generated, number of call participants, call success rate, duration, and so on.

Business value

A single engineer created the entire app in two months. He spelled out all the edge use cases and let the AI copilot implement the flows. This allowed us to create features and integrations in less time than it takes to scan the API docs. Generative AI can also write and run tests at the same time with the code, ensuring clean output. The entire project cost $31,200.
1
USD saved using AI-driven development (89% savings).
1
One engineer doing the work of a 5-person Scrum team.
1
Increase in development velocity without sacrificing the quality.

LOOKING FOR AN EXPERT DEVELOPMENT TEAM?

Set up the foundation for your future success – arrange a free consultation with our product development pros.

Contact us

More cases

Melody

Music search engine for Grammy-awarded producers

Learn more

CEMAsys

CSR solution for Forbes Global 2000 companies

Learn more

Aim Analytics

Data-driven TV and radio ad campaign management

Learn more