In June 2024, Grio launched the GRŌ app: an AI gardening assistant for novice gardeners. The GRŌ app was a superb manner for us to grasp the total extent of AI’s capabilities whereas figuring out pitfalls and limitations to keep away from on future initiatives.
Challenges and Limitations of AI Integration
As we started creating the GRŌ app, we bumped into some preliminary challenges related to the AI integrations:
Selecting the Supply AI
After we initially started planning GRŌ, we thought-about a number of of the highest open-source AIs, together with Gemini (previously Bard) and ChatGPT. We rapidly found that one of many largest hurdles to AI integration can be one of many first you face: the AI waitlist. As new open-source AIs are launched, they’re solely supplied to a restricted variety of customers.
This presents an attention-grabbing conundrum for brand spanking new apps. You possibly can both delay your app launch as you await entry to the most recent model of an AI or you should use an older, open model. When you use an older model, your AI is probably not as correct, however you could make it to market sooner.
For GRŌ, we found a great compromise between the 2 choices. We started with GPT-3.5, as we have been instantly granted entry. As soon as we have been in a position to make use of GPT-4, we upgraded to make use of the newer model of the system. Whereas this may increasingly have added a while to the mission, beginning with 3.5 allowed us to get a great sense of the best way to use the system.
Crafting Efficient Queries
The GRŌ app makes use of two AI integrations. Within the first, customers enter details about their location and the AI produces specialised suggestions for crops and care based mostly on their biome and their USDA Hardiness Zone.
The problem with this integration was not getting the AI to return data, it was getting the AI to return the proper data in the best format. Our builders frolicked experimenting with the best way to arrange the question inside the code to make sure that customers acquired related, concise responses every time.
Monetary Burden of Chat Function
The second AI integration we used within the GRŌ app was for the person’s private AI gardening assistant. With the assistant, customers are in a position to sort in questions and obtain rapid responses.
This integration permits us to supply a degree of service that’s not obtainable in different gardening apps in the marketplace. Nevertheless, this function additionally creates an added monetary burden, as we should pay for each query that customers ask.
To scale back the monetary burden with out compromising the expertise, our builders carried out two key options. First, they created parameters for a way the gardening assistant would reply questions. For instance, the geographic information collected within the first integration is submitted together with the person’s question, permitting the AI to supply personalised responses that align with the person’s local weather.
Second, our builders created parameters for which questions the gardening assistant would reply. To stop the assistant from getting used for all AI questions the person could have all through their day, our builders specified that the gardening assistant may solely reply gardening and plant-related questions. If the person asks one thing outdoors of this scope, the AI assistant responds with: “I’m an AI Horticulturist. I can solely reply questions on crops and gardening.”
Limitations of the Dataset
The most important limitation of this AI integration was that we have been depending on ChatGPT’s data base. For a gardening app, this isn’t a prohibitive difficulty, as ChatGPT’s dataset consists of numerous gardening and horticulture sources. Nevertheless, for an organization creating a really technical or domain-specific app, a normal dataset wouldn’t be adequate.
In order for you a extremely correct AI, then you have to create a specialised dataset that it could possibly be taught from. That is cost-prohibitive for many organizations, making it troublesome to create a completely correct AI integration.
For the GRŌ app, we offer up-front language that explains that the app makes use of ChatGPT integrations and that AI doesn’t at all times present correct outcomes. We additionally permit customers to re-ask questions in the event that they really feel that they haven’t acquired a adequate reply within the AI’s preliminary response.
Advantages of AI Integration
The GRŌ app highlighted two giant advantages of AI integration:
Providing Customized App Experiences
When customers start their GRŌ journey, they supply details about their backyard location. With AI integration, we have been in a position to immediately gather this data and return personalised plant and care suggestions based mostly on their biome.
AI has allowed builders to create far more personalised experiences for customers, thereby bettering the general person expertise.
Providing Chat Help 24×7
With the GRŌ AI gardening assistant, customers have rapid solutions to their questions 24/7. The place earlier than, corporations have been restricted by the variety of individuals they may rent, it’s now simple to supply round the clock service for a fraction of the associated fee. This degree of availability improves person expertise and buyer loyalty.
Enhance Your App with AI Integrations As we speak
AI integrations are revolutionizing the methods we create and make the most of apps. At Grio, we’re discovering progressive methods to make use of AI integrations and maximize your app’s potential. Arrange a free session at present and learn how our AI specialists may also help you create an distinctive app.
#GRŌ #Choices #Integrations
