# AI is here

**URL:** <https://community.glideapps.com/t/ai-is-here/56611>\
**Category:** Ask for Help\
**Created:** [January 12, 2023, 10:59am UTC](https://community.glideapps.com/t/ai-is-here/56611 "2023-01-12T10:59:53Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![IlseB](https://sea2.discourse-cdn.com/flex002/user_avatar/community.glideapps.com/ilseb/32/44063_2.png) [@IlseB](https://community.glideapps.com/u/IlseB)\
**Post date:** [January 12, 2023, 10:59am UTC](https://community.glideapps.com/t/ai-is-here/56611/1 "2023-01-12T10:59:53Z")

</div>

Hi,

Has anyone started to experiment with gpt-3 and use cases for our glideapps?

Thanks

---

<div class="post-metadata">

**Author:** ![Robert\_Petitto](https://sea2.discourse-cdn.com/flex002/user_avatar/community.glideapps.com/robert_petitto/32/25193_2.png) [@Robert\_Petitto](https://community.glideapps.com/u/Robert_Petitto)\
**Post date:** [January 12, 2023, 12:12pm UTC](https://community.glideapps.com/t/ai-is-here/56611/2 "2023-01-12T12:12:55Z")

</div>

I have. My first use case was using rollups to determine, let’s say, a user-score. Based on that score and a couple other factors, I used an if then else column to output a text prompt. If users ask for a “suggestion”, I call up OpenAI, send the prompt and wait for the response. I then display the response in the app. Users can then ask for “another suggestion” and they’ll receive another one. If using Glide API, it has a decent turnaround time.

---

<div class="post-metadata">

**Author:** ![bjgray](https://sea2.discourse-cdn.com/flex002/user_avatar/community.glideapps.com/bjgray/32/50137_2.png) [@bjgray](https://community.glideapps.com/u/bjgray)\
**Post date:** [January 21, 2023, 5:47am UTC](https://community.glideapps.com/t/ai-is-here/56611/3 "2023-01-21T05:47:07Z")

</div>

I was dabbling with the **GPT for Sheets** extension in GS and thought I’d try it out on a problem I’d posted about a few weeks ago ([Transform JSON with JQ](https://community.glideapps.com/t/transform-json-with-jq/56088)). I needed to parse out just the _amount_ and _description_ from a block of json code.

In the screenshot below, the json code—which is dynamic content pulled in through an API— is in the column on the right. The middle column is the GPT prompt I provided and the left column is the resulting markdown text. The next screenshot is the table as it appears in the app. Not perfect but the potential is pretty mind-blowing.

 ![Screen Shot 2023-01-20 at 11.18.17 PM](https://us1.discourse-cdn.com/flex002/uploads/glideapps/original/3X/7/4/742929da8842e9bac618327a592249e8b3efabb4.png)

 ![Screen Shot 2023-01-20 at 11.37.15 PM](https://us1.discourse-cdn.com/flex002/uploads/glideapps/original/3X/4/3/4366976a25712d10ef61576fe001fefbdcc8fad2.jpeg)

---

<div class="post-metadata">

**Author:** ![kevingorne](https://sea2.discourse-cdn.com/flex002/user_avatar/community.glideapps.com/kevingorne/32/53704_2.png) [@kevingorne](https://community.glideapps.com/u/kevingorne)\
**Post date:** [January 24, 2023, 2:37am UTC](https://community.glideapps.com/t/ai-is-here/56611/4 "2023-01-24T02:37:31Z")

</div>

The company I work for used chat gpt to accurately map any table I have in google sheets to my data model in my database and it works pretty well. Since database columns can be pretty freeform, chat gpt’s NLP still does a good job at figuring out which column in g-sheets maps to which column in our database / data warehouse