Do you have an example of something you are trying to solve for? We have used Airtable for project management by creating sprints, projects, tasks with automations to generate recurring tasks.

This isn’t Airtable-specific, but for AI product management use case inspiration, check out Claire Vo’s podcast:

So, for example, if one of your PM tasks is to write detailed product specs for developers, you could look at using a field agent to help draft a detailed project spec. Create a table for project specs, put the info you want to get into the spec in the row (as one or several fields), and give the field agent a prompt specifying what’s in your company’s typical project spec and how it’s formatted PLUS an example spec (one-shot prompt). I recommend leaving a human in the loop ;-)  

I’ll be forthright and note that I’m a something of a genAI/LLM skeptic (I know, I know—courting controversy in the AI section of the forum). That said, just want to contribute to this conversation with the following:

The thing more and more studies around implementing AI in the workplace are finding is that if the approach is too general, chances are the initiative won’t go anywhere (see this article summing up an MIT study). GenAI is also something that runs best when your data is clean, structured well, up-to-date, which I think is another issue a lot of companies end up running into.

​@Lorraine got to the core of this by effectively asking ‘what are you looking to solve’? You’re more likely to find success here if you’re able to identify a specific problem, bottleneck, time-sink, or friction point that you’d like to address, and ask if genAI can actually help—without compromising reliability, compliance requirements, confidentiality/security, etc. (Keeping a human in the loop like ​@nroshak mentioned can help address some of those risks.)

Right off the bat, based on what you mentioned in your post, using a field agent to summarize meeting notes would be the first use case that pops to mind. Extracting certain pieces of data from contract or invoice PDFs might be another. 

When it comes to Airtable functionality specifically, I’m also very big advocate for asking “what issues can I address without genAI?” i.e. is there an AI-less automation capability I haven’t fully explored yet? A third-party integration or connector that would address a pain point without a high cost (e.g. a Make or Zapier connection)? A script that would give my base a little extra juice to complete a specific task? 

At the end of the day LLMs are word prediction machines—highly complex ones, sure, but I find thinking about them in that way helps me better grasp what they’re better suited for (summarizing, assisting with drafting, maybe some light research). Knowing that, if I’m really trying to improve efficiency instead of creating more data that needs to be reviewed or polished by a human, I can better assess when using genAI makes sense, or when I’m better off turning to a specific, deterministic automation; creating a repertoire of record templates; utilizing extensions; or enhancing UI/UX experiences via interfaces to provide structure and clarity to different workflows.

This reply is a long-winded way for me to argue that even if Airtable is now AI-native, it doesn’t mean you’re condemning your team to inefficiency and obsolescence if you rarely use Omni or field agents. Just some food for thought as more and more teams, like yours, ask if implementing genAI is worth the time investment.

Hey everyone, just wanted to chime in and build on what’s been said here, especially the point about not always needing AI for everything. I totally agree that sometimes you can achieve what you need with straightforward rules or automations rather than jumping straight into using AI. For example, another user here recently asked about using AI for matchmaking, and in that case, you can often just create your own criteria and let a script handle it without AI. That way, you’re not dealing with a black box and have more control. But I do see the appeal of using AI, as it might seem like a more approachable way to create an algorithm by using a prompt rather than a script. 

Now, to answer the original post and share how I use it, when I do incorporate AI, I make sure to integrate it with scripting so I get the AI’s responses in JSON format. This is key: instead of just getting a blob of text, I get structured data with each task as a separate object. For example, when I’m working on development stories for clients, I feed in context from previous projects and examples so the AI knows how I typically structure tasks. It then automatically generates new tasks in JSON format, which I convert directly into Airtable records using a script. This means I’m not manually doing this each time, it’s all automated and continually improving as the AI learns from new data. 

For more specific project management workflows, you could use a similar approach to generate project tasks directly from descriptions or meeting notes. If you’re using an AI meeting note taker, you could send those notes into Airtable and automatically create action items from them. Another use case is tagging or categorizing project data. AI can pull out the most relevant keywords or categories from large amounts of information, giving you a cleaner way to organize tasks. Those categories can also make your later prompts more effective, since you’re feeding the AI more focused and structured data to work with.

In short, it’s about being thoughtful, use AI where it genuinely adds value and rely on simpler automations when that’s all you need. You’re not missing out if you keep it simple where you can. Hope that helps!

Hello All - In the past I have received notification when responses are added to my posts but did not in this case. Maybe AI is to blame? 

​@Lorraine - I agree with the what is your use case recommendation but I really didn’t know where to start. Thus, the post to get inspiration from use cases others have identified. 

​@nroshak - Have not watched the demo just yet but will do. Appreciate the Product Spec example.

​@stagandforge - We are already using Teams AI to summarize meeting minutes or, when not in our teams, add transcript to Copilot with prompt to generate the minutes in the desired format. 

​@stagandforge  and ​@airvues - I appreciate the suggestions and definitely see auto generation of project tasks, Action Items, etc. as positive use cases I can explore. I do want to avoid third party tools or scripting solutions if possible. a) Because scripting is not in my skill set any more than brain surgeon or rocket scientist. b) I am hopeful to use OOTB solutions or at least to identify/test/deploy those solutions first to wow the masses and especially the brass. Then I may be able to get funding/interest to bring in persons with the right skills to build out more complex use cases. 

Thank you all for the inputs. I have some things to explore. 

Regards,

Jason Knighten