Anyone with troubles with the Filtering and AI text generation in Automation?
I’m running into a strange issue. I have a table of records that I filter by ‘Created date’ before passing it into “Generate text with AI.” At this AI stage, I’m just filtering entries where the Source is a scientific source (peer-reviewed papers, preprints, PubMed, etc.).
This used to work perfectly. But today the output only contains 1 row, even though my input table has more valid entries.
If I feed the AI a table containing only the two Nature entries, it correctly outputs both.
But when I pass in the full table (with more rows from other sources), it only outputs one Nature entry.
Example Title: Paper 1 Source: nature Title: Paper 2 Source: nature Title: Blogpost Source: yahoo
Expected output: Paper 1 (nature.com) Paper 2 (nature.com)
Actual output with full table: Paper 1 (nature.com)
I also tried simplifying the prompt to: Based on the table: Grid of Records Print all rows where the Source has "nature"
But it still only returns one row. I first thought this might be a visualization issue (limited to 10 results), but when I run the automation, it really only gives me a single result. Is it an issue with the system? I dont know how to troubleshoot this.
Page 1 / 1
Hey @Daniel2909,
Would you mind sharing the exact prompt you are using?
Also, rather than handling the “filtering” within the AI step itself, did you think about having an AI Agent field on the table itself to adequately categorize the record as Scientific Source or Non-Scientific Source, then on the automation have a Find Records field (apply the filter where Source is exactly Scientific) and only then push data to the AI automation step?
Hm, as a datapoint this seems to be working okay for me, screenshots below. Any chance you could share some screenshots of your data and automation setup? If you could DM me a link to your base I could take a look for you too!
Setup:
Result:
Thank you very much for your replies. The field agent was extremely helpful here. I did the following: added a column with the field agent filtering the records. So all scientific entries will be tagged as “Research”, and it is working perfectly fine. My prompt is very basic:
I have 100 records in my table. In the preview step, Airtable only displays the first 10 results, which I know is expected behavior. However, when I actually run the automation, instead of processing all 100 records, it only processes 10 of them. Here’s my real example:
Record #8 is tagged as Research.
Records #91 and #98 are also tagged as Research.
That means 97 entries are tagged Out, and only those three are Research.
If I filter directly on the region containing records #91 and #98 (by date), both are correctly returned. But if I run the automation across the entire dataset (all 100 records), the output includes only record #8.
At first I assumed this was just a preview limitation, but even when the automation runs fully, the email output contains only record #8 (almost as if it is restricted to the first 10 records of the grid). The “Maximum records” setting is already set to 1000, so that shouldn’t be the issue (?). And the Date region covers all records.
The dataset is very simple: a table with one column tagged either Research or Out. I’ve used this automation successfully before, but only recently started seeing this behavior.
Title
Source
Created
Tag
Paper
nature
29/09/2025
Research
blog
x.com
30/09/2025
Out
Hm yeah, I’m seeing the same behaviour, feeding it 30 records, with 20 that meet the criteria and it only outputs 14
Reckon you may want to open a support ticket to get a concrete answer
Hey @Daniel2909,
I’m glad to hear that! Make sure to reach out if you have any other question :D