The sheet is where it ends up
Most of what people ask dassi to do with Google Sheets doesn’t really happen in Sheets. The work happens on other sites (a retailer’s product page, a company directory, the Stripe dashboard) and the sheet is just where the result gets written down. dassi reads those sites in your browser, signed in as you, and types the rows in the way you would.
Competitor prices are the most common: we write that up as competitor price monitoring, with a column per week. Then there’s building lead lists out of a directory, a row per company and a source link on each. In one run we did against the Y Combinator directory it narrowed 6,238 companies to 31 recent fintech startups that were hiring and wrote up the 15 we asked for in about three and a half minutes, and one of those was tagged fintech but reads more like a voice AI company. I’d want a person to look at that row before anyone emails them. Sometimes the direction flips, and rows from the sheet get typed into a web form one at a time; the data entry job covers that. When we tested that on a public practice form it submitted two rows and stopped on the third, where the email had no domain ending and the phone number was six digits long, and it left that row alone rather than making up a correction.
Worked example: prices every Monday
Column A holds product URLs: your listing plus the same item on three competitor sites. Taught once, the Monday run goes:
- Open each URL, pick the right size or color if the page asks, and read the price, stock status and seller.
- Write those into this week’s columns with the date.
- Compare against last week and highlight any price that moved more than 5%.
- Put two lines at the top of the tab: the lowest price, and anything that looked odd.
I’d keep that second line even if you drop everything else. We ran this check on a 24 oz Owala FreeSip bottle across Owala, Amazon, Target and Walmart and it took just under five minutes. Standard colors sat between $29.97 and $29.99 everywhere, but Amazon also had third-party sellers asking up to $61.99 for retired colors, and if an average swallows that, the sheet ends up telling you the going price is somewhere in the forties.
Formula help
dassi also helps inside the sheet itself. Ask for month-over-month change on Revenue, or spend per vendor for this quarter only, and it reads your headers and ranges, writes the formula into the cell, and says what it did. I find it most useful for QUERY and ARRAYFORMULA, which I look up every single time. The formula suggestions page has examples.
Approvals, and when not to use it
Writing to a sheet is easy to undo, because version history keeps every change, so most people let dassi write without asking. Submitting a form, sending an email or publishing something on another site is different, and those are the steps to flag for approval. It also stops and asks when a page shows something it can’t reconcile with what’s in the sheet.
Some cases don’t need dassi at all. If a price sits in plain HTML on a public page, IMPORTXML might do the job for free. dassi makes sense when the page needs a login, a variant picked, or somebody deciding whether two listings are the same product. It also types through the page like a person, so a few hundred rows is comfortable and fifty thousand belongs in a CSV import. Retail sites watch for automated traffic, so dassi goes at a human pace, and when a site shows a CAPTCHA it hands that to you instead of trying to get past it. Page content goes to the AI model you picked (through dassi’s relay if you’re on included credits), and scheduled runs happen in your own browser, which has to be open on Monday morning.