How to Use Gemini in AppSheet to Build a Job Tracking App

Coding Liquids tutorial cover featuring Sagnik Bhattacharya for How to Use Gemini in AppSheet to Build a Job Tracking App.
Coding Liquids tutorial cover featuring Sagnik Bhattacharya for How to Use Gemini in AppSheet to Build a Job Tracking App.

To use Gemini in AppSheet to build a job-tracking app, sign in with a paid AppSheet licence, open My Apps, choose Create, then App, then Start with Gemini, and describe your jobs, statuses and who works on them. Review and edit the proposed tables, create the app, then delete its sample data.

The Complete Excel Course with AI Integration

From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.

By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner

Buy now

Two points set expectations. Gemini builds the structure (tables, columns, links and a few starter views), not your business rules, so the app you get is a draft to test with staff. And it creates the app on an AppSheet database rather than a Google Sheet, which matters for row limits and for anyone who wanted to keep editing the data in Sheets.

Which AppSheet plan do you need for Gemini?

Gemini for App Creation is included in the paid AppSheet licences: Publisher Pro, Starter, Core and Enterprise Plus, and you may also have access to Core through a Google Workspace licence. AppSheet's help page for Gemini says you do not need Gemini Enterprise or a Gemini add-on for Workspace. The pricing page lists Starter at $5 per user a month, Core at $10 and Enterprise Plus at $20, and lets you build and test with up to 10 users at no cost; the Gemini help describes access in terms of paid licences, so confirm on your own account whether the free test mode shows the Start with Gemini button.

LicencePer user a monthGemini for App CreationAppSheet database rows
Free (building and testing)$0Check your account1,000 per database
Starter$5Included2,500 per database; 5 databases
Core$10, or included in most paid Workspace plansIncluded2,500 per database; 10 databases
Enterprise Plus$20Included, plus Gemini in AppSheet Solutions200,000 per database

Gemini in AppSheet Solutions is a different feature, included with Enterprise Plus only: it adds AI tasks inside automations, such as extracting data from an image or PDF, and consumes credits. This tutorial uses only app creation. If you are weighing plans for a small crew, AppSheet's pricing and what each tier adds covers the decision.

How do you describe a job tracker to Gemini?

A good Gemini job-tracker description covers five things: what a job is, what you record about it, who works on it, which statuses jobs move through and which views you want. The Gemini help page says to give as much detail as you can, and that rule matters because the output is only as specific as the input.

Take a four-person marketing agency that runs design and print jobs for clients and keeps losing track of who is doing what. Compare two prompts.

The weak one: "A job tracker for my agency."

The better one (illustrative):

Track production jobs for my marketing agency. Each job belongs to a
client and has a title, a brief, a due date, an owner from our team of
four, an estimated number of hours and a status: Briefed, In design,
With client for approval, In production, Delivered. Clients have a name,
a contact email and a phone number. Team members have a name and an
email. Owners log hours worked on each job with a date and a note. I
want to see which jobs are due this week, who has too many jobs open,
and which jobs have been With client for approval for more than five
days. Staff will use phones.

The better prompt names the statuses, which become the options of a drop-down. It names the second table (hours logged), which stops everything being squashed into one wide sheet. And it ends with the questions you want the app to answer, which shapes the views.

What Gemini sends back

Google's own example prompt, about managing facility inspections, returned four tables (Inspectors, Reports, Inspections and Facilities), with columns listed under each and links between them, such as Facility and Inspector columns in Inspections that point at the Name column of the other tables. Expect the same shape for your prompt. A plausible, illustrative result for the agency's brief:

TableColumns proposed (illustrative)
ClientsId, Name, Contact email, Phone
Team membersId, Name, Email
JobsId, Title, Client (link), Brief, Due date, Owner (link), Estimated hours, Status
Time entriesId, Job (link), Team member (link), Date, Hours, Note

An Id column is created automatically as a unique key and cannot be edited or deleted, per Google's help. Read the result as a first draft and check three things: did it invent columns you do not want, did it miss one you need, and does Status hold your five values exactly.

Why edit the structure before you create the app?

Edit the proposed structure in the chat window before you create the app, because changing tables and columns there saves an hour later. Step 2 of the flow lets you edit the proposed structure before the app exists, as AppSheet's help describes:

  • Add a table with the Add table button, naming it and adding columns.
  • Edit a table by clicking it: rename it, add a column, change a column's name or type, edit the options of a single-select or multi-select column, or delete a column or the whole table.
  • Start over discards the structure and returns you to the prompt.

For the agency's tracker, make four edits. Open Jobs, check that Status is a single-select with exactly Briefed, In design, With client for approval, In production and Delivered. Add an Approval sent date column, so you can find jobs stuck with the client. Add a Priority single-select (Normal, Urgent). And add an Image column on Jobs if proofs will be photographed on site; Google's own example adds exactly that, with Add column, a name and the Image type.

If your first draft is off in a big way, Start over with a better prompt rather than editing for twenty minutes. A description that names tables explicitly ("a Clients table, a Jobs table and a Time entries table") is likely to get a closer first draft than one that leaves Gemini to guess.

What do you do with Gemini's sample data?

Delete every row of sample data before anyone uses the app, because Google's help says Gemini creates the app on an AppSheet database with sample data that may be inaccurate. Clicking Create app opens the new app in the AppSheet editor. Gemini returns pre-built views and actions, and a welcome screen lets you try the live app before you click Customize with AppSheet to open the editor.

Do that now, in this order:

  1. Open each table in the app or in the AppSheet database and delete every sample row.
  2. Check that the Clients and Team members tables hold only real people, then add them.
  3. Add two real jobs from last week and see whether the views show what you expected.

Why it matters in numbers: each sample row counts towards the database's row cap, and fake jobs in a dashboard make real reports meaningless. A manager who looks at "jobs due this week" and sees five invented ones will distrust the whole tool.

Does Gemini build the app on your Google Sheet?

No, Gemini for App Creation begins with your description and builds an AppSheet database, not an app on your own Google Sheet, as Google's help describes it. You have two sensible choices:

  1. Accept the AppSheet database if the tracker is new and small. It has no formulas to maintain, and you can edit data in the AppSheet editor. The row cap is the cost: 2,500 rows per database on Starter and Core, counted across all tables in the database.
  2. Keep your data in Google Sheets if you want to edit rows in a spreadsheet, use Sheets formulas or share data with other tools. Create the app from your own data instead (in Sheets, Extensions, then AppSheet, then Create an app), and use Gemini's proposal only as a design reference for the columns you should have. AppSheet's quick start also shows how to import a sheet into an AppSheet database with Create, Database, Import from Sheets, which is a way to move existing records across.

Do the sum before you choose. At 40 new jobs a week, a database with Clients, Team members, Jobs and Time entries (say 4 hours entries per job) adds 40 job rows and 160 time rows a week, which is 200 rows a week, and 2,500 rows disappear in 12 or 13 weeks. If you track time per job, the AppSheet database on Starter or Core is the wrong home, and a sheet is the right one. If you do not log time and only track jobs, the cap lasts about 62 weeks at 40 jobs a week. For a longer comparison of the two stores, read AppSheet versus a plain Google Sheet.

When the cap is reached, AppSheet's limits page says existing data is retained and you keep read and edit access, but you cannot add new data. That is a bad day to find out, so set a calendar reminder at 80 per cent.

If the spreadsheet side is the slow part, try The Complete Excel Course with AI Integration on Coding Liquids.

What should you fix by hand after Gemini finishes?

After Gemini finishes, fix the business rules by hand: required fields, the initial owner, a My jobs view, a due-this-week view, sign-in and who can change Status. Gemini produces a working structure; the rules that make it fit your business are still yours. Spend an hour on these, in this order:

FixWhy
Make Title, Client, Owner and Due date requiredA job without them cannot be tracked
Set Owner's initial value to the signed-in userSaves a tap and stops jobs landing unowned
Add a view "My jobs" filtered to the signed-in userStaff mostly want their own list
Add a view for jobs due this week, sorted by dateAnswers the question in your brief
Require sign-in on the appWithout it AppSheet counts each device as a guest user
Limit who can change StatusOne person marking jobs Delivered skews reports

Once the basics work, the natural next steps are in two other tutorials: emailing customers and staff when a status changes and adding photos and a customer signature. Add them one at a time in the editor; I would not expect either to come out of the first app-creation prompt.

A second prompt for a different business

The same method works for any small team that tracks work. A four-fee-earner law firm tracking client matters could write (illustrative):

Track client matters for a small law firm. Each matter has a client, a
matter type (conveyancing, probate, employment, dispute), a fee earner,
an opening date, a next-action date and a status: Open, Waiting on
client, Waiting on third party, Closed. Keep a diary of notes against
each matter with a date and the author. I want a view of matters whose
next-action date has passed.

The principle is unchanged: name the tables, name the statuses, name the questions. Note, though, that client confidentiality changes the setup. Restrict access by role, think about who can see which matters, and ask your Workspace administrator about the security settings that apply to your plan; for sensitive data, get the data-protection position checked by a qualified adviser before staff enter real records.

How do you test a Gemini-built app with staff?

Test the app with two people for three days before you trust it, rather than rolling it out to the whole team. A simple script:

  1. Day one: each person enters their open jobs, around ten each, using the app on their own phone. Time how long a job takes to enter. If it takes more than a minute, there are too many required fields.
  2. Day two: they work from the app only, no side spreadsheet. Note every time they reach for something the app cannot show.
  3. Day three: the manager opens the due-this-week view and checks it against what they know. Anything missing is a data entry problem or a filter problem, and the pilot shows which.

Ask each person three questions afterwards: what did you do twice, what could you not find, and what would you drop? A list of five changes from the pilot is a better brief for the next version than any further prompt to Gemini. If you want a comparison point for the same job-tracker idea in another product, Airtable versus AppSheet sets out the differences in cost, offline use and AI.

If you want help fitting a tracker like this into how your team actually works, I offer one-to-one AI consultations for small businesses.

Three rounds of prompt, and what each one fixed

Most owners do not get the right structure first time, and that is normal. Here is how a realistic session for the agency's tracker might run (the outputs are illustrative, not captured from a live run):

RoundWhat you typed or didWhat came backWhat you changed
1The short prompt, "A job tracker for my agency"Two tables, Jobs and Clients, with generic columns such as Name and DescriptionStarted over, because there was no status, owner or hours
2The detailed brief aboveFour tables, five statuses, links between themEdited Status options, added Priority and Approval sent date
3Created the app and opened it on a phoneA working app full of sample dataDeleted the samples, made four columns required, added "My jobs" view

Notice that the improvement came from specifics, not from clever phrasing. The detailed brief took five minutes to write and saved an hour of editing. A habit worth copying: keep your best prompt in a document with the date, so a colleague can recreate the app or build a sister app for another team.

How much does the first week cost in money and time?

A four-person pilot on Core costs $40 a month at list price, or $0 extra if the Workspace edition includes Core, and roughly half a working day to build by my estimate. The $40 is four users on Core at $10 a month, and AppSheet's help says most paid Workspace business, enterprise, education, frontline and nonprofit editions include Core. The build, by my estimate rather than any vendor figure: 30 minutes to write the prompt and edit the structure, an hour to clear samples and set required fields, two hours to add the views and the sign-in rule, and a few short pilot sessions. That is roughly half a working day to a first version, against weeks of planning for a bought system and a fortnight of data migration.

The decision rule I would apply: if the pilot shows the team using the app daily after two weeks, keep it and add one improvement a month. If they drift back to a spreadsheet, ask why before you build anything further, because the usual answer is that entering a job takes too long, not that the app lacks features.

A mistake that costs a week: building twelve tables first

Picture an owner who asks Gemini for "a full business management app" and gets back a dozen tables for clients, suppliers, invoices, stock, staff, leave and more. The structure looks impressive. Staff use two tables and ignore ten, because the others duplicate tools they already use. The app is slow to understand and hard to change.

The remedy is scope. Build one app for one job: tracking work. Cut the proposed tables to the three or four the team will really open (Clients, Team members, Jobs, Time entries). If you later need stock or leave, create a separate small app, because AppSheet's licence counts users across all your apps, not apps, so more small apps cost nothing extra in licences.

What limits and admin settings apply to Gemini in AppSheet?

AppSheet counts licences by unique app users across all apps, caps every database at 20 tables and 100 columns per table, and lets an administrator switch Gemini for App Creation off for each team.

  • Licences: AppSheet counts unique app users across all apps. Ten staff on Core is $100 a month at list price, unless your Workspace plan already includes Core.
  • Database limits: 20 tables per database, 100 columns per table, 2,000 characters per cell and 5,000 in a LongText column, on every plan.
  • Admin control: an administrator can turn off Gemini for App Creation per team. If the button is missing, ask your Workspace or AppSheet admin before assuming a plan problem.
  • Terms: Google's help says that for Workspace customers Gemini in AppSheet is treated as a Workspace Generative AI Service, and for standalone customers as an AppSheet Generative AI Service, each under its own terms; read them before entering client data.

A sensible plan is a two-week trial with Gemini: describe the app, edit the structure, clear the sample data, test with two people, and decide then whether to continue, move the data to Sheets or buy a finished tool. The AppSheet introduction helps you check costs first, and the AI hub lists other tools worth comparing before you commit.

Gemini in AppSheet: questions before you build a job tracker

Do I need Gemini in my Google Workspace plan?

No. AppSheet's help says Gemini for App Creation is included in paid AppSheet licences (Publisher Pro, Starter, Core and Enterprise Plus) and that you do not need Gemini Enterprise or a Gemini Workspace add-on. Core is also included through many Workspace editions, so Workspace customers may already have access.

Does it build on my Google Sheet?

Not directly. AppSheet's help says Gemini creates the app on an AppSheet database with sample data, starting from your description. To use an existing Google Sheet, create the app from your data instead, or import the sheet into an AppSheet database, then use Gemini ideas to improve the design.

Can an admin switch it off?

Yes. Administrators can turn Gemini for App Creation off at team level, so if you have several teams you switch it off in each. Turning it off affects app creation only. It does not affect AI in automations or Gemini features in Workspace.

How many jobs will the database hold?

On Starter or Core the database holds 2,500 rows, counted across all its tables, against 1,000 on the free plan and 200,000 on Enterprise Plus. At 40 jobs a week, 2,500 rows lasts about 62 weeks if the jobs table is the only one, and less once other tables count.

How do I build a job-tracking app with Gemini in AppSheet?

Sign in with a paid AppSheet licence, open My Apps, choose Create, then App, then Start with Gemini, and describe your jobs, statuses and who works on them. Review and edit the tables Gemini proposes, create the app, then delete its sample data and test it with two staff before trusting it.

What happens to the sample data Gemini adds?

Gemini creates the app on an AppSheet database and includes sample data that may be inaccurate, so delete every sample row before you use the app. Each sample row counts towards the database's row cap, and invented jobs in a dashboard make real reports meaningless.

Further reads

Sources: AppSheet help-centre articles on Gemini for App Creation, subscriptions and database limits, and the AppSheet pricing page, read on 17 Nov 2026.

Rather not work this out alone?

An AI implementation consultation is a one-to-one call with me, not an agency or a sales team. You leave knowing what AI you can use in your business and where to start.

Book a 1:1 call with me