Run your first agent
Pick the agent that fits the job and a model that won't waste money. Then send your first task.
You’ll ask a specialist to research customer-acquisition ideas and save them in a sheet. Before writing the prompt, choose an agent and model that fit the job.
Step 1
Choose the agent for the job
Select New chat in the workspace, then choose the team that owns the work, then choose the agent from that team.
Match the agent’s job and tools to the result. A specialist won’t waste model usage working out a role it doesn’t have.
If none of the available agents fit, create an agent for the job.
Step 2
Choose a cost-efficient model
Pick a fast, lower-cost model with Low reasoning for this first research task. Models and provider prices change, so compare the choices in the picker and use the session total as your guide.
Image placeholder
Capture the model picker at 100% zoom with several current models and Low reasoning visible. Crop close enough to read names and price information.
Step 3
Ask for a clear outcome
State the result you want and name the fields it must contain. Leave the tool choices to the agent instead of writing a click-by-click procedure.
Example task
Find 10 customer-acquisition strategies that an early-stage SaaS company can try. Save them in a sheet with these fields: strategy, description, implementation steps, estimated cost, timeframe, difficulty, impact, and source URL.
Step 4
Review and steer the session
Once the agent finishes, check its response and the sheet. Ask for corrections in the same session; “add three more,” for example, continues with the context already there.
Sheets are team-scoped databases that hold information beyond a single session. People can edit the rows directly, and agents in that team can work with the data later.
Image placeholder
Capture a completed research session beside its populated sheet at 100% zoom. Include the follow-up composer and session cost without making the table text too small to read.
Step 5
Build on the result
To turn the sheet into an app, start a session with Ricky - App builder. Tell Ricky which sheet to use and what the app should let people do.
Each app binds to one Hivy Sheet. A larger build takes more time and model tokens, so check the session cost before asking for another round of work.
Video placeholder
Record a 45-second clip at 100% zoom. Select Ricky - App builder, ask it to use the existing sheet, open the finished app, and end on the session cost.
Before you start
- Match the agent to the job.
- Choose a fast, lower-cost model with Low reasoning.
- Name the result and its required fields.
- Check the sheet, then make corrections in the same session.
- Choose Ricky - App builder if you want an app, and watch the session cost as it works.