Skip to main content

Use AI services in automations

Connect external AI tools and services to your workflows with an automation. Send record values, get an answer back, and save it in a field of your record.

Written by Leo

Tape doesn't have its own AI. Instead, an automation sends values from a record to the AI service you already use and saves the answer back in Tape. You decide which values are sent and which field gets the answer.

You can use this for many things, including:

  • Summarizing long notes on a lead into two sentences.

  • Drafting a first reply to a customer request.

  • Sorting new requests into a category, such as "Billing" or "Support".

Automation with the trigger When a record is created in Expenses, the action Send a HTTP request using POST to an AI service, and Update the current record that saves the HTTP response in the field Approval note

Note:

  • Your data goes to the AI service. Everything you put in Request is sent to that service. Check its terms and your company's rules for AI before you send customer data.

  • Want an AI assistant to work in Tape instead? Connect it with Tape MCP. Learn more about Tape MCP →

Send a record to an AI service

Automations are available on the Free, Plus, and Premium plans. The number of automation actions per month depends on your plan. See the limits for your plan on our pricing page. To set this up, you need to be a workspace admin of the workspace the database app is in.

To send a record to an AI service:

  1. Open the database app and click the Create and view automations icon at the top.

  2. Select New automation and choose a trigger, for example When a record is created.

  3. Select Add action → Send HTTP request. The step appears as Send a HTTP request.

  4. Select the method next to using, then select POST.

  5. In URL, enter the address that your AI service gives you for requests.

  6. Select Show headers. Enter the header your service asks for, for example "Authorization" as Header and your key as Value. Select Add header to add a second header with "Content-Type" and "application/json".

  7. In Request, write the instruction for the AI in the format your service expects, usually JSON. Type @ to insert record values, such as the notes of the record.

  8. Select Create automation.

In Expenses, the automations icon opens a new automation, When a record is created starts it, Add action and Send HTTP request add the step, POST is selected, <a href="https://api.example-ai.com/v1/chat" rel="nofollow noopener noreferrer" target="_blank">https://api.example-ai.com/v1/chat</a> is entered as URL, Show headers opens a header row with Content-Type and application/json, and the Request asks to summarize the value Expense inserted with @

Save the answer in your record

The answer is saved in the variable under and save response in, called HTTP response unless you rename it. If the service answers in JSON, Tape saves it as an object.

To save the answer:

  1. Below the request, select Add action → Perform calculation or Execute script to read the answer from the response, for example var_http_response followed by the path your service uses for the text.

  2. Select Add action → Update record. The step appears as Update the current record.

  3. Choose the field for the answer, type @, and select the result of the step before.

  4. Select Save changes.

In the automation New expense automation, Add action and Perform calculation add the step with var_http_response.text, Add action and Update record add Update the current record, Approval note is chosen, typing @ and selecting Calculation result inserts the result, and the pointer rests on Save changes

Where the text sits in the response depends on the service. Learn more about the Perform calculation action →

Test your automation

Select Manual run, choose the records under Run on, then select Save and start run. The request goes to the real AI service. If the field stays empty, add an Execute script step with console.log(var_http_response) after the request. Its output shows in the run's steps. Learn more about Execute script →

In the automation Expense approval, selecting Manual run opens Run on selected records with a record search, Save and start run, and the list of expenses; choosing Client dinner adds it, and the pointer rests on Save and start run

Let's look at an example

Let's say Juliet Adams works in sales at Acme Inc. and wants a short summary of the notes on every new lead. Here's how:

  1. She creates an automation with the trigger When a record is created in the Leads app.

  2. She adds Send HTTP request with POST, the address of her AI service, and her key in the headers.

  3. In Request, she asks for a two-sentence summary and inserts the "Notes" field with @.

  4. She reads the answer with Perform calculation and saves it in the field "Summary" with Update record.

Every new lead now gets its summary on its own.

Errors with AI services

  • The step takes too long. Each request has to answer within a short time, and a run ends after three minutes. Ask for shorter answers. Learn more about limits →

  • The field stays empty. An error from the service, such as a wrong key, doesn't fail the step. Print the response with console.log in an Execute script step and check it in the run.

  • The service blocks the request. Allow Tape's IP addresses. Learn more about IP addresses →

Learn more: Ideas and examples from other customers are in the guide for using AI in Tape →

FAQs

Does Tape have its own AI?

No, Tape doesn't include its own AI. You connect the AI service you choose with an automation, or connect an AI assistant with Tape MCP.

Which AI services can I use?

Any service that accepts web requests. If the service needs an OAuth 2.0 login instead of a key, use Send API request. Learn more about Send API request →

What's the difference to Tape MCP?

With an automation, Tape sends data to an AI service when something happens in Tape. With Tape MCP, an AI assistant works with your Tape data when you ask it to. Learn more about Tape MCP →

Still have more questions? Find answers and get help from community experts in the Tape community →.

Did this answer your question?