AI Workflow Agent — Put AI to Work Inside Every Automation
Add a prompt-driven AI step to any automation—reading Record data and attachments, returning structured output, and handing results to every downstream action, all inside your governance.
What Is AI Workflow Agent?
The AI Workflow Agent is a SmartSuite automation action that embeds a large language model call directly into a workflow. Builders compose a prompt from static instructions and live Record data, choose a connected AI model, and map the response—as free text or structured fields—into the automation's next steps. It is available on all current SmartSuite plans, and every workspace includes 100 free AI requests through the built-in SmartSuite Trial connection—learn more about Automations.
Prompt Builder
Compose Prompts from Live Record Data
Every AI Workflow Agent step starts with a prompt you control—static instructions mixed with dynamic variables pulled from the trigger Record or from earlier steps in the same automation, including single-record and multi-record field values. When the selected model supports attachments, the step can read files like PDFs as input.
The result is AI that works from the same governed Record data as the rest of your workflow—not from copy-pasted context.
Structured Output
Return Structured Fields, Not Just Text
Choose Simple Output to stream the model's raw response into a text area, or Custom Output to return structured data mapped to defined field types—Text, Number, Boolean, Date/Time, Email, URL, Phone, Single Select, and Duration. Output can be a single value, a list, a JSON object, or a list of JSON objects.
Per-field instructions sharpen accuracy, and Single Select outputs must match a configured option exactly—so what AI writes stays inside the structure your Tables enforce.
Workflow Orchestration
Route AI Results Into Every Downstream Step
Downstream actions consume the AI response like any other automation data—create, update, or delete Records, send emails, find Records, branch on conditions, or loop over lists. Classification results can drive routing; extracted values can populate Fields the moment a Record arrives.
Each AI Workflow Agent action uses one model, and multiple agent actions can be chained inside a single automation—so different steps can call different models where it makes sense.
Model Choice
Connect the AI Providers Your Organization Approves
The AI Workflow Agent runs on provider connections established in the AI Center—OpenAI, Anthropic, Google Gemini, Amazon Bedrock, Azure OpenAI, Perplexity AI, Nscale, IBM watsonx, and xAI. Each connection carries its own credentials, from API keys to AWS IAM and Azure endpoints, managed by admins rather than pasted into automations.
Every workspace also ships with the built-in SmartSuite Trial connection—100 free AI requests, no API key required—so teams can build and test before committing to a provider.
Governed AI
Govern Every AI Step Through the AI Center
Admins set up each provider connection once in the AI Center, scope it to specific Solutions or the whole workspace, and decide independently whether the AI Workflow Agent may use it. Builders pick only from approved connections—credentials never leave admin hands, and every call is logged with its feature, Solution, date, and result.
AI capability and oversight arrive together: workflows get smarter while admins keep a complete record of what ran, where, and on which model.
AI Workflow Agent
AI Workflow Agent steps run inside governed automations—triggered by Record events and feeding structured results straight into the risk, compliance, and audit workflows your team already runs.
Classify inbound regulatory-change items into a Single Select impact category the moment a Record is created. Strict option matching keeps classifications inside your taxonomy, and downstream steps route each item to the owning compliance area.
Parse a free-text risk submission into structured fields—category, affected process, and a normalized description—so new risks enter the register ready for scoring and review instead of waiting on manual re-entry.
Read an uploaded PDF—a certificate, report, or attestation—and return key values as structured output. Evidence Records arrive populated and consistent, which helps control owners stay audit-ready.
Generate a status-update draft from a finding's remediation fields and hand it to a downstream Send Email step. Issue owners review and send—AI accelerates the writing, people keep the judgment.
Pair the AI Workflow Agent with Automations conditional logic to branch on an AI classification—high-severity risk submissions route to human review while routine ones proceed automatically.
AI Workflow Agent steps run inside governed automations—triggered by Record events and feeding structured results straight into the risk, compliance, and audit workflows your team already runs.
The AI Workflow Agent is an automation action that embeds an AI model call inside a SmartSuite automation. It sends a prompt built from your instructions and live Record data to a connected model, then routes the response—as text or structured fields—into subsequent automation steps.
No. You configure it inside the no-code automation builder: write a prompt with field placeholders, pick a connection and model, and define the output shape. SmartSuite recommends testing the automation before turning it live.
It supports connections to OpenAI, Anthropic, Google Gemini, Amazon Bedrock, Azure OpenAI, Perplexity AI, Nscale, IBM watsonx, and xAI, each configured in the AI Center. Every workspace also includes the built-in SmartSuite Trial connection with 100 free AI requests and no API key.
Yes. Custom Output maps the model's response to typed fields—Text, Number, Boolean, Date/Time, Email, URL, Phone, Single Select, and Duration—as a single value, a list, a JSON object, or a list of JSON objects. Single Select responses must exactly match a configured option on the target field.
The SmartSuite pricing page lists AI Workflow Agents as available on all current plans. Every workspace includes 100 free trial requests; continued or heavier use runs on a provider connection your admin adds in the AI Center.
Provider connections are created, scoped, and toggled centrally in the AI Center, so builders never handle credentials. Activity logging records each call's feature, Solution, date, and execution result, and Run History traces the automation runs AI steps execute inside—helping teams stay audit-ready.
Each AI Workflow Agent action uses a single model and provider, but you can chain multiple agent actions within one automation—so different steps can call different models where the task warrants it.
Build AI-Powered Workflows in Days, Not Months
Start with 100 free AI requests, wire a prompt into your first automation, and watch structured results flow downstream—governed from day one.