Ai Agent Action in Workflows is Live!
Add autonomous AI decision-making into ACEIRT Fusion Workflows with the new AI Agent action, including CRM-aware logic, memory, and full execution traces.
Bring autonomous AI into your workflows
Workflows no longer stop at rigid if/then branches. The new AI Agent action adds autonomous reasoning and decision-making inside any workflow, so the system can evaluate context and choose the next best action for you.
The AI Agent workflow action is now available. Add it to any workflow to let an AI agent interpret context, use your CRM data, and return decisions as text or JSON with full execution traces.
Instead of mapping every path manually, you hand the agent natural-language instructions and let it work like a CRM-savvy teammate inside the workflow.
How AI Agent changes workflows
Traditional workflows in ACEIRT Fusion follow deterministic rules: you define every condition, every branch, and every next step ahead of time.
The AI Agent action introduces a new pattern:
- A trigger runs (for example, a form submission).
- The workflow calls the AI Agent action with relevant context.
- The agent reasons over that context, your CRM data, and your instructions.
- The agent decides what to do next and outputs structured results for the workflow to use.
This lets you handle nuanced cases without exploding the number of branches, such as lead qualification, routing based on free-text responses, or deciding which follow-up sequence to start.
CRM-aware, traceable, and configurable
The AI Agent is designed to work with your CRM data and workflow tools rather than as a black box.
- Full CRM awareness: The agent can search and pull contacts across your CRM and understands CRM users. For example, you can instruct it to "assign this deal to Sarah" and it will map that request to the right user and record.
- Ready-to-use templates: Start fast with common patterns like lead follow-up, no-show appointment recovery, and other prebuilt agent behaviors.
- Per-tool control: For each tool the agent can use, control whether to let AI decide all field values or lock down specific fields that should never change.
- Conversational memory: The agent maintains a rolling summary of past executions so it can stay consistent about a lead or account across multiple runs.
- Structured output: Choose between natural language text output or strict JSON, so downstream workflow steps can parse and use results reliably.
- Execution transparency: Every run of the AI Agent action produces a detailed trace, so you can inspect which information it used, which tools it called, and why it made specific decisions.
- Multiple model options: Select from several underlying models so you can balance quality, speed, and cost for each workflow.
Add the AI Agent action to a workflow
Use the AI Agent action anywhere you currently use deterministic logic, such as right after a form submission or a CRM update.
Open or create a workflow
- Go to your Workflows area in ACEIRT Fusion.
- Open an existing workflow or create a new one where you want AI-driven decisions.
- Confirm that your usual trigger is set up, such as a form submission, new lead, or status change.
Insert the AI Agent action
- In the workflow editor, add a new action after your trigger.
- Select AI Agent from the list of available actions.
- Place it where you would normally add complex conditional logic or multiple branches.
Choose how to configure the agent
- In the AI Agent action settings, choose whether to build your own agent or start from a template.
- Connect any CRM objects or tools you want the agent to access so it can search contacts, deals, or appointments.
- Decide whether the agent should return text or JSON output for downstream steps.
Save, test, and inspect the trace
- Save your workflow and trigger a test run, for example by submitting the form you used as a trigger.
- Open the workflow run details and review the AI Agent trace to see the reasoning, tools used, and the final output.
- Adjust prompts, tools, or output options until the agent consistently makes the decisions you want.
When the workflow run completes, you should see the AI Agent step with a detailed execution trace and an output payload (text or JSON) that later steps can use.
Configure the agent: build your own vs templates
Use tabs below to pick the setup path that matches how much customization you need.
Creating your own AI Agent inside a workflow gives you full control over instructions, tools, and outputs.
-
Define the agent's role and goals
- In the AI Agent settings, give the agent a clear role, such as "You are a CRM agent that qualifies inbound leads based on their responses and existing CRM data."
- Explain what success looks like: for example, "Decide whether this lead is hot, warm, or cold, and suggest the next best follow-up step."
-
Connect CRM context and tools
- Enable tools that let the agent search contacts, look up deals, or check appointment history.
- For each tool, use the let AI decide all field values toggle if you want the agent to fill everything in, or turn it off and lock specific fields that should always follow your own rules.
-
Set output format
- Choose text output when a later step only needs a message, summary, or instruction for a human or another system.
- Choose JSON output when you want to route leads or set fields based on the agent's decision, such as
{ "lead_status": "hot", "owner": "sarah" }.
-
Leverage conversational memory
- Keep the agent's memory enabled so it maintains a concise summary of prior runs for the same contact or deal.
- Use that memory to instruct the agent to stay consistent in tone, offers, or next steps when it sees the same person again.
Templates are the fastest way to try the AI Agent action with proven patterns.
-
Pick a template
- In the AI Agent settings, select a template such as Lead follow-up, No-show recovery, or another prebuilt scenario.
- Review the template description so you understand which trigger and CRM data it expects.
-
Customize instructions lightly
- Adjust wording to match your brand voice or product specifics while keeping the core logic of the template intact.
- Add any guardrails you care about, such as "Never offer a discount larger than 10 percent" or "Always prioritize existing customers."
-
Map CRM fields and tools
- Confirm which CRM objects the template uses (contacts, deals, appointments) and map them to your workspace if needed.
- Decide which fields are safe for the AI to set automatically and which fields you want to lock.
-
Set output usage
- For follow-up templates, route the agent output directly into email, SMS, or task-creation steps.
- For recovery or re-engagement templates, use JSON output so downstream branches can act differently based on the agent's classification of the situation.
Write prompts like you speak to a skilled teammate. Use clear goals, examples of good and bad behavior, and mention CRM concepts directly, such as "If the contact has an assigned owner, keep it; otherwise assign to Sarah." Typos and informal language are fine — the agent understands natural instructions.
Example: form submission that the agent handles
A common setup is:
- Trigger: Form submitted for a "Request a demo" form.
- Action: AI Agent evaluates the form responses plus existing CRM data.
- Output: JSON with fields like
lead_status,priority, andnext_step. - Next steps: Branch the workflow based on those fields, such as:
- Hot leads → create a task and assign the deal to Sarah.
- Warm leads → enroll in a nurture sequence.
- Cold leads → log for future remarketing.
This pattern replaces many nested conditions with one AI Agent action that keeps learning from context and past interactions.
Try it now
Start by adding a single AI Agent action to an existing workflow that already works with deterministic rules, such as your main inbound lead flow.
Once you see the detailed traces and structured outputs, expand usage to more complex journeys like no-show recovery, multi-step follow-up, and routing across teams.
Last updated 1 day ago
Built with Documentation.AI