Agents in Factors are purpose-built AI collaborators designed to automate the most grueling parts of your Go-To-Market (GTM) workflow. Instead of your team manually stitching together data from different silos, Agents handle the heavy lifting—surfacing actionable insights so your reps can focus on high-value engagement.
Today’s manual process of account enrichment is fragmented. Teams often bounce between Slack, CRM, and LinkedIn to find context. Agents bridge this gap by delivering the right information to the right person at the exact moment they need it.
How to Build Your First Agent
Getting started is straightforward. Navigate to the automation hub to begin:
Select Automations from the top navigation menu.
Click on Agents and select Create Agent.
Enter your specific Agent Instructions to define its goal.
Click on create agent, note that you need to publish agent to access it across the application and via the API
Writing Effective Agent Instructions
Agents function by leveraging specific Tools. To trigger a tool within your instructions, use the @ symbol. You can further refine the data by applying filters inside each tool.
Note: When using the Web Research tool, include a clear and specific search query, and explicitly reference it using the @Web Research tag
Currently, Factors supports six specialized tools:
Tool | Purpose & Data Provided |
Account Profile | Accesses firmographic data and specific account properties. |
Web Research | Scouts the live web for recent news, funding, or specific company updates. Use @ to enter the query that you want to add to the web research tool |
Account Journey | Tracks all touchpoints (G2, LinkedIn, Website visits) for a specific account. |
HubSpot Deal | Pulls real-time deal stages, amounts, and sales history. |
HubSpot Company | Retrieves CRM-level company records and owner information. |
Top Identified User | Surfaces the most active or relevant individual stakeholders for an account. |
Example: Use the @Web Research tool combined with @Account Journey to create a "Pre-call Brief" agent that summarizes why an account is spiking and what their recent news is.
Note:- LLMs can hallucinate when handling date-specific data. To improve accuracy, include this instruction in the timestamp section of your prompt.
TIMESTAMP RULES
NEVER compute epoch → UTC from scratch using mental arithmetic. LLM epoch math is unreliable and will produce wrong dates and times. INSTEAD — always use this anchor method:
REFERENCE ANCHOR: 2026-03-16 00:00:00 UTC = epoch 1,773,619,200 Step 1: seconds_into_day = event_epoch − 1,773,619,200 Step 2: If result is negative, subtract 86,400 per prior day. If result is ≥ 86,400, add 1 day per 86,400 seconds. Step 3: Convert seconds_into_day → HH:MM:SS Step 4: Output: 2026-03-16 HH:MM:SS UTC (adjusted for day offset)
VERIFICATION REQUIREMENT: Show your anchor subtraction explicitly in a scratchpad before writing the final timestamp. Do not skip this step. SELF-CHECK: 1773676066 − 1773619200 = 56,866s = 15h 47m 46s → 2026-03-16 15:47:46 UTC Use this as a known-good test case to verify your method is correct before computing any other timestamps in the same run.
SCRATCHPAD: Perform all anchor subtraction and verification internally. Do NOT include scratchpad, working, or intermediate math in the final output.
Deploying Your Agents
Once published, your Agents aren't just static bots—they become part of your data layer. You can deploy them across four key areas:
Alerts: Get AI-summarized insights delivered directly to Slack or Email when an account shows intent.
Workflows: Use Agent output to trigger automated sequences or CRM updates.
Column Properties: Add Agent-generated data (like "Pain Point Summary") as a column in your account views.
Segments and reports: Build sophisticated audiences and reports based on Agent findings through agent propertoes (e.g., "Accounts with recent leadership changes").
Via API : You can access the agent output via the APIs to build your own automations outside factors. Postman Request:
'https://api.factors.ai/afx/call/deal_won_story_slack_VISWz?domain=domain.com&response=json' \--header 'Authorization: your_token'