Review deals and your forecast with agents

Prev Next

Who can use this: Any team member

Available on: Any Gong plan

Ideal for: Sales manager, RevOps

Required permissions: Manage company agents, skills, and automations

Your forecast is only as good as the deals behind it, and most of what describes a deal is typed into the CRM by the rep who owns it. A deal can sit untouched for weeks and still count toward the quarter. A rep can move a deal to a later stage before the buyer has agreed to anything. A manager may not find out until the forecast call, when there is little time left to act.

You can have agents check your pipeline against what buyers actually said on calls and in emails. You get a nudge on deals that have gone quiet, a check on each stage move, and an honest read on the deals that matter most.

Gong has an agent template for each of these agents, so you can start from a working agent and adjust it to your process:

Agent

What you get

When it runs

Stalled deal nudge

An email to the deal owner and their manager that explains why the deal looks stalled and suggests one next step, so they can move it forward or take it out of the forecast.

By itself, 60 days after the last activity on an open deal.

Deal field verifier

An email to the manager that says whether the buyer's words support the deal's new stage, and what the buyer said about timing and commitment, each backed by a dated quote.

By itself, each time an open deal changes stage.

Deal health inspector

A health score, the buying signals and risks with evidence, one to three next actions, and a forecast recommendation.

When you run it on a deal, for example before a forecast call or a 1:1.

Note:

The stalled deal nudge and the deal field verifier are automatically triggered agents, so their Generate with AI step consumes Gong credits each time it runs. The deal health inspector is a manually triggered agent and doesn't consume credits. Learn more about agents and credits.

Flag stalled deals

Get a nudge on a quiet deal while there's still time to act on it. When 60 days pass with no activity on an open deal, this agent runs by itself, reads the last call or email, and emails the deal owner and their manager. For example, the email looks like this:

Example email

Subject: Stalled deal – Harbor Analytics

  • Why it looks stalled: On the last call, the buyer said the project was waiting on budget approval from finance. No call or email has been logged since.

  • Next step: Ask the champion whether the budget review has happened, and offer a short call with finance this week.

Before you start, check that your CRM records a last activity date on deals, because the agent counts the days from that date and can't run without it. In Salesforce, the template uses the deal's Last Activity field.

To build the agent:

  1. From the left sidebar, click Agents.

  2. Click New agent.

  3. Find the Stalled deal nudge agent template, and click Select.

  4. (Optional) To rename the agent, click the edit icon at the top left, and enter a new name.

  5. If the Date condition step has no field selected, click it, and select the field your CRM uses for a deal's last activity date.
    Note: A different date, such as the last stage change, changes what the agent measures.

  6. (Optional) To change when the agent runs, click the Date condition step, and enter a new number of days.

  7. Click the Generate with AI step, and review the Instructions. The template fills in:

Summarize why this deal appears stalled based on the last call/email, and suggest one concrete next step. If no calls, say so.
  1. Click the Send an internal email step. The template sends the email to the deal owner and the owner's manager. To send it to someone else, in Recipients, select CRM roles, team members, or both.

  2. Click Save.

Tip:

What counts as stalled depends on the deal. In discovery, 30 days may be too long, while 30 days in contracting may be normal. An agent has one trigger, so to use a different number of days for different segments or stages, duplicate the agent. On each copy, in the Date condition step, change the number of days, and click Add filter to limit it to one segment or stage.

To turn the agent on and check that it works, see Create an automatically triggered agent.

Check stage moves against what buyers said

Catch a stage move the buyer hasn't agreed to before it reaches the forecast call. Each time an open deal moves to a new stage, this agent runs by itself and checks the move against what the buyer said on calls and in emails. It emails the manager a verdict on the new stage (Supported, Challenge, or Insufficient Evidence) and what the buyer said about timing and commitment. The agent can see the stage change, but not the deal's current close date or forecast category, so the manager compares those with what the buyer said. For example, the email looks like this:

Example email

Subject: Deal risk check – Northfield Systems

  • Stage (Negotiation): Challenge. "We haven't seen pricing yet." (March 12)

  • Timing: "Legal review usually takes about six weeks for us." (March 12)

  • Commitment: No call or email mentions an approved budget.

To build the agent:

  1. From the left sidebar, click Agents.

  2. Click New agent.

  3. Find the Deal field verifier agent template, and click Select.

  4. (Optional) To rename the agent, click the edit icon at the top left, and enter a new name.

  5. Click the Generate with AI step. In Instructions, replace the template's text with the following instructions, which ask only for what the agent can check:

This opportunity just moved to a new stage. Check the move against what the customer said on the calls and emails for this opportunity.

1. Stage: Give a verdict on the new stage (Supported, Challenge, or Insufficient Evidence). Include a dated quote as evidence and any risk it points to.
2. Timing: Summarize what the customer said about timing, approvals, and when they expect to sign, with dated quotes.
3. Commitment: Summarize what the customer said about budget, decision makers, and how committed they are, with dated quotes.

If no calls are found, say so and stop. Don't invent details.

If your stage definitions and qualification criteria are attached, use them.
  1. (Optional) To have the agent judge the stage by your own rules, in the Generate with AI step, click Add files, and add your stage definitions and qualification criteria.

  2. Click the Send an internal email step. The template sends the email to the manager of the deal owner. To send it to someone else, in Recipients, select CRM roles, team members, or both.

  3. Click Save.

The agent doesn't change any fields in your CRM. To turn the agent on and check that it works, see Create an automatically triggered agent.

Inspect a priority deal

Run this agent from an opportunity page in Gong Assistant when you need an honest read before a forecast call, a deal review, or a 1:1 with the rep. It reads the deal's calls, emails, and stakeholder activity, and doesn't rely on the CRM stage. It returns:

  • The confirmed buying signals and the stated risks, each with evidence

  • Any stage or stakeholder assumption that no conversation has confirmed

  • A health score and the reason for it

  • One to three next actions most likely to move the deal forward

  • A forecast recommendation of Commit, Best Case, Pipeline, or At Risk, with the reasoning

To build the agent:

  1. From the left sidebar, click Agents.

  2. Click New agent.

  3. Find the Deal health inspector template, and click Select.

  4. Review the Instructions. The template fills in:

You are a deal inspection specialist. Your goal is to assess the real health of a specific sales opportunity, independent of its CRM stage.

Analyze the selected deal's calls, emails, and stakeholder activity.

Identify confirmed buying signals and stated risks, each with supporting evidence.

Flag any stage or stakeholder assumption that isn't actually confirmed in a conversation.

Score overall deal health and explain the score.

Recommend the next 1-3 actions most likely to move the deal forward.

State a forecast call: Commit, Best Case, Pipeline, or At Risk, with reasoning.
  1. (Optional) To have the agent score deals by your own rules, click Add files, and add your forecast category definitions or deal review checklist.

  2. If you added files, in Instructions, describe what each file contains and how the agent should use it.

  3. (Optional) To let everyone in your company use the agent, in the top right, click Personal, and select Company.

  4. Click Save.

To run the agent, see Run agents in the Gong Assistant.

More agents you can build

You can build more agents that watch your pipeline for other signs that a deal needs attention. Each one below says what the agent does and when it runs, and gives the instructions to paste. To build one, follow the steps in Create an automatically triggered agent, set the trigger described, and paste the instructions into the Instructions field of the Generate with AI step. Then add a Send a Slack message or Send an internal email step to deliver the result.

Close-date check agent

Find out whether a deal is ready for its close date while there is still time to act. A set number of days before the close date, or after it has passed, the agent reads the recent calls and emails and tells the deal owner whether the deal is on track, at risk, or unclear. It cites the moment from a call that supports its read, and gives one next step. If the calls don't show enough to judge, the agent says so.

  • Trigger: A date condition on the opportunity's close date, a set number of days before or after it

  • Notify: The deal owner, by Slack message

Role: You are a pipeline-hygiene assistant flagging an approaching close date.

Use: the account and its recent calls or emails, plus limited CRM context. Do not assume access to fields, boards, or systems beyond that.

Produce: a short read of whether the account looks ready for its close date, with one next step.

Focus on: what the recent calls show about readiness (agreed next steps, open blockers, stakeholder alignment) against the close date coming up, and the single most useful thing to do now. State the evidence (a specific moment) before each read. If the recent record does not show enough to judge readiness, say so plainly rather than guessing.

Readiness rubric: call it "on track" only if a next step past the close date has been explicitly agreed; "at risk" if a blocker was named; "unclear" if neither.

Format: a short readiness read that cites a moment or two from the calls, then a one-line recommended next step. Under 70 words.

Tone: plain and direct, written for the account owner.

Don't: invent information not in the call or email; assert a probability or a forecast number the calls do not support; invent blockers.

Stage-gate check agent

Check that a deal has earned its new stage. When a deal advances, the agent checks each requirement for that stage against the recent calls, and tells the rep and manager whether each one is covered, missing, or unknown, with a short quote or moment and one action to close each gap. Before you use it, add your team's requirements for each stage to the instructions, for example "a champion confirmed before Proposal". Without them, the agent has no bar to check against.

  • Trigger: A CRM change when a deal's stage changes

  • Notify: The rep and their manager, by Slack message or email

Role: You are a RevOps assistant checking methodology compliance at a stage gate.

Use: the recent calls on the account, plus limited CRM context. Do not assume access to fields, boards, or systems beyond that.

Produce: a short compliance read when the account advances a stage.

Focus on: each required element for that stage (for example prospect-verified outcomes, or a champion), the evidence for it on a recent call, and whether it was captured or is missing. If the recent calls do not show whether an element was met, mark it unknown rather than covered.

Format: one line per required element: the element, a short quote or moment, then covered / missing / unknown, then a one-line action to close each gap. Under 25 words per line.

Tone: plain and direct, written for the rep and manager.

Don't: invent information not in the call or email; require elements the team has not defined; mark something covered without evidence.

Loss-pattern alert agent

Let a sales leader step in before a deal slips away. When a known loss-pattern signal shows up on an open deal, such as a stalled decision, a budget freeze, a competitor gaining ground, or a lost champion, the agent sends the leader a short alert with a quote, the account context, why it matters now, and a suggested action. If the signal is weak or was resolved on the call, the agent says so instead of raising an alarm. This agent needs a published AI tracker that detects the signals you care about.

  • Trigger: An AI tracker that detects a loss-pattern signal on a call or email

  • Notify: A sales leader, by Slack message or email

Role: You are a deal-risk analyst watching open accounts for loss patterns.

Use: the call or email that triggered this, plus limited CRM context. Do not assume access to fields, boards, or systems beyond that.

Produce: a short alert that a known loss-pattern signal showed up on an open account, so a leader can step in early.

Focus on: which loss-pattern signal appeared (for example a stalled decision, a budget freeze, a competitor gaining ground, a lost champion), the exact quote or moment, and why it matters now. If the signal is weak or was resolved on the call, say so plainly and do not raise it.

Format: three or four bullets: the signal (with a short quote), the account context, why now, suggested leadership action. Under 60 words.

Tone: plain and direct, written for the leader.

Don't: invent information not in the call or email; assert a loss risk the call does not support; invent a competitor or a reason.

To win back deals that were lost, build an agent that reviews each lost deal 180 days after it closes and sends the deal owner a re-engagement plan. See Re-engage closed-lost deals with agents.

For agents for other teams, with their instructions, see the Custom Agents prompt library in Gong Academy.