Autopilot populates and updates Salesforce fields automatically using AI extraction from call transcripts and emails. It is the backbone of methodology enforcement and CRM hygiene in Momentum.
Three variants — choose based on trigger
| Variant | Trigger / Use Case |
|---|
| Autopilot Classic | Trigger: immediately after a call ends. Analyzes: the just-completed call transcript. A lookback period can also be configured to extend analysis beyond the latest call. Best for: per-call data — next steps, objections, MEDDPICC fields, sentiment from THIS call. |
| Retropilot | Trigger: a Salesforce record created or record updated event — including any specific field being updated (e.g., stage change, renewal date change, custom field update). Analyzes: conversational data within a customer-defined time window going back in time from the trigger date. Best for: comprehensive summaries at deal milestones — renewal prep, QBR, stage-based rollups. |
| Autopilot Batch | Trigger: manual — runs on a defined set of records. Analyzes: historical data across all matched records within the specified time window. Best for: data cleanup, backfilling new methodology fields, campaign enrichment, or a one-off analysis — e.g., identifying deals where the prospect raised a specific pain point now that your company has released a solution for it. |
Field type rules — this is where most errors happen
Autopilot auto-adapts output based on the Salesforce field type. Your prompt must be written differently depending on the target field type.
Picklist fields
Picklist rule: exact match only. Output must be one value from the allowed picklist — character-for-character. No explanation. No surrounding text. If the correct value cannot be determined: output nothing (never guess). A wrong picklist value is worse than an empty field — it corrupts CRM data.
- Define each value, don’t just list it. The AI needs to know what each option means, not just what it’s called. A label like “Partner’s End User” is ambiguous without a definition.
- Add few-shot examples. For classification tasks, showing 3–5 transcript snippet → output pairs is one of the most reliable ways to improve consistency. The AI learns the pattern from the examples rather than inferring it from your rules alone.
- Specify whose statements count. Rep statements, hypotheticals, future plans, and guesses don’t qualify. Only explicit statements from the customer or prospect should drive classification.
- Write tiebreaker rules. When two values could apply, tell the AI which wins. When things are ambiguous or contradictory, always resolve to your null value (None, Unknown, etc.) rather than guessing.
ROLE
You are a [Company] CRM analyst. Your job is to review call transcripts and
accurately classify [field topic] based on what the customer explicitly states.
OBJECTIVE
Analyze the transcript and select the single best matching value from the list
below. If the transcript does not clearly state the answer, select "None."
CONTEXT
[1–2 sentences explaining your business and why this field matters.]
Use these values exactly as written:
[Value 1]: [Definition — what this value means in your context.]
[Value 2]: [Definition — what this value means in your context.]
None: The transcript does not clearly state this, or the information is too
ambiguous to classify confidently.
RULES
- Select ONLY one value. Output only the value and nothing else.
- Only classify based on what the customer/prospect explicitly says.
- Do NOT classify based on rep descriptions, guesses, hypotheticals,
recommendations, or future-state plans unless the customer confirms it.
- If the transcript is ambiguous, contradictory, or conditional: select None.
- If the rep states something but the customer does not confirm it: select None.
- If multiple values could apply, prioritize [define your tiebreaker logic].
OUTPUT FORMAT
Respond ONLY with the single final value. No commentary, no explanation,
no punctuation, and no additional text.
EXAMPLES
Example 1:
Transcript: "[Customer quote clearly matching Value 1]"
Output: [Value 1]
Example 2:
Transcript: "[Customer quote clearly matching Value 2]"
Output: [Value 2]
Example 3:
Transcript: Rep asks. Customer says "We're still figuring that out."
Output: None
Textarea fields
Textarea rules: plain text, no preamble. No preamble: never start with “Based on the transcript…” Prefer near-verbatim language from the transcript. Always define the absent-data behavior — see the Overview for silence rule patterns including how to prevent trailing statements. ROLE
You are a sales intelligence analyst extracting [field topic] from a call transcript.
GOAL
Extract [specific information — e.g., the customer's stated pain points].
RULES
- Include: [what qualifies — be specific]
- Exclude: [what doesn't qualify]
- Use language close to what the customer actually said
- No preamble, no explanation, no trailing statement
- If [information] is not discussed: output nothing
OUTPUT FORMAT
[Plain text. Bullet list / 2–3 sentences / single value — specify.]
Field type quick reference
| Field Type | Prompt Rules |
|---|
| Picklist | List all allowed values in RULES. Return one value — exact match. Silence = output nothing. |
| Multipicklist | List all allowed values. May return more than one, separated by semicolons (;). Specify in OUTPUT FORMAT. Silence = output nothing. |
| Textarea (short) | 1–2 sentences. No preamble. Silence behavior required — see Overview. |
| Textarea (long) | Structured bullet list or paragraph. Specify max length. Verbatim language preferred. |
| Rich Text | Plain prose only. The field renders formatting, but raw markdown symbols (**, ##) will appear as literal characters — not bold or headers. Use line breaks for structure. No preamble. |
| Boolean (checkbox) | TRUE or FALSE only. Define exactly what constitutes TRUE. Silence = FALSE. |
| Number | Numeric value only. No units unless field expects them. Silence = output nothing. |
| Date | YYYY-MM-DD format. Extract from explicit mentions only. Do not infer. Silence = output nothing. |
| Reference (Lookup) | Exact record name as it exists in Salesforce. Only when clearly identifiable. If not found: return nothing. Important: Autopilot can only populate lookup fields referencing Contact records (e.g. Economic Buyer). Lookup fields to other objects, such as Accounts or Custom Objects, are not supported at this time. |
| Save Behavior | When to Use It |
|---|
| Confirm to Write | Weeks 1–2. Human reviews before saving. Validate prompt quality before automating. |
| Write if Empty | Standard operation. Only populates blank fields. Recommended default for most fields. |
| Automatic Write | High-confidence, frequently-changing data. Use after validating prompt output quality. |
Autopilot examples
The examples below show both prompt styles where it applies. For Textarea fields, a plain-language prompt often works well to start. For Picklist fields, the structured format is strongly recommended — value definitions, source scoping, and tiebreaker rules each directly improve classification accuracy in ways plain language can’t reliably replicate.
Example 1 — Next Steps (Textarea, Classic)
ROLE
You are a sales operations analyst capturing agreed next steps from a call.
GOAL
Extract all next steps or action items explicitly agreed upon during this call.
RULES
- Include: items with a clear owner and a stated or implied timeline
- Exclude: vague statements like "we'll be in touch" with no specific action
- Format each as: [Owner]: [Action] — [Timeline if stated]
- If no next steps were agreed: return absolutely nothing — no output, no explanation,
no placeholder text. Complete silence is the correct output.
- Outputting any text when no qualifying data exists is a critical error.
- Do not append any statements about what was not found or not discussed.
Your output must contain only confirmed next steps and nothing else.
OUTPUT FORMAT
Bullet list. Plain text.
Natural language equivalent — same task, simpler prompt. For a straightforward extraction like next steps, a plain-language version often gets you 80% of the way there and is a good starting point before adding rules:“Extract any next steps agreed on this call, including who owns them and when. Format as a bullet list: [Owner]: [Action] — [Timeline]. If nothing was agreed, return nothing.”Add structure when you see the AI including vague commitments, rep-stated actions the customer didn’t confirm, or adding trailing statements when nothing was found.
Example 2 — Site Management Model (Picklist, Classic)
ROLE
You are a CRM analyst. Your job is to review transcripts of sales calls and
accurately identify whether the customer manages sites after they are built,
or whether their end users manage their own sites.
OBJECTIVE
Analyze the transcript and determine which site management model the customer
explicitly states they are currently using. Select the single best matching
value from the list below. If the transcript does not clearly state the current
model, select "None."
CONTEXT
Understanding who manages sites after launch is important for implementation
planning, support design, and account strategy.
Use these values exactly as written:
Partner: The customer retains ownership and ongoing management of the site
after it is built. Their team handles updates, edits, maintenance, or
changes for the end user.
Partner's End-User: After the site is built, the end user takes over and
manages their own site directly.
None: The transcript does not clearly state who manages sites after launch,
or the information is too ambiguous to classify confidently.
RULES
- Select ONLY one value. Output only one value and nothing else.
- Only classify based on what the customer/prospect explicitly says about
their current post-launch model.
- Do NOT classify based on rep descriptions, guesses, hypotheticals,
recommendations, or future-state plans.
- If the rep states the model but the customer does not confirm it: select None.
- If the transcript describes both build and handoff, prioritize who manages
the site after launch.
- If the transcript is ambiguous, contradictory, or conditional: select None.
OUTPUT FORMAT
Respond ONLY with the single final value. No commentary, no explanation,
no punctuation, and no additional text.
EXAMPLES
Example 1:
Transcript: "After we launch the site, our team handles all updates for the client."
Output: Partner
Example 2:
Transcript: "Once the site is live, we hand it off and the client manages everything."
Output: Partner's End-User
Example 3:
Transcript: No site management model mentioned by the customer.
Output: None
Example 4:
Transcript: Rep asks "So your customers manage their own sites?"
Prospect: "We're still figuring that out."
Output: None
Example 3 — Closed Won Story (Textarea, Retropilot)
Triggered on Opportunity stage change to Closed Won. Analyzes the full conversation history within the deal window and writes a CRM-ready win story to a dedicated Salesforce field.
[ROLE]
You are a Senior Sales Strategy expert and elite CRM data specialist with deep
expertise in win/loss analysis. You are exceptionally precise, detail-oriented,
and committed to data integrity. You specialize in distilling sales conversations
into concise, high-signal CRM entries for the Closed Won Story field, capturing
exactly why the deal was won and how it was won.
[CONTENT TO USE]
Review the call transcripts provided and extract CRM-ready content by applying
the [CLOSED WON FRAMEWORK].
[GOAL]
Analyze the transcripts and produce accurate, specific, and story-driven CRM
data for the Closed Won Story field in Salesforce. Your output must answer two
core questions: why did they buy, and how did they buy — based only on
information explicitly stated in the transcripts.
[RULES]
- Be thorough in your analysis.
- Keep your answer to 3 sentences or fewer.
- If there is no sufficient answer, return an empty string.
- You MUST keep your answer to 1,500 characters or below.
- Adapt your analysis to the prospect's specific industry, business model, and
deal context.
- Prioritize the details most directly tied to the final buying decision.
- Produce an output that could stand alone as a complete win story for a sales
leader who was not on the call.
- You MUST avoid vague generalizations such as "good fit," "liked our product,"
or any language that does not provide actionable deal intelligence.
- Capture specific details: why they chose to buy, what pain or trigger drove
the decision, which capabilities or differentiators were decisive, who the
key decision-makers were, what the competitive or status quo alternative was,
and what ultimately pushed them to close.
[CLOSED WON FRAMEWORK]
Capture why the opportunity was won and how it was won — including the specific
pain points, triggering events, differentiators, decision dynamics, and
competitive context that led the prospect to choose your solution over all
alternatives.
Key areas to listen for:
- The primary pain point, business issue, or triggering event
- The key decision-maker, champion, or internal buying influence
- The capabilities, differentiators, or proof points that won confidence
- The competitive alternative, incumbent solution, or status quo displaced
- The urgency driver: deadline, growth pressure, or strategic initiative
- Any pricing, packaging, implementation, or commercial factor that helped close
[EXAMPLE FORMAT TO FOLLOW]
CFO-led; 3-system reconciliation post-acquisition created urgency to close
books 5 days faster. Beat incumbent on native integrations and onboarding
support. VP Ops was key champion.
Example 4 — Closed Lost Reason (Picklist, Batch)
Run manually across a set of closed-lost records to backfill a reason classification field — a classic Batch use case for retroactive data cleanup or methodology enforcement.
ROLE
As an expert in identifying and analyzing reasons for lost deals based on
aggregate account interactions, your responsibility is to determine the primary
motive that led to the failed sales deal.
RULES
- You MUST present the closed lost reasoning as one of the pre-established
options in the picklist without providing any additional explanation.
- You MUST select from the predefined options.
- You MUST exercise considerable discretion in not selecting the "Other" option.
OPTIONS
- Budget: The deal didn't close due to the prospect's monetary constraints.
- Competition: The competition offered a better product/service or pricing.
- Lost Momentum: The sales process stagnated and didn't progress.
- Authority: The wrong person was targeted and did not have decision-making power.
- Don't see Value: The prospect didn't perceive the value of our product/service.
- Unresponsive: Failed to effectively communicate or get a response from the prospect.
- Timing: The proposal or close didn't occur at an appropriate time.
- Feature: The absence or inadequacy of a feature in the product/service.
- Bad Fit: The product/service didn't fit well with the prospect's requirements.
- Rejected by Sales: The sales team prematurely deemed the prospect unsuitable.
- Other: Any other reason beyond the ones listed above.
EXAMPLE OUTPUT
Unresponsive
Autopilot — QA checklist
- Correct variant: Classic for per-call, Retropilot for event-triggered history, Batch for bulk or one-off
- Picklist: allowed values listed, output = exact value only
- Multipicklist: semicolon separator specified in output format
- Textarea / Rich Text: no preamble, no markdown symbols (they render as literal characters in Rich Text fields), silence behavior defined
- Silence rule explicitly stated
- No over-extraction (inferring content not in the transcript)
- Save behavior matches current confidence level
