AI Sales Agent - Interactive Workflow

AI Sales Agent: Interactive Workflow

Welcome to the interactive workflow designer. This application translates the analysis of customer "closure" replies into a tangible, explorable workflow. The goal is to help you build an accurate and responsive AI sales agent.

New: Multi-Intent Logic

This workflow is now designed for a multi-intent system. A customer email (e.g., "Send me a quote and a demo") will be classified into a prioritized list of intents (e.g., `["DEMO", "PRICING"]`).

Your bot's logic should use the first intent (`intents[0]`) to define the main call-to-action. Then, it should loop through any other intents in the list to add secondary paragraphs that answer the customer's other questions.

Click the tabs on the left to see the workflow for each intent. The prompt below is designed to provide your AI with the necessary classification.

Note: These workflows are designed for an email-based conversation. This means the AI already knows the customer's email, and responses are asynchronous.

Prompt for Intent Classification

Use the following system prompt for your LLM to classify an incoming customer email. The LLM should be instructed to respond *only* with the JSON object.

You are a specialized AI classification engine for a sales agent. Your sole purpose is to analyze a customer's email reply and return a prioritized list of all intents found.

Analyze the email based on its primary goals.

**Rules:**
1.  Identify ALL intents present in the email.
2.  Return a list of intents, sorted by priority (highest priority first).
3.  CRITICAL: Respond *only* with a JSON object in the format: {"intents": ["INTENT_ID_1", "INTENT_ID_2", ...]}
4.  If no intent is clear, respond with: {"intents": ["UNKNOWN"]}

**Intent List (in order of priority):**

--- 1. POSITIVE CLOSURE (HIGH VALUE) ---
* **PURCHASE**: Customer wants to buy or order.
    (e.g., "I want to order," "I'll take it," "Buy now," "Where do I pay?")
* **DEMO**: Customer wants a product demonstration.
    (e.g., "Want a demo," "Show me how it works," "Can I see it in action?")
* **MEETING**: Customer wants to schedule a general meeting.
    (e.g., "I want a meeting," "Let's talk," "Book a time.")
* **CALL**: Customer wants a human to call them.
    (e.g., "Want someone to call me," "Call me," "Have a rep contact me.")

--- 2. INFORMATION GATHERING (MEDIUM VALUE) ---
* **PRICING**: Customer is asking for price or a quote.
    (e.g., "What's the price?," "How much?," "Send a quote.")
* **QUESTION**: Customer asks a specific, high-interest logistical/feature question.
    (e.g., "What's your support email?," "Where are you based?," "Does it integrate with X?")
* **RESOURCES**: Customer wants a brochure, whitepaper, or spec sheet.
    (e.g., "Can you send me a brochure?," "Do you have a whitepaper?")
* **COMPARISON**: Customer wants a comparison against a competitor.
    (e.g., "How are you better than [Competitor]?," "Send me a comparison chart.")

--- 3. OBJECTIONS & REJECTIONS (CRITICAL) ---
* **OBJECTION_PRICE**: Customer says it's too expensive or they have no budget.
    (e.g., "Too expensive," "Not in the budget," "That's a lot.")
* **OBJECTION_TIMING**: Customer is not ready *now*.
    (e.g., "Not right now," "Ask me in 6 months," "We're in planning.")
* **OBJECTION_COMPETITOR**: Customer is already using or happy with a competitor.
    (e.g., "I'm using [Competitor]," "We are happy with our current solution.")
* **REJECTION**: Customer is explicitly not interested.
    (e.g., "No thanks," "Not interested," "Please remove me from your list," "Unsubscribe.")

--- 4. CONVERSATIONAL & STATUS ---
* **CONDITIONAL**: Customer agrees, but with a new condition.
    (e.g., "OK, but send me the details first," "Yes, after I see the specs.")
* **AFFIRMATION**: Customer gives a simple "yes" to a bot's proposal.
    (e.g., "OK," "Confirmed," "Sounds good," "Perfect," "Proceed.")
* **VAGUE_POSITIVE**: Customer expresses happiness but no clear action.
    (e.g., "Great!," "Awesome," "Love it," "Nice.")
* **FOLLOW_UP**: Customer is asking for an update on a previous request.
    (e.g., "Any update on this?," "Just checking in," "Did you send the quote?")
* **CONFUSION**: Customer doesn't understand.
    (e.g., "I'm confused," "I don't get it," "What do you mean?")
* **OOTO**: Out of office, auto-reply.
    (e.g., "I will be out of the office...")

Master Bot Workflow Logic

This is the central logic for handling all incoming emails. The system is designed to handle both the **customer-facing reply** and the **internal actions** (like notifying a human) in parallel.

Customer Email Received

LLM Classification
{"intents": ["INTENT_1", ...]}

Workflow Engine: Parallel Processing

1. Construct Customer Reply

1. Get `intents[0]` (Primary)

2. Loop `intents[1...n]` (Secondary)

Define Main CTA

e.g., "Great! Here is the scheduling link..."

Add Secondary Paragraphs

e.g., "...and regarding your question about pricing..."

2. Process Internal Actions

1. Loop through *all* intents.

2. Check if intent needs human.

Human Handoff Required?

Yes

Notify Human Agent

(e.g., "CALL", "OBJECTION")

Log & Continue

(e.g., "PRICING", "RESOURCES")

Execute Actions

(1) Send constructed email to customer.

(2) Trigger internal notifications / CRM updates.

End Interaction

Intent: Direct Purchase

Customer Says:

"I want to order this," "I'll take it," "Where do I pay?"

Analysis:

This is the highest-value intent. The customer is ready to buy. The goal is to remove all friction and close the deal immediately.

Optimal AI Response & Action:

Action: Provide Checkout Link or Assign Human

"That's fantastic news! We're excited to have you on board. You can complete your order directly through our secure checkout: [Checkout/Order Link]"

(Alternative)

"That's fantastic news! To finalize your order, I have looped in my colleague [Human's Name] from our accounts team. They will be in touch shortly to confirm the details and process your payment."

Logic: Provide the most direct path to payment. If a link isn't possible, a warm handoff to a human is the next best step.

Intent: Request Demo

Customer Says:

"I want a demo," "Show me how it works," "Can you set up a live demo?"

Analysis:

Extremely high-value. A demo is a firm commitment of time and a strong buying signal. The goal is to get this scheduled *immediately*.

Optimal AI Response & Action:

Action: Provide Scheduling Link

"A demo is the best way to see the power of [Product]! Please use our booking link to find a time that works best for you and our specialists:"

[Your Scheduling Link (e.g., Calendly)]

"If you can't find a suitable time, please let me know, and I will have our team reach out directly."

Logic: Automate scheduling to remove back-and-forth. This is the #1 priority. A human handoff is the fallback.

Intent: Request Meeting

Customer Says:

"I want a meeting," "Let's talk," "Can we schedule a time to discuss?"

Analysis:

This is slightly more vague than a "demo" but still a very strong buying signal. It needs to be routed to the correct human (e.g., a sales rep or account executive).

Optimal AI Response & Action:

Action: Provide Scheduling Link

"Absolutely. We'd be happy to discuss your needs. The best way to do that is to book a time with one of our account executives. Please use the link below to find a time that works for you:"

[Your Scheduling Link for Sales]

Logic: Same as the demo. Route to a calendar to avoid email tag. Ensure this link goes to a person, not just a generic demo.

Intent: Request Call

Customer Says:

"Want someone to call me," "Here's my number, call me," "Can you have a rep reach out?"

Analysis:

The customer wants immediate, human-to-human contact. This is an urgent, high-priority lead. The AI cannot solve this; it must route it for immediate action.

Optimal AI Response & Action:

Action: Create CRM Task & Confirm Receipt

"Thank you. I have personally notified our team, and a representative will be calling you shortly."

(If they provided a number): "They will reach you at [Customer's Phone Number]."

(If they did NOT provide a number): "To ensure we can connect, could you please confirm the best phone number for us to use?"

Logic: The AI's job is to (1) Create an URGENT task in the CRM for a human, and (2) Reassure the customer that their request has been received and is being acted upon.

Intent: Pricing Inquiry

Customer Says:

"What's the price?," "How much does this cost?," "Send me a quote."

Analysis:

A standard, high-interest buying question. The response depends on your business model (public pricing vs. custom quotes).

Optimal AI Response & Action:

Action (If Public Pricing): Provide Link

"You can find our full pricing details right here on our pricing page: [Link to Pricing Page]."

Action (If Custom Quote): Ask Qualifying Questions

"We provide custom quotes to ensure you're only paying for what you need. To build your quote, I just need to know two things:"

"[Qualifying Question 1, e.g., How many users?]"
"[Qualifying Question 2, e.g., Which features are most important?]"

"Alternatively, we can discuss this in a brief demo. [Schedule Demo Link]"

Logic: Never just say "it depends." Either provide the price directly or *immediately* begin the process of gathering the data needed for a quote.

Intent: Specific Question

Customer Says:

"Does it integrate with Salesforce?," "What's your support like?," "Is it on-prem or cloud?"

Analysis:

The customer is "in the weeds." This is a great sign. They are seriously evaluating the product. The AI should try to answer this from its knowledge base.

Optimal AI Response & Action:

Action: Answer from Knowledge Base & Offer Next Step

"Great question. Yes, we have a native integration with Salesforce."

"You can read more about it in our documentation here: [Link to Docs]"

"This is actually a topic our specialists cover in the live demo, as it's a very popular feature. [Schedule Demo Link]"

Logic: (1) Answer the question. (2) Provide proof (a doc link). (3) Use the question as a pivot to a higher-value action (a demo).

Intent: Request Resources

Customer Says:

"Send me a brochure," "Do you have a whitepaper?," "Can I get a spec sheet?"

Analysis:

The customer is interested but may be in an earlier research phase. The goal is to fulfill the request but also try to accelerate them to a higher-intent action.

Optimal AI Response & Action:

Action: Provide Resource & Pivot to Demo

"Of course. Here is a link to our [Brochure/Whitepaper]: [Link to Resource]"

"While that document provides a great overview, the best way to see how [Product] can help you specifically is in a 15-minute live demo."

"You can schedule one here: [Schedule Demo Link]"

Logic: Always fulfill the request. But don't *just* fulfill it. Use it as an opportunity to "upsell" them from a passive (reading) to an active (demo) buying stage.

Intent: Request Comparison

Customer Says:

"How are you better than [Competitor]?," "Send me a comparison chart."

Analysis:

This is a strong buying signal. They have a shortlist. The AI must answer this confidently, focusing on strengths, not attacking the competitor.

Optimal AI Response & Action:

Action: Answer with Strengths & Pivot to Demo

"That's a great question. While [Competitor] is a good solution, our customers find we are much stronger in a few key areas:"

"[Strength 1, e.g., Ease of Use]: Our interface is consistently rated #1 for usability."
"[Strength 2, e.g., Support]: We provide 24/7 human support, which is something we're very proud of."

"I'd be happy to show you these differences in a live demo: [Schedule Demo Link]"

Logic: Confidently state 2-3 key differentiators. Do not get into a feature-by-feature war. Immediately pivot to a demo where a human can handle the nuances.

Intent: Objection (Price)

Customer Says:

"That's too expensive," "Not in the budget," "Your price is too high."

Analysis:

This is an objection, not a "no." The customer is engaging. The AI's job is to re-frame the conversation from *cost* to *value* and *ROI*.

Optimal AI Response & Action:

Action: Acknowledge & Pivot to Value (or Handoff)

"I understand. It's an important investment. Many of our best customers felt the same way, but found that [Product] saved them an average of [X hours / $Y] per month."

"We also have more flexible packages that might be a better fit. I've looped in my colleague [Human's Name] who is a specialist in finding the right solution for your budget. They will reach out shortly."

Logic: An AI should not negotiate price. (1) Empathize. (2) Re-frame with ROI. (3) Immediately hand off to a human sales rep who *can* negotiate or offer other packages.

Intent: Objection (Timing)

Customer Says:

"Not right now," "Ask me in 6 months," "We aren't ready yet."

Analysis:

This is a classic "stall." The AI's job is to accept this gracefully and set up a future follow-up, converting a "no" into a "not yet."

Optimal AI Response & Action:

Action: Accept & Schedule Nurture

"No problem at all. Thank you for the update. Timing is everything."

"If it's alright with you, I will make a note to check back in around [Date, e.g., 6 months from now]."

"In the meantime, you might find our monthly newsletter on [Topic] useful. [Newsletter Sub Link]"

Logic: (1) Be polite and respectful. (2) Get permission for a future follow-up (and log this in the CRM). (3) Offer a low-friction way to stay in touch (nurture).

Intent: Objection (Competitor)

Customer Says:

"I'm already using [Competitor]," "We're happy with our current solution."

Analysis:

This is a tough objection. The AI's job is not to argue, but to find a small crack. The goal is to highlight a key differentiator and ask a "no-risk" question.

Optimal AI Response & Action:

Action: Acknowledge & Differentiate

"That's great. [Competitor] is a solid product. We actually have many customers who've switched from them."

"They tell us our [Key Differentiator, e.g., AI automation] saves them 10+ hours a week, which they couldn't do before."

"If you're ever curious to see *just* that feature, I'd be happy to set up a quick 10-minute chat. No pressure."

Logic: (1) Acknowledge the competitor. (2) Use social proof ("many customers..."). (3) Isolate ONE key differentiator. (4) Offer a low-risk, high-value alternative to a full demo.

Intent: Rejection

Customer Says:

"Not interested," "No thanks," "Unsubscribe," "Please remove me from your list."

Analysis:

This is a clear "no." The AI's #1 job is to comply immediately and professionally. Failure to do so damages brand reputation and can violate spam laws.

Optimal AI Response & Action:

Action: Comply, Confirm, and Close

"Thank you for letting us know. I have removed you from our mailing list, and you will not receive any further outreach from me."

"Wishing you all the best."

Logic: (1) Action: Log "Do Not Contact" in the CRM. (2) Response: Confirm the action has been taken. Be polite, be professional, and end the conversation.

Intent: Conditional

Customer Says:

"OK, but send me the details first," "Yes, if you can send me a spec sheet," "I'll do the demo *if* you can confirm it integrates with [X]."

Analysis:

This is a "yes" with a hurdle. The AI must treat this as a multi-intent request: (1) Fulfill the condition. (2) Re-affirm the original action.

Optimal AI Response & Action:

Action: Fulfill Condition, then Re-state Action

"Happy to! Here is the spec sheet you requested: [Link to Resource]"

"Once you've had a chance to look that over, please feel free to use the scheduling link for your demo: [Schedule Demo Link]"

Logic: This is what the multi-intent system is for. The prompt will classify this as `["AFFIRMATION", "RESOURCES"]` (or `["DEMO", "QUESTION"]`). The bot must answer the secondary intent first, then immediately follow up with the primary call-to-action.

Intent: Affirmation

Customer Says:

"OK," "Confirmed," "Sounds good," "Perfect," "Please proceed."

Analysis:

This is a simple "yes" to a proposal the AI *just* made (e.g., "Can I send you the link?" -> "OK"). The AI needs to complete the action it proposed.

Optimal AI Response & Action:

Action: Execute the Proposed Action

"Great. Here is that resource I mentioned: [Link to Resource]"

"As I mentioned, the best way to see this in action is in a demo: [Schedule Demo Link]"

Logic: This is a continuation. The AI must remember the *context* of the last email, execute the action, and (if appropriate) pivot to the next logical step.

Intent: Vague Positive

Customer Says:

"Great!," "Awesome," "Love it," "Nice."

Analysis:

This is a positive "emotional" response, but not a "business" response. It contains no action. The AI must convert this positive energy into an action.

Optimal AI Response & Action:

Action: Acknowledge & Pivot to Action

"Glad to hear it!"

"The best way to see this in action is in a live demo. Would you be open to scheduling a 15-minute call?"

"[Schedule Demo Link]"

Logic: Don't just say "thanks." Acknowledge the positive sentiment and immediately channel it into the primary sales goal (demo, meeting, etc.).

Intent: Follow-Up

Customer Says:

"Any update on this?," "Just checking in," "Did you send the quote?"

Analysis:

This is a potential "drop." The customer is waiting for something. This is high-priority and needs to be routed to a human *immediately* to check the status.

Optimal AI Response & Action:

Action: Handoff to Human for Status Check

"Thank you for following up. I am checking on the status of your [quote/request] with the team right now."

"I have looped in my colleague [Human's Name] who will get back to you with an update shortly."

Logic: The AI cannot know the status of a human's task. It must (1) Reassure the customer, and (2) Immediately escalate to the human responsible for the original request.

Intent: Confusion

Customer Says:

"I'm confused," "I don't get it," "What do you mean?"

Analysis:

The AI has failed to be clear. Its last message was too complex. The goal is to simplify or hand off to a human who can explain better.

Optimal AI Response & Action:

Action: Simplify & Re-explain, or Handoff

"My apologies for not being clear. Let me try to simplify:"

"[Simple, one-sentence re-explanation of the last point]"

"This is often easier to explain in a quick call. I have notified my colleague [Human's Name] who can explain this much better than I can. They will be in touch."

Logic: An AI that confuses a customer should not try the same thing again. The safest and most helpful response is to escalate to a human.

Intent: Out of Office

Customer Says:

"I will be out of the office until [Date]..."

Analysis:

This is a simple system message. The AI should not reply to the customer. It should parse the return date and update the CRM.

Optimal AI Response & Action:

Action: NO REPLY. Log in CRM.

1. DO NOT REPLY TO THE EMAIL.

2. Parse Date: Attempt to find a return date (e.g., "until June 5th").

3. Update CRM: Pause all outreach to this contact until the return date.

Logic: Replying to an OOTO message is a dead giveaway of a simple bot. A smart agent logs the information and pauses communication to look human.