- Benchmark X 360

Build complexity Smart Support+ No database · All native Voiceflow tools · Variables + Conditions + Function tools + System tools

Seven enhancements at a glance

Each enhancement can be built in any order. The build details tab has step-by-step instructions for each one.

Enhancement 1

Severity classification and routing

Agent classifies issue as High/Medium/Low before resolving. High severity is intercepted in KB Mode Selection and routed directly to escalation.

New Playbook Updated Playbook

Enhancement 2

Device type personalisation

Agent asks device type once. All troubleshooting steps are tailored to that device — Windows/Mac/Dell/Apple.

New variable New Playbook

Enhancement 3

Web search fallback

Implemented in Session 5 as Tier 3 of the three-tier fallback chain. No action needed.

✓ Complete

Enhancement 4

Satisfaction rating at close

One-question 1–5 rating at session end. Rating is stored in the satisfaction_rating variable.

Buttons New variable

Enhancement 5

Language support (already done)

Implemented in Session 5. No changes needed here.

✓ Complete

Enhancement 6

Repeat issue detection

Included in the Enhancement 4 Playbook instruction paste. No separate action needed.

✓ Included in 4

Enhancement 7

Final Playbook paste — all enhancements combined

Single paste per Playbook consolidating all hashtag sections: #severity-check, #device-filter, #repeat-check, #satisfaction-rating, #cards-formatting.

Single paste Cards on

New variables to declare first

Before building any enhancement, add all four new variables in the Variables panel.

Variable	Default value	Description
severity	none	Issue severity classification: high, medium, or low
device_type	none	User’s device type captured once at session start
satisfaction_rating	none	User’s satisfaction score 1–5 captured at session close
is_repeat_issue	none	Whether the user has experienced this issue before: yes or no

Also add all four to the Set step at the start of the Support Workflow, resetting each to none alongside the existing variables.

Step-by-step build instructions

Click any enhancement to expand the instructions. Build in any order.

1 Severity classification and routing Low effort ▼

Add a Severity Triage Playbook after Language Selection. The severity routing is then handled inside the KB Mode Selection Playbook — no separate Condition block needed on the canvas.

Step 1 — Create Severity Triage Playbook

Left sidebar → Playbooks → + → Create new playbook

Name

Severity Triage

LLM Description

Classifies the user’s issue as high, medium, or low severity based on their description. Sets the severity variable. Does not ask the user to classify it themselves.

Instructions

# Goal Classify the issue severity and set {severity}. # Classification rules High — system completely down, cannot work at all, data loss risk, or security breach suspected. Medium — significant slowdown or partial failure, work is impaired but possible. Low — minor issue, workaround available, or cosmetic. # Instructions Based on the user’s issue description, classify severity as “high”, “medium”, or “low”. Set {severity} accordingly. Do not ask the user to classify it themselves. Do not explain the classification to the user. Just set the variable and exit.

Exit condition: severity must be set to high, medium, or low — not none or blank.

Step 2 — Add Severity Triage to the Support Workflow canvas

On the canvas, wire Severity Triage directly between Language Selection and KB Mode Selection. No Condition block needed — the routing happens inside KB Mode Selection’s instructions.

Step 3 — Intent: severity check in KB Mode Selection #severity-check

The KB Mode Selection Playbook needs a severity check at the very top of its instructions. When {severity} = “high”, it will deliver one firm closing message and exit immediately — no KB mode question, no resolution tiers. The complete instruction including this section is pasted in Enhancement 7. Note this hashtag so you understand what that section does when you see it.

The high severity path must exit cleanly to avoid the chat looping back. This is achieved by adding a dedicated Exit Condition to KB Mode Selection and wiring it to the End block on the canvas — covered in Step 4 below.

Step 4 — Add Severity High Exit Condition to KB Mode Selection Playbook

Open the KB Mode Selection Playbook → Exit Conditions → + Add condition. It currently has two: KB Mode Individual and KB Mode Combined. Add a third:

Field	Value
Name	Severity High
LLM description	This condition is satisfied only when severity has been set to “high”. The agent has delivered a closing message and the session must end immediately.
Required variable	severity — must equal “high”

On the canvas, the KB Mode Selection block will now show a third exit port: Severity High. Wire this port directly to the End block. This ensures the chat closes cleanly when severity is high, with no looping.

With this wiring in place: high severity → KB Mode Selection delivers closing message → exits via Severity High port → End block → session closes and appears in transcript.

2 Device type detection and personalisation Low effort ▼

Ask the user their device type once at session start. All troubleshooting steps in the KB are then filtered to that device.

Step 1 — Create Device Detection Playbook

Name

Device Detection

LLM Description

Asks the user what device they are using and sets the device_type variable. Runs once after language selection.

Instructions

# Goal Ask the user their device type and set {device_type}. # Instructions Ask exactly this: “What type of device are you using? a) Windows laptop or desktop (Dell or other) b) Mac laptop (MacBook) c) Mac desktop (iMac) d) Dell workstation e) Tablet or other device” Set {device_type} to the matching value: a) → “windows” b) → “macbook” c) → “imac” d) → “dell_workstation” e) → “other” Confirm with one sentence and exit.

Add exit condition requiring device_type to be set — not none or blank.

Step 2 — Intent: device type filtering in resolution Playbooks #device-filter

Both resolution Playbooks need an instruction to show only steps relevant to {device_type} when presenting KB results. The complete instruction including this section is pasted in Enhancement 7. Note this hashtag so you understand what that section does when you see it.

Step 3 — Add Device Detection to the Support Workflow

Insert the Device Detection Playbook block immediately after Language Selection, before Severity Triage. This ensures device type is captured upfront — before any severity routing or KB mode selection that might exit the flow early.

3 Web search fallback — already complete No action needed ▼

Web search fallback was fully implemented in Session 5 as Tier 3 of the three-tier resolution chain. Both resolution Playbooks already have the Web Search system tool enabled and the fallback priority instructions in place.

Nothing to do here. The complete chain — Primary KB → Fallback 1 KB → Web Search — is already wired and tested. This slot is kept in the sequence so participants can see where web search fits architecturally within the seven-enhancement design pattern.

4 Satisfaction rating at session close Very low effort ▼

This enhancement adds two behaviours to both resolution Playbooks. Rather than updating the Playbooks now, note both hashtags below — the complete single paste covering all changes is in Enhancement 7.

Intent: repeat issue check #repeat-check

At the very start of the resolution chain, the agent asks the user once whether this issue has happened before. If yes, it skips all three resolution tiers, tells the user a support ticket is being logged, and proceeds directly to session close. If no, it proceeds with the full resolution chain as normal.

Intent: satisfaction rating at session close #satisfaction-rating

After every session — whether resolved through the KB chain or escalated due to a repeat issue — the agent presents a 1–5 rating using Buttons. The selected value is stored in {satisfaction_rating} and the session closes with a confirmation message.

Do not paste anything here yet. Both behaviours are included in the complete instruction paste in Enhancement 7. Pasting here separately would create duplicate or conflicting instructions.

5 Language support — already complete No action needed ▼

Language support was fully implemented in Session 5. The Language Selection Playbook, the language variable, and the language instructions in both resolution Playbooks are already in place.

Nothing to do here. This slot is kept in the sequence to reinforce for participants that the seven enhancements map to a coherent design pattern — each one addresses a real gap in a production support agent.

6 Repeat issue detection — included in Enhancement 4 No action needed ▼

Repeat issue detection is already embedded in the full Playbook instruction block provided in Enhancement 4. When you replaced the Playbook instructions in Step 1 and Step 2 of Enhancement 4, the repeat issue check was included at the top of the resolution chain.

Nothing to do here. The repeat issue check — asking the user if this has happened before and routing accordingly — is live as part of the Enhancement 4 instruction paste. This slot is kept so participants can see it as a distinct design concept: conditional routing based on issue history.

7 Final Playbook paste — all enhancements combined Single paste per Playbook ▼

This is the single update step for all three Playbooks. Each paste consolidates every hashtag section noted in earlier enhancements: #severity-check, #device-filter, #repeat-check, #satisfaction-rating, and #cards-formatting. Replace all existing instructions — do not append.

Step 1 — Enable Cards system tool

Open Resolve Issue — Individual → System tools panel → confirm Cards is toggled on. Repeat for Resolve Issue — Combined. Cards is on by default so this is a quick check only.

Step 2 — Replace KB Mode Selection Playbook instructions

Open KB Mode Selection. Select all existing instructions and replace entirely with:

# Severity check [#severity-check] Before asking about KB mode, check {severity}. If {severity} = “high”: Tell the user exactly this: “This is a high severity issue. I am logging an urgent support ticket for the IT team to investigate. Please avoid shutting down affected systems and disconnect from the network if possible. The IT team will contact you shortly.” Exit immediately. Do not ask any follow-up questions. Do not offer further assistance in this session. If {severity} = “medium” or “low” or “none”: Proceed to ask the user which KB mode they prefer. # KB mode question Ask the user exactly this: “How would you like me to search for a solution? A) Search five individual topic guides separately B) Search the combined cross-reference guide” Set {kb_mode} = “individual” or “combined” based on the user’s choice. Confirm with one sentence and exit.

Step 3 — Replace Resolve Issue — Individual Playbook instructions

Open Resolve Issue — Individual. Select all existing instructions and replace entirely with:

# Goal Resolve the user’s technical issue by working through a structured three-tier resolution chain. # Repeat issue check [#repeat-check] Before searching for a solution, ask once using Buttons: “Have you experienced this exact issue before?” → “Yes, this has happened before” → “No, this is the first time” Set {is_repeat_issue} = “yes” or “no”. If {is_repeat_issue} = “yes”: Skip all resolution tiers. Tell the user: “Since this has happened before, I am logging a support ticket for the IT team to investigate directly.” Proceed to Session close below. If {is_repeat_issue} = “no”: Proceed with full resolution chain below. # Resolution chain — follow this order exactly ## Tier 1 — Primary KB search Search the Knowledge Base using the KB tool. Apply the metadata filter: kb_type = individual When providing troubleshooting steps, show only the steps relevant to {device_type}. [#device-filter] If a relevant answer is found: provide it as a clear step-by-step response. Ask: “Does this resolve your issue?” If yes: proceed to Session close. If no: proceed to Tier 2. ## Tier 2 — Fallback KB search (external source) Search the Knowledge Base again using the KB tool, but this time query with kb_type = fallback1. If a relevant answer is found: provide it and note: “I found this in an extended reference guide.” Ask: “Does this resolve your issue?” If yes: proceed to Session close. If no: proceed to Tier 3. ## Tier 3 — Web search Search the web using the Web Search tool. If a relevant answer is found: provide it and note: “I found this from a public web source — please verify these steps with your IT team before applying.” Ask: “Does this resolve your issue?” If yes: proceed to Session close. If no: tell the user: “I was unable to find a solution in my available sources. I am logging a support ticket for the IT team to investigate.” Proceed to Session close. # Session close [#satisfaction-rating] Ask: “How satisfied are you with this support session?” Use Buttons: 1 — Very dissatisfied 2 — Dissatisfied 3 — Neutral 4 — Satisfied 5 — Very satisfied Set {satisfaction_rating} to the number selected. Close with: “Thank you for the feedback. Your ticket reference has been noted. Goodbye!” # Response formatting [#cards-formatting] When providing multi-step troubleshooting answers with three or more steps, use Cards to present the response. For each card: – Title: the step name (e.g. “Step 1 — Restart device”) – Description: the action to take in 1–2 sentences – No image – No button Do not attach source URLs or KB references to cards. Use plain numbered text only for single-step answers or confirmations. Respond in {language} throughout. Keep all technical terms in English.

Step 4 — Replace Resolve Issue — Combined Playbook instructions

Open Resolve Issue — Combined. Select all existing instructions and replace entirely with:

# Goal Resolve the user’s technical issue by working through a structured three-tier resolution chain. # Repeat issue check [#repeat-check] Before searching for a solution, ask once using Buttons: “Have you experienced this exact issue before?” → “Yes, this has happened before” → “No, this is the first time” Set {is_repeat_issue} = “yes” or “no”. If {is_repeat_issue} = “yes”: Skip all resolution tiers. Tell the user: “Since this has happened before, I am logging a support ticket for the IT team to investigate directly.” Proceed to Session close below. If {is_repeat_issue} = “no”: Proceed with full resolution chain below. # Resolution chain — follow this order exactly ## Tier 1 — Primary KB search Search the Knowledge Base using the KB tool. Apply the metadata filter: kb_type = combined When providing troubleshooting steps, show only the steps relevant to {device_type}. [#device-filter] If a relevant answer is found: provide it as a clear step-by-step response. Ask: “Does this resolve your issue?” If yes: proceed to Session close. If no: proceed to Tier 2. ## Tier 2 — Fallback KB search (external source) Search the Knowledge Base again using the KB tool, but this time query with kb_type = fallback1. If a relevant answer is found: provide it and note: “I found this in an extended reference guide.” Ask: “Does this resolve your issue?” If yes: proceed to Session close. If no: proceed to Tier 3. ## Tier 3 — Web search Search the web using the Web Search tool. If a relevant answer is found: provide it and note: “I found this from a public web source — please verify these steps with your IT team before applying.” Ask: “Does this resolve your issue?” If yes: proceed to Session close. If no: tell the user: “I was unable to find a solution in my available sources. I am logging a support ticket for the IT team to investigate.” Proceed to Session close. # Session close [#satisfaction-rating] Ask: “How satisfied are you with this support session?” Use Buttons: 1 — Very dissatisfied 2 — Dissatisfied 3 — Neutral 4 — Satisfied 5 — Very satisfied Set {satisfaction_rating} to the number selected. Close with: “Thank you for the feedback. Your ticket reference has been noted. Goodbye!” # Response formatting [#cards-formatting] When providing multi-step troubleshooting answers with three or more steps, use Cards to present the response. For each card: – Title: the step name (e.g. “Step 1 — Restart device”) – Description: the action to take in 1–2 sentences – No image – No button Do not attach source URLs or KB references to cards. Use plain numbered text only for single-step answers or confirmations. Respond in {language} throughout. Keep all technical terms in English.

Replace, do not append. Select all existing content in each Playbook’s instruction field before pasting. This clears any Zendesk references or outdated instructions from earlier sessions.

All five hashtag sections are now live across both resolution Playbooks and KB Mode Selection: #severity-check · #device-filter · #repeat-check · #satisfaction-rating · #cards-formatting

Updated workflow structure after all seven enhancements

The consolidated Support Workflow now includes all new blocks. New blocks added by Session 6 are shown in purple.

Support Workflow (consolidated)

Block 1 — Start trigger Block 2 — Language Selection Playbook [Session 5] Block 3 — Device Detection Playbook ← Enhancement 2 Block 4 — Severity Triage Playbook ← Enhancement 1 Block 5 — Set step (reset all variables including 4 new ones) Block 6 — KB Mode Selection Playbook [Session 5 — updated with #severity-check] ├─ Exit: Severity High → End block directly ← Enhancement 1 ├─ Exit: KB Mode Individual → Block 7a onwards └─ Exit: KB Mode Combined → Block 7b onwards Block 7 — Message step (Scripted) Block 8 — Resolve Issue Playbook (enhancements 4, 5, 6, 7 embedded in instructions) Block 9 — End step

Complete variable reset Set step

Variable	Reset value	Added in
kb_mode	none	Session 4
language	en	Session 5
severity	none	Session 6
device_type	none	Session 6
satisfaction_rating	none	Session 6
is_repeat_issue	none	Session 6

Test cases for Session 6

Run each test after building the corresponding enhancement.

Test 1 — Severity routing

I think we have been hacked. Someone has accessed our customer database without authorisation. I cannot access any of my files now.

Expected: severity = high. KB Mode Selection intercepts the issue, skips KB mode question, and tells the user a support ticket is being logged directly. No resolution tiers attempted.

Test 2 — Device personalisation

Select MacBook as your device. Then ask:

My screen is flickering and I cannot connect my external monitor.

Expected: Steps shown are Mac-specific only (System Preferences, Detect Displays, SMC Reset). No Dell or Windows steps appear.

Test 3 — Repeat issue detection

Start a session with any issue. When asked if the issue has happened before, answer yes.

Expected: Agent skips all resolution tiers, tells the user a support ticket is being logged for the IT team, and proceeds directly to the satisfaction rating.

Test 4 — Satisfaction rating

Complete any support session through to the end.

Expected: Agent presents the five-option rating Buttons at session close. Selecting a rating sets {satisfaction_rating} and the session closes with the confirmation message.

Test 5 — Cards formatting

My laptop is running very slowly and I also cannot connect to the company VPN.

Expected: Response is presented as visual cards — one per major troubleshooting stage — instead of a numbered text list.

TechSupport Agent — Session 6 Upgrade Guide