with

Interactive AI Enablement Experience for Fin, powered by Supersonik

Salesforce, in collaboration with Supersonik, launched an AI-powered enablement experience on September 9th to train SMB Account Executives on the newest addition to its portfolio: Fin. Designed to be thorough yet approachable, the experience helps AEs get up to speed through a self-serve journey, from understanding the product to confidently positioning and selling it. Along the way, AEs can explore specific capabilities in greater depth, ask questions about Fin, and take part in roleplays designed to prepare them for conversations with different types of buyers.

Key findings

1

246 Salesforce SMB AEs opted in during the first week. 78% of assessed reps already met the benchmark for identifying and positioning a Fin opportunity.

2

AEs used the experience to surface real-world selling questions, not just complete the training. In 63 of 257 sessions, reps proactively asked the agent specific questions about Fin.

3

Fin Insights is the feature that resonated most strongly, cited in 30 of the 60 sessions that captured what stood out to learners.

4

Procedures is the feature learners struggle with most, making it the clearest coaching priority. The agent frequently needed to reinforce how Procedures connect with Helpdesk and Copilot.

Demo structure

Sent to 7,860 SMB Account Executives at Salesforce and shared through Trailhead; opt-in, and open until 16 October 2026. The experience has a mandatory core enablement experience1, four optional capability deep dives2 and an optional roleplay3.

Structure of the Fin enablement experience: core enablement experience, four capability deep dives, roleplay
  1. 1 Core enablement experience: the guided walkthrough every learner is meant to complete, covering what Fin is, who it is for and its four capabilities (Helpdesk and Copilot, Procedures, Insights, Operator). It has four areas; completion is measured against them.
  2. 2 Capability deep dive: an optional module on one capability, chosen by the learner after the core enablement experience and graded 1 to 10 at its end.
  3. 3 Roleplay score: the learner sells Fin to a simulated buyer and is graded 1 to 10 on four independent pillars (objection handling, ICP and persona relevance, product understanding, value narrative). The session score is the sum of the pillars scored divided by the number of pillars scored; an undiscussed pillar is null, not 0.

Week one at a glance

Exhibit 1

Where learners stand after the first week

Learners4
246
Salesforce AEs who opted in this week; 11 came back for a second session
Valid sessions5
257
Average 15:04 per session, median 10:41; 65 hours of enablement in total
How much do they know about Fin?6iDemo score: the learner's understanding of the core enablement experience, scored 1 to 10. Returned as null when the learner has not given enough evidence to assess understanding, so a session with completion above 0 and no score progressed but was not assessed; it is not a low score. The cohort figure is the average over assessed sessions.
6.2out of 10
74 sessions assessed
How far did they get?7iCompletion: the share of the core enablement experience's four areas covered in a session, averaged over the sessions that covered at least one. Sessions that covered none are counted separately, not as zero.
82%
Completion of the core enablement experience across the 115 sessions
How well do they sell it live?3iRoleplay score: the learner sells Fin to a simulated buyer and is graded 1 to 10 on four independent pillars (objection handling, ICP and persona relevance, product understanding, value narrative). The session score is the sum of the pillars scored divided by the number of pillars scored; an undiscussed pillar is null, not 0.
6.2out of 10
15 sessions took the roleplay
Source: Supersonik insights for the Fin enablement agent on Trailhead, 11 to 18 September 2026. 365 sessions exported; 108 not marked valid excluded. Grades are the reviewed grading carried on each session record. Enabled8 means a demo score of 5 or higher; 58 of the 74 assessed sessions (78%) meet it.
  1. 4 Learner: a distinct Salesforce email address across the valid sessions.
  2. 5 Valid session: a session where there has been interaction between the agent and the learner. All counts in this report use valid sessions only.
  3. 6 Demo score: the learner's understanding of the core enablement experience, scored 1 to 10. Returned as null when the learner has not given enough evidence to assess understanding, so a session with completion above 0 and no score progressed but was not assessed; it is not a low score. The cohort figure is the average over assessed sessions.
  4. 7 Completion: the share of the core enablement experience's four areas covered in a session, averaged over the sessions that covered at least one. Sessions that covered none are counted separately, not as zero.
  5. 8 Enabled: a demo score of 5 or higher. A learner at that level should be able to recognise a Fin opportunity in an existing account and position it at a high level; from 7 upwards, to take Fin into a customer conversation.

In this report

1. Progress

Exhibit 2

How far learners went through the experience

115 of 257 sessions completed the core enablement experience; 15 took the roleplay and 10 a capability deep dive

115
15
10
Core enablement experienceEntered at least one of its four areas
RoleplaySold Fin to a simulated buyer
Capability deep divesChosen after the core enablement experience
Each part: sessions that reached it, and the average grade of those assessed
Core enablement experienceDemo score at the assessment
115sessions
6.274 graded
RoleplayLive selling scenario
15sessions
6.215 graded
Capability deep dives19 deep dives across four capabilities
10sessions
4.919 graded
Counts are sessions and milestones are not exclusive. The core enablement experience is graded at its assessment, so fewer sessions are graded than entered it. By capability: Helpdesk and Copilot 9 graded at 3.8, Procedures 4 graded at 5.2, Insights 3 graded at 5.0, Operator 3 graded at 7.7.
Source: coverage, capability and roleplay grades per session, 257 valid sessions. Columns count sessions that reached each milestone; the rows beneath add the grade of those assessed.

2. Quality

Scores run 1 to 10 on the shared rubric, based on how well the learner explained each checkpoint in their own words.

Exhibit 3

Demo score distribution

74 sessions assessed; 33 solid or better

9 to 10 out of 10, mastery: 15%7 to 8 out of 10, solid: 30%5 to 6 out of 10, partial: 34%3 to 4 out of 10, weak: 19%1 to 2 out of 10, absent: 3%74sessions
9 to 10 out of 10, mastery15%
7 to 8 out of 10, solid30%
5 to 6 out of 10, partial34%
3 to 4 out of 10, weak19%
1 to 2 out of 10, absent3%
Source: demo score per session6, 74 sessions assessed out of 257. Every score is grounded in evidence from the learner (what they said, asked or demonstrated), not in what the agent explained; listening to a complete explanation without engaging is not understanding. Weak understanding earns a low score, absence of evidence earns no score. Scores follow the grading anchors9; capability scores are never averaged into the demo score. 58 sessions (78%) sit at 5 or higher, the enabled bar8.
  1. 9 Grading anchors, shared by demo score, capability scores and roleplay pillars: 9 to 10 mastery (correct, specific and unprompted; ready to use with a customer), 7 to 8 solid (covers the key points with minor gaps or one nudge), 5 to 6 partial (general understanding but missing important specifics; repeated prompting needed), 3 to 4 weak (material omissions or errors; struggles to explain why it matters), 1 to 2 absent (no meaningful understanding demonstrated, or substantively incorrect).

Exhibit 4

Roleplay score distribution

15 sessions; 9 solid or better

7 to 8 out of 10, solid: 60%5 to 6 out of 10, partial: 20%3 to 4 out of 10, weak: 20%15sessions
9 to 10 out of 10, mastery0%
7 to 8 out of 10, solid60%
5 to 6 out of 10, partial20%
3 to 4 out of 10, weak20%
1 to 2 out of 10, absent0%
Source: roleplay score per session3, 15 sessions. Four independent pillars: objection handling (responds to resistance directly, reframes it and keeps the conversation moving), ICP and persona relevance (connects the conversation to the buyer's priorities and context), product understanding (describes Fin accurately, specifically and within the right scope), value narrative (connects Fin to a business outcome rather than listing features). The session score is the sum of the pillars scored divided by the number of pillars scored; if no pillar was discussed the session carries no roleplay score. The roleplay score is independent from the demo score and completion.

3. Resonance

Exhibit 5

What resonated with learners

60 sessions recorded what stayed with the learner

Insights as proof of value
30sessions
Resolution at scale
17sessions
The handoff that keeps context
17sessions
Procedures as guardrails
15sessions
Industry fit and customer stories
15sessions
Operator and the improvement loop
10sessions
›Examples by theme
Insights as proof of value: performance, gaps and improvement priorities30 sessions
  • Insights dashboards proving value to executives and identifying improvements
  • Insights showing automation rate, CSAT, and improvement opportunities
  • Insights showing success and ROI from autonomous agents resolving customer service inquiries
  • Insights surfacing performance gaps and high-volume issues from real conversation data
  • Insights turning Fin performance data into clear improvement priorities
  • and 26 more
Resolution at scale: fewer queues, less manual work, higher CSAT17 sessions
  • Fin AI Agent resolving a lost-card request end to end
  • Fin resolving complex onboarding and integration workflows end to end
  • Fin controlling support costs and volume spikes with high resolution rates
  • Automating common retail requests for fast self-service during volume spikes
  • Raising CSAT while offloading support cases for technology customers
  • and 13 more
The handoff that keeps context and helps human agents17 sessions
  • Seamless AI-to-human handoff with full context
  • Seamless handoffs preserving customer history and context so customers do not repeat themselves
  • Using Fin summaries and live transcripts to avoid customers repeating themselves during handoff
  • Intentional human handoff for sensitive banking cases
  • Context-rich handoff to the right human
  • and 12 more
Procedures and guardrails controlling sensitive workflows15 sessions
  • Procedures as a control layer for complex support workflows
  • Procedures providing strict control for sensitive processes
  • Procedures keep Fin inside approved company processes with human handoff for sensitive cases
  • Procedures correctly escalating suspected fraud to a human
  • Procedures controlling high-risk actions with policy compliance and escalation
  • and 11 more
Industry fit and customer stories15 sessions
  • High tech relevance for volume and complexity
  • Applicability to telco, utilities, and energy service scenarios
  • Fin handling energy and utilities customer requests directly
  • Fin reducing citizen friction and call center queues
  • AI handling routine healthcare questions at scale
  • and 13 more
Operator and the continuous improvement loop10 sessions
  • Operator helping teams fix gaps faster
  • Insights and Operator let operations teams identify gaps and draft improvements without engineers
  • Support teams can launch, test, and improve Fin without engineering support
  • Operator drafting structured improvements such as help articles or procedures for team review
  • Operator-driven continuous improvement loop
  • and 6 more
Helpdesk and Copilot for human agents8 sessions
  • Copilot support guidance for human agents
  • Copilot view for human agents
  • Copilot-supported handoffs saving time and improving customer experience
  • Copilot providing fast, consistent, source-backed responses for human agents
  • Helpdesk and Copilot providing guidance and content for financial services support
  • and 3 more
The overall Fin story and architecture6 sessions
  • Fin architecture/product overview
  • Fin’s architecture and overall capabilities
  • Relevance of Fin examples to Salesforce Service AE context
  • Five-capability Fin walkthrough
  • Fin’s overall architecture and connected product approach
  • and 1 more
Fin next to Agentforce: coexistence and where Fin wins6 sessions
  • Fin helping an existing Agentforce Voice customer across channels
  • Fin layered on top of Salesforce Agentforce as an enhancement rather than rip-and-replace
  • Agentforce and Fin coexisting, with Fin for high-volume repetitive CX tasks
  • Fin as a self-service front end for Agentforce automotive customers
  • Fin as a potential fit for tyre-change requests compared with Agentforce
  • and 2 more
The roleplay itself4 sessions
  • Role-play as the next interactive step
  • Business-value role-play for SMB CX leaders
  • Role-play as the most interesting next step
  • Role-play practice experience
Outcome-based pricing2 sessions
  • Outcome-based pricing for high call volumes
  • Outcome-based pricing tied to actual resolutions
Source: resonance recorded by the agent, 60 sessions. A session counts once per theme and can sit in several, so counts overlap. Themes are matched on what the learner literally named.

Exhibit 6

What learners asked to see

27 sessions asked the agent to show something specific; the capabilities and the product walkthrough came first

Capabilities and the product walkthrough
11sessions
Industry scenarios and the roleplay
5sessions
Integrations and channels
4sessions
Pricing, proof points and competitive positioning
4sessions
Architecture, models and how to turn it on
3sessions
›Examples by theme
Capabilities and the product walkthrough11 sessions
  • Fin AI Agent
  • the capabilities, the handoff, Procedures
  • insights into performance and business outcomes, voice agent fin, Operator
  • high level overview of Finn
  • walk through the platform and different use cases
  • and 15 more
Industry scenarios and the roleplay5 sessions
  • Role play
  • brokerage
  • airline scenario
  • e-commerce demo
  • retail value for Carrefour, French language capability
  • and 2 more
Integrations and channels4 sessions
  • integration with service platforms, email integration
  • Slack integration
  • email channel for communication
  • Fin's integration with Agentforce service
  • languages, Turkish
  • and 7 more
Pricing, proof points and competitive positioning4 sessions
  • deep dive of the pricing model
  • pricing tiers
  • financial banking use case, biggest customers of Fin worldwide, number of resolutions per month
  • Fin buying committee, dual impact after deploying Fin
  • difference between Fin and Casey
  • and 11 more
Architecture, models and how to turn it on3 sessions
  • custom models
  • control intelligence
  • LLM in the background
  • UI for how to turn it on
Source: requests recorded by the agent, 27 sessions. Each session sits in the theme that dominates its requests.

Exhibit 7

Competitors mentioned by learners

Agentforce in 12 of the 23 sessions that named one

Agentforce
12
Service Cloud
4
Zendesk
3
›All alternatives named
In-house AI builds
2
Freshdesk
2
HubSpot
2
Casey
2
Decagon
2
Sierra
2
Salesforce Voice
1
Genesys Cloud
1
ServiceNow
1
Source: alternatives named by learners, 23 sessions. A session naming two counts once for each. Agentforce and Agentforce Service are counted together; Salesforce and Service Cloud are counted together.

4. Questions

Exhibit 8

What learners asked the agent about Fin

63 of 257 sessions; how Fin connects comes before how to sell it

How Fin works and connects
19sessions
How do I actually sell this
14sessions
Fin next to Agentforce and the Salesforce stack
11sessions
Apex, models and architecture
8sessions
Pricing and billing
6sessions
Proof points and differentiation
5sessions
›Examples by theme
How Fin works: deployment, integrations, channels and handoff19 sessions
  • Which languages do you support?
  • How does this integrate with Salesforce?
  • How to integrate with email?
  • What is your integration with Slack?
  • Does all this work on mobile?
  • and 43 more
How do I actually sell this: pitch, SKU, scenarios for my customer14 sessions
  • Which is the correct Salesforce SKU to sell Fin?
  • How would this work for a customer like Arista Networks?
  • Do you have a customer example in manufacturing?
  • Can you provide an easily understandable pitch for my SMB customers?
  • What kind of healthcare companies have you had success with already?
  • and 29 more
Where Fin sits next to Agentforce, Service Cloud and Casey11 sessions
  • Could Fin help an existing Agentforce Voice customer?
  • Is Procedures like Agentforce instructions in topics and actions?
  • What is the difference between Agentforce and Fin?
  • How does Casey fit into this?
  • How do we position Fin against Agentforce?
  • and 44 more
How Fin is built: Apex, models and architecture8 sessions
  • Can you tell me more about the custom models?
  • Is it based on an open weights model?
  • What is Fin built on? And how do I enable it?
  • Are all of Fin's models first party?
  • Is there a context in the control layer where Data 360 comes into play?
  • and 11 more
Pricing and billing6 sessions
  • What is the pricing for Intercom plus Fin?
  • What happens to the billing charge if a resolved conversation is reopened later?
  • What are seats? Is it every Fin agent or every support agent Fin is tied to?
  • How is it priced based on resolution?
  • How does Fin's billing model resolve a prospect objection that AI is too expensive?
  • and 26 more
Proof points and differentiation5 sessions
  • How is this better than Genesys Cloud?
  • Is 76% a good number?
  • What is the primary customer problem that Fin solves?
  • What is Fin's key advantage over AI features from other help desks like HubSpot or Freshdesk?
  • What has been the real-world staffing impact for companies that have deployed Fin?
  • and 9 more
Source: questions extracted from each transcript by the agent, grouped by hand. 53 sessions that asked only about the session itself (skipping, speed, credit) are left out. Each session sits in the theme that dominates its questions.

5. Comprehension and misconceptions

Comprehension is what the agent recorded a learner working through; misconceptions are the beliefs it had to correct.

Exhibit 9

Comprehension

Procedures in 37 of 50 sessions

How Procedures define and control what Fin does
29sessions
How Procedures connect to Helpdesk and Copilot
8sessions
What Fin runs on: Apex, Intercom and where it lives
6sessions
Fin next to Agentforce, Service Cloud and Casey
4sessions
The role of Insights in improving Fin
3sessions
›Examples by theme
How Procedures define and control what Fin does29 sessions
  • How Procedures control Fin
  • How to explain Fin controls for sensitive financial services support
  • Procedures in plain language and why they matter
  • How Procedures define the steps and actions Fin follows
  • How Procedures connect to Fin AI Agent and human handoffs
  • and 36 more
How Procedures connect to Helpdesk and Copilot8 sessions
  • How Procedures relate to Helpdesk and Copilot
  • How Procedures connect to Helpdesk and Copilot
  • How Procedures connect to Fin, Helpdesk, and Copilot
  • How Procedures connect with Helpdesk and Copilot
  • Role of Insights in improving Fin
  • and 7 more
What Fin runs on: Apex, Intercom and where it lives6 sessions
  • What Fin is built on and how to enable it
  • Whether Fin is a separate app or where it runs
  • Relationship between Fin and Apex
  • Fin product architecture
  • Relationship between Fin and Intercom
  • and 2 more
Fin next to Agentforce, Service Cloud and Casey4 sessions
  • How Fin integrates and coexists with Salesforce Service Cloud and Agentforce
  • Difference between Agentforce and Fin
  • Difference between Fin and Casey
  • How Fin complements Agentforce Service
  • Relationship between Helpdesk/Copilot and Agentforce Service
  • and 5 more
The role of Insights in improving Fin3 sessions
  • Role of Insights in the Fin story
  • Insights’ role in the Fin story
  • Insights' role in measuring Fin performance and identifying improvement opportunities
Source: comprehension topics recorded by the agent, 50 sessions. Each session sits in the theme that dominates its entry.

Exhibit 10

Misconceptions

31 sessions carried at least one, most about how the capabilities relate to each other

How Procedures, Helpdesk and Copilot relate
9sessions
What Insights is for
5sessions
What Operator does on its own
5sessions
What Fin is built on
3sessions
How Fin fits with Salesforce, Intercom and billing
3sessions
The proof point numbers
3sessions
Knowledge and the five capabilities
3sessions
›Examples by theme
How Procedures, Helpdesk and Copilot relate9 sessions
  • Believes Procedures are primarily context that Helpdesk and Copilot draw from
  • Believes Helpdesk and Copilot identify conversation reasoning and call Procedures
  • Believes Procedures are only guidance for human agents after handoff
  • Believes Procedures can be initiated manually from the Helpdesk inbox
  • Believes Copilot manages or tunes Fin directly
  • and 6 more
What Insights is for5 sessions
  • Believes Insights is primarily for showing or auditing the specific steps Fin followed in a process
  • Believes Insights shows the amount of time and money Fin is saving the company
  • Believes Insights controls the process Fin follows
  • Believes Insights surfaces operational information during a live end-user workflow rather than measuring Fin performance
  • Believes Insights is for presenting products to customers
What Operator does on its own5 sessions
  • Believes Operator is specifically for human engineers to manage, monitor, and configure Fin
  • Believes Operator can deploy complex Procedures automatically
  • Believes Operator automatically drafts missing customer data connectors
  • Believes Fin automatically learns and improves itself continuously
  • Believes Fin continuously improves itself automatically via its retrieval model
  • and 2 more
What Fin is built on3 sessions
  • Believes Salesforce Apex is equivalent to Fin’s Apex model layer for building an agent
  • Believes Fin is completely built using Apex
  • Believes Fin is essentially ChatGPT connected to documentation and support tools
  • and 2 more
How Fin fits with Salesforce, Intercom and billing3 sessions
  • Believes Fin plus Intercom is required when using Fin with Salesforce Service Cloud
  • Believes customers only pay when Fin resolves a process end to end and never when it transfers to a human
  • Believes Fin and Casey do the same thing
  • Believes Fin requires switching the customer’s helpdesk system
  • Believes Apex is a free public model
  • and 2 more
The proof point numbers3 sessions
  • Believes Fin’s resolution proof point is 50 to 80% rather than the established 76% average across customers
  • Believes many Fin customers are above 96% resolution rather than above 90%
  • Believes Fin has a 78% resolution rate
  • Believes the 76% resolution benchmark is a customer-specific support-case offload rate rather than an average across queries Fin sees
Knowledge and the five capabilities3 sessions
  • Believes knowledge articles require a person to activate or use them, rather than being reference content Fin can use directly
  • Believes Knowledge Integration and System Actions are two of the five Fin capabilities, instead of Helpdesk and Copilot and Operator
  • Believes there is a Fin capability called Observer that measures interactions and turns gaps into improvements
Source: misconceptions recorded by the agent, 31 sessions. Each session sits in the theme of its dominant misconception; the dropdown lists them all.

6. Returning learners

Exhibit 11

Returning learners

Two of the 11 learners with more than one valid session; none yet has a section graded twice

LearnerScoreCoverageAttemptsImprovementCoach nextRoleplay
Learner F2 valid sessions3.01 of 52n/aCore enablement experience6.5
Learner H2 valid sessions6.01 of 52n/aHelpdesk and Copilot6.8
Learners are identified by email across their valid sessions; two of the 11 are shown as examples. Score: the demo score of the core enablement experience at the latest attempt. Coverage: sections ever graded. Attempts: valid sessions. Improvement: the latest attempt against the previous one, only when the core enablement experience was graded twice. Coach next: lowest latest grade below 6, or the first section not yet covered. Roleplay: latest roleplay grade. Learners are lettered, never named.

Appendix: how this report was produced

DefinitionsEvery defined term is footnoted where it first appears. Demo score, completion and roleplay score are the three headline measures; enabled is a demo score of 5 or higher. Coverage, demo, capability and roleplay grades are the reviewed grading carried on each session record. Questions, resonance, requests, alternatives, gaps, comprehension and misconceptions come from the agent and cover all valid sessions.
Grading1 to 10 against the shared rubric:
  • 9 to 10 mastery: correct, specific and unprompted; ready to use with a customer
  • 7 to 8 solid: covers the key points with minor gaps or one nudge
  • 5 to 6 partial: general understanding but missing important specifics; repeated prompting needed
  • 3 to 4 weak: material omissions or errors; struggles to explain why it matters
  • 1 to 2 absent: no meaningful understanding demonstrated, or substantively incorrect
A section never reached carries no grade. A returning session starts where the learner left off and updates the standing grade.
Qualitative signalsInsights are automatically extracted by the AI agent and grouped into recurring themes based on similarity across sessions. The analysis covers the key signals surfaced during the experience, including questions asked by AEs, messages that resonated, misconceptions, areas of strong or weak comprehension, competitor mentions, and capabilities learners requested to explore in greater depth.