Customer support answers what the customer opened. Customer success owns whether the account reaches the outcome you sold, before the renewal is already lost. That is the split. It is not “nice vs. tickets,” and it is not a reason to buy two chat products.
Guzli sells a support chatbot and wrote this page. It is not an independent customer-success or customer-experience study. An AI chatbot can take repetitive support work and surface success signals (repeat tickets, booking a quarterly business review, expansion intent). It does not replace a customer success manager on a named account.
Method: desk research on operator writeups (opened 19 September 2026), not shadowed CSM or support shifts.
AI context: AI in customer service. Tool criteria: AI chatbot buyer’s guide.
Verdict
- Support: reactive, transactional, measured on response, resolution, customer satisfaction on the closed ticket (CSAT), escalation.
- Success: proactive, relational, measured on time-to-value, adoption, gross and net revenue retention (GRR/NRR), churn, expansion.
- The failure mode is information routing: the customer success manager (CSM) becomes a shadow helpdesk, or support never tells success that the same account filed five billing tickets.
Who this is for
Leaders writing a RACI (who is responsible, accountable, consulted, informed). Small teams that wear both hats and need a trigger list anyway. Anyone about to “let the bot do success.”
Who this is not for
You only wanted a two-line definition. You cannot name who owns a refund versus a renewal conversation.
Customer support
The customer starts it. The work has a beginning and an end: ticket, chat, call. The job is restore the promised service — bug, bill, how-to, outage.
Own: first response time, time to resolve, first-contact resolution (the ticket closed without a transfer), CSAT on the closed ticket, escalation rate, cost per resolution, contact rate per account (shared with success).
Do not own: whether the account adopted the feature they bought. A 4-minute first reply on a product the customer never launched is still a support win and a success miss.
Customer success
The vendor starts more of it: kickoff, quarterly business review (QBR), risk review, enablement. The CSM owns the outcome in the contract — go-live, seats used, workflow live — and the commercial health of the book.
Own: time-to-value (how long until the sold outcome is live), product adoption in a defined cohort, health score you can explain, GRR (revenue kept after churn and downgrades, excluding expansion) and NRR (the same book including expansion), renewal rate, expansion, logo and revenue churn.
Do not own: every password reset. If customers message the CSM directly for tier-1 break-fix, you do not have a success motion. You have an unscalable support queue with a friendlier title.
HubSpot’s 2026 success-metrics article (a vendor blog, opened 19 September 2026) mixes leading indicators (engagement, effort) and lagging ones (renewal, lifetime value). Treat QBRs completed and calls logged as activity. They can be green while NRR falls. Gainsight’s GRR vs NRR explainer (a CS-platform blog, same date) is the ordinary-language split used above: GRR excludes expansion; NRR includes it. Operator blogs such as The Marketing Juice and Fini use the same job split — support owns the ticket; success owns the account. None of those are a shadowed-team study.
Side-by-side
| Support | Success | |
|---|---|---|
| Timing | After the customer reaches out | Before the customer is in trouble |
| Unit of work | Conversation / ticket | Account / cohort |
| Goal | Restore service | Reach outcome; keep and grow revenue |
| Typical channel | Chat, email, phone, social, portal | Kickoff, QBR, in-product, executive review |
| Scales by | Automation, deflection, staffing | Segmentation and CSM:account ratio |
Customer experience (CX) — net promoter score (NPS), effort, brand — sits across both. Do not give support an NPS target and call it success.
Overlap is normal; ownership must still be singular
A billing outage is support. The same outage the week before renewal is support plus a success risk flag. Onboarding “how do I invite a user?” can be a help article (support) and a time-to-value milestone (success). Write the owner on the ticket type, not in Slack folklore.
Small team, one person: still split the calendar. Morning is the queue. Friday is health review. If the queue always wins, you do not have success. You have support with a strategy deck.
Handoffs that do not drop context
Sales → success (new logo). Within a day of close: goals, stakeholders, in-scope vs out, risks, success criteria, go-live date. Incomplete packages show up as “the CSM is rereading the CRM.”
Support → success. Trigger on pattern, not vibe:
- Same intent, same account, N tickets in M days
- Severity / security / data-loss
- Usage down after a burst of how-to tickets
- Renewal inside 90 days and a rising contact rate
The ticket is the artifact. Success acts on the pattern. Support should not wait for the CSM to “just look at this one.”
Success → support. CSM forwards with a warm sentence and the account goal. A cold “file a ticket” feels like a brush-off; answering it themselves trains the customer to skip the desk.
Kickoff line to the customer: chat/email/phone for break-fix; CSM for outcomes and renewals. Repeat it when someone tests the boundary.
Where AI belongs (and does not)
Support is more automatable: bounded, documented, repetitive. Success scales with judgment and a book of accounts.
A useful agent:
- Resolves tier-1 from the same help center both teams edit
- Opens a ticket in Salesforce, Zendesk, or Guzli’s help desk with a summary
- Collects a meeting when the user is actually in an onboarding or expansion moment
- Flags repeat-ticket accounts for the CSM (that is a report, not a personality)
A harmful agent: books every visitor into the CSM’s Calendly, or answers renewal-commercial questions from a stale FAQ.
Guzli can sit on website live chat (and inbound voice on Standard+) for the support half, train on articles and solved tickets, and hand off with context. See current plans and pricing. Hobby $49/mo chat-only. Voice starts Standard $149 + 16¢/min. That is support infrastructure. It is not a success org.
Worked path (desk research)
Illustrative desk-research path, not a measured customer book: a 40-account SaaS roster. Billing “why was I charged?” is 30% of tickets. Support owns the Stripe explanation and the refund policy. Success gets a weekly list of accounts with three or more billing tickets in 14 days — those churn for “unpredictable invoice,” not for slow first reply. Automate the explanation; do not automate the save conversation.
FAQs
Can both teams share one chatbot?
Yes, if routing and the knowledge base are shared and the bot does not pretend to be the customer success manager. A useful agent resolves tier-1 from the same help center both teams edit, opens a ticket with a summary, and flags repeat-ticket accounts. It does not book every visitor into the CSM’s calendar or answer renewal-commercial questions from a stale FAQ.
What do we automate first?
Repetitive support intents with a written policy. Then scheduling for onboarding. Then, maybe, a health-alert digest for accounts that trip the trigger list. Not “AI quarterly business reviews.”
How do we prove the split is working?
Support: volume, first-contact resolution, CSAT on the closed ticket, escalation. Success: time-to-value, adoption, net revenue retention. The 40-account billing path above is the combined story — automate the Stripe explanation; do not automate the save conversation. Worked math for the support half: AI chatbot ROI.