Three insights customer support teams can use from Refresh North America 2025
AI is loud right now. But if it doesn't make getting help faster and easier for customers and employees, it misses the point.
Freshworks reported a 15% year-on-year revenue bump to just over $215M in Q3, fueled by demand for AI-driven service. Enterprise teams aren't chasing hype - they want faster response, higher CSAT, and work that feels meaningful. As Freshworks CEO Dennis Woodside put it, agents shouldn't spend their day resetting passwords and answering returns. That's work AI can take off the plate with better outcomes for the customer.
Insight #1: AI only works if it's built into the workflow
Freshworks' approach is clear: integrate AI into the tools support teams already use. Freshservice (employee experience) and Freshdesk (customer experience) sit under one AI layer serving three personas - end users (Freddy AI for self-service), agents (Freddy Copilot), and leaders (Freddy Insights). It's one system, not a bolt-on.
Customers moving off heavy platforms want the same thing you do: speed to value and less babysitting. Stack Overflow's team, for example, pushed onboarding into Freshservice and automated the handoffs across IT and People Ops. On-time laptop delivery shot up, and the team shifted from reactive firefighting to proactive ops.
What to do next:
- Map your top 10 repeat intents (password reset, returns, access requests). Automate those first.
- Expose a single self-service entry point. Don't make people hunt through channels.
- Give agents an AI copilot inside the ticket, not in a separate tab.
- Instrument the basics: deflection rate, time to first response, time to resolution, CSAT.
- Review one workflow per week. Remove clicks, merge steps, trim fields.
Insight #2: Complex operations need simple platforms
TaylorMade Golf shifted from an on-prem, services-heavy setup to Freshservice. That move helped them handle remote work, co-build workflows with app owners, and roll out an internal AI assistant called "Caddy." Employees ask Caddy questions, get knowledge or service requests routed automatically, and tier-1 noise drops.
As Freshworks product leadership noted, most orgs run enterprise-scale use cases with mid-market headcount. The platform has to be complete but straightforward, so your team - not an army of consultants - can stand it up and keep it moving.
What to do next:
- Publish a clear service catalog and tie every request type to an owner and an SLA.
- Name your assistant and make it the front door. If it feels familiar, people will use it.
- Turn your internal knowledge into AI-ready articles: short, stepwise, single outcome per page.
- Set a deflection target for tier-1. Track it weekly and iterate prompts/knowledge.
- Keep humans in the loop for judgment calls. AI routes and drafts; agents decide.
Insight #3: People-first AI turns support into a performance engine
McLaren Racing treats business tech like race-day kit - it has to be fast, reliable, and constantly improving. With six garage setups rotating worldwide and 150 people on the road, they run incidents and reliability through one Freshworks workflow across track and factory. AI adds another "opinion" with probabilities, while people make the call.
The goal: services that feel almost frictionless, without overbuilding. Uncomplicate the path from issue to fix. Keep learning loops tight so problems don't reach the "car" - whatever your version of that is.
What to do next:
- Use a single intake channel for all incidents and requests. No side doors.
- Standardize incident categories and severities. Consistency speeds triage.
- Automate containment steps for common failures before escalation.
- Treat AI suggestions as a second opinion. Require human confirmation on high-impact actions.
- Run weekly reliability reviews. Feed fixes back into automations and knowledge.
Your 30-day plan to put this into practice
- Week 1: Identify the top 5 repetitive tickets. Write simple SOPs and convert them to automations.
- Week 2: Stand up a single self-service portal. Point email, chat, and Slack to it.
- Week 3: Pilot an agent copilot in one queue. Measure draft quality and handle time.
- Week 4: Launch a named assistant to deflect tier-1. Ship 20 high-traffic knowledge articles.
- Across all weeks: Track deflection, FRT, TTR, CSAT. Review the numbers every Friday and adjust.
Helpful resources
- Freddy AI overview - how end-user, agent, and leader experiences connect.
- AI courses by job - practical training paths for support and service teams.
Disclosure: theCUBE is a paid media partner for Refresh North America 2025. Sponsors do not have editorial control over theCUBE or its coverage.
Your membership also unlocks:
