Risotto raises $10M to reshape AI-driven customer support
Support leaders know the drill: platforms like Zendesk, ServiceNow, and Freshworks run the show, but workflows still feel clunky. Risotto just closed a $10 million seed round led by Bonfire Ventures, with 645 Ventures, Y Combinator, Ritual Capital, and Surgepoint Capital joining in. The goal is simple: make AI actually useful inside the support stack.
Instead of replacing your ticketing system, Risotto plugs into it. It sits between systems like Jira and the internal tools your team uses to resolve tickets, acting as the orchestration layer that keeps AI on task.
What Risotto actually does
Risotto uses a third-party base model, but the real value is the infrastructure between the model and the client. That layer manages uncertainty, guards against bad outputs, and routes actions to the right internal tools.
Their "flavor" is practical: prompt libraries, eval suites, and thousands of real examples from real support interactions. That gives the system a tighter feedback loop and a higher chance of doing what your runbooks expect.
Early results: 60% automation at Gusto
In collaboration with Gusto, Risotto reports about 60% of support tickets being automated. That's deflection and resolution combined, not magic. Outcomes will vary by ticket mix, knowledge quality, and system access, but the direction is clear: more automation without ripping out your current stack.
Why this matters for support leaders
- Lower handle time: AI handles repetitive tickets and drafts quality responses for agents to approve.
- Fewer handoffs: The orchestration layer talks to internal tools directly, reducing swivel-chair work.
- Consistency at scale: Prompt libraries and evals enforce policy and tone across teams and shifts.
- Governance built-in: Guardrails keep model uncertainty in check, which is the real blocker for most teams.
- No platform overhaul: Works alongside existing systems like Jira and Zendesk, so you keep your data and workflows.
How to prep your org for AI-assisted support
- Map your top 20 intents by volume and impact. Start with clear, well-documented flows.
- Centralize knowledge. Fix duplicate macros, outdated docs, and scattered SOPs.
- Define trust thresholds. What can be auto-resolved vs. agent-approved vs. always-human?
- Instrument everything. Track FCR, AHT, CSAT, reopens, and time-to-resolution with and without AI.
- Run a 4-6 week pilot in one queue. Measure before expanding.
- Lock down data. Add redaction, access controls, and audit trails before connecting internal systems.
Risks and guardrails to consider
- Model drift and bad outputs: Use eval suites and regression checks before every change.
- Over-automation: Keep humans in the loop for policy, billing, security, and edge cases.
- Change management: Train agents on review workflows, escalation rules, and when to override AI.
- Ticket hygiene: Poor categorization and messy fields will tank performance; clean data first.
The bigger shift
If this approach scales, AI moves from "add-on" to the core of service operations. That means fewer manual touches, faster resolution, and a more reliable layer across fragmented internal tools.
There's also immediate cleanup to do. Some teams still need multiple people just to wrangle Jira. Streamlining current workflows and data structures lays the groundwork for AI to actually help.
What this means for your roadmap
- Short term: Automate FAQs, password/account tasks, and policy-driven requests with human review.
- Mid term: Connect internal tools for refunds, adjustments, verifications, and account actions.
- Long term: Shift from ticketing-first to intent-first routing and actioning with clear oversight.
Further learning
- Build team skills for AI-assisted support with practical courses: Complete AI Training - Courses by Job
The takeaway: AI won't fix broken processes, but paired with clean workflows and strong guardrails, it can handle a large chunk of repetitive support work. Tools like Risotto make that possible without a full platform swap-and that's where real value shows up for support teams.
Your membership also unlocks:
