Program report · filed November 19, 2026
AI in SalesMeasured report
SupportNinja CEO: AI Customer Service Fails When Teams Pick Tools First
SupportNinja CEO Craig Crisler argues CX teams deploy AI backwards: containment rates can hide unresolved issues, and data analysis must come before automation picks a channel.
By Marcus Bennett3 min read602 words
Program notes
- SupportNinja CEO Craig Crisler says AI customer service projects should start with a defined problem or goal, then select and test the tool against it
- A higher containment rate doesn't guarantee resolution; failed bot interactions migrate to other channels and reappear on different dashboards
- Existing data sources — CRM records, transcripts, surveys, escalations, repeat contacts, sentiment trends — can expose friction points before automation
- SupportNinja cited a benefits administrator that cut status-update contacts with proactive alerts across four stages of the claims journey

A higher containment rate doesn't mean the customer got what they needed. That single distinction sits at the center of a new argument from SupportNinja's CEO and Co-Founder Craig Crisler, who says too many CX teams are deploying AI for the sake of deploying AI — then watching the metrics without knowing whether anything actually got resolved.
The failure mode is familiar to anyone running a support operation under pressure. CX leaders face demands to automate faster, cut ticket volume, and prove a return on their AI spend. Under that strain, many identify a busy support channel, drop in a chatbot or an agentic tool (AI that can take actions rather than just answer questions), and hope the numbers move.
What's wrong with containment as a target?
The problem, per the argument, is what the dashboards miss. A bot can answer quickly, fail to resolve the issue, and push the customer into another channel. The interaction then disappears from one dashboard and reappears somewhere else — which reads as a win in the first system while the customer repeats the effort elsewhere.
Crisler's framing flips the sequencing most teams use. "You start with the problem or the strategic goal you're trying to achieve, then you find a solution, and then test against that to understand whether or not it works," he said. In practice, that means CX leaders should stop asking where a chatbot can be deployed and start asking why customers are getting in touch at all.
Where to find the friction first
The diagnostic data usually already exists inside the business. The sources named include:
- CRM records and email threads
- interaction transcripts
- customer surveys
- escalations and repeat contacts
- sentiment trends
Crisler's point is that AI belongs at this stage too — before automation, not just after it. "The beauty of AI is that you can analyze data at scale. Using that tooling to analyze first and understand, 'okay, well, this is a friction point in our CX journey. Let's look at that and see if there's something we can fix around that,'" he said.
Why efficient handling isn't the same as fixing the problem
The piece offers one concrete pattern: customers repeatedly contacting a business to ask for a claim status update. Handling those calls more efficiently will improve operational metrics. It does nothing about the effort that caused the calls in the first place — the customer still had to reach out.
SupportNinja has pointed to a benefits administrator that took the other path, deploying proactive alerts across four stages of the claims journey. Instead of waiting for status calls, the administrator told customers where their claim stood at each stage. The approach targets the root contact driver rather than the contact itself — the difference between absorbing demand and removing it.
What this changes for CX teams
For teams mid-rollout, the implication is a reordering of the workflow: mine existing CRM, transcript, and survey data for friction points, define the strategic goal, then select and test automation against that goal. Containment rate becomes a secondary metric, not the scoreboard. For smaller teams without dedicated analytics staff, the same AI tooling can do the upfront analysis before it's ever pointed at customers — which lowers the cost of doing the diagnostic work first.
The broader signal is that AI customer service is moving out of its demo-driven phase. As vendors keep shipping agentic tools, the separating line between programs that work and programs that churn will likely be measured in repeat contacts and resolution data — not in how slick the rollout looked.
via cxtoday.com (Source)
More from Marcus Bennett
Show full bio
Staff writer covering people, launches and day-to-day practice in Sales technique and methodology, sales enablement at Quota Brief.
120 articles
Also rated
- Microsoft bets contact center harness will outlast the AI model
- ServiceNow Data: One Customer Problem Spans Three to Five Apps
- Microsoft Puts Dynamics Inside Copilot as Contact Center AI Goes GA
- Article 50 of the EU AI Act Lands on Sales and Marketing Desks
- Gartner's 2026 CRM CEC Quadrant: Five Leaders, One Battle — Governed AI Autonomy