Home > Columns > CRM Columns

CrmXchange Weekly Pulse - Sept 21–28, 2026

CrmXchange

Presented By: CrmXchange


CrmXchange Weekly Pulse

The big theme: contact center AI is being judged on proof (resolution and governance), not features.

  • Cresta: only 9% of conversations are fully autonomous and 76% are human plus AI.
  • UK complaints about voicebots blocking access to a human are becoming a regulatory issue.
  • Salesforce's Dreamforce launches (Casey, human+AI workforce management) are still being unpacked.

CX & Contact Center Weekly Briefing: Sept 21–28, 2026

Top themes this week

1. Buyers now want proof that contact center AI works, not feature lists
Analysts and buyers are judging AI by whether it resolves issues, integrates with existing systems and shows ROI. ISG's new buyers guide scores vendors on governance and integration, not demos; it names NiCE, Verint and Genesys as leaders. Zeus Kerravala put it as "resolution and resolution quality is the new unit of value." Several pieces this week say containment and deflection aren't enough to measure success.
Sources: CX Today · Silicon ANGLE

2. Most service is still humans and AI working together
Cresta's workforce report finds only 9% of conversations are fully autonomous, while 76% involve humans and AI together. 93% of leaders say conversations are getting more complex. Vendors are building for this: SoundHound's "Human Assisted Resolution" lets an AI agent get a quick approval from a person without transferring the call, and Salesforce added "agent approvals" to its Service Cloud console.
Sources: CX Today · CXM

3. Getting to a human is becoming a regulatory issue
CXM reports a 34% quarterly jump in complaints to UK banks about voicebots blocking access to a person. It attributes the figure to FCA Consumer Duty disclosures but doesn't show how it was calculated. The argument is that a high containment rate can hide customers who simply gave up. No Jitter's companion piece proposes "tiered autonomy": a human signs off whenever an outcome affects a customer's money, rights, access, privacy or safety.
Sources: CXM · No Jitter

4. AI efficiency is becoming table stakes, so outcomes are the new edge
As transcription, automated QA and agent assist become standard in every platform, No Jitter argues that speed stops setting anyone apart. What will: getting the answer right, catching problems early, and letting agents make real decisions. ScorebuddyCX data shows why that's hard. 76% of managers trust AI quality scores but only 57% of agents do, and just 14% of teams review those scores consistently.
Sources: No Jitter · CX Today

5. Fraud and trust pressure on phone channels
UK regulator Ofcom opened investigations into Vonage and Voxbone over scam calls made from numbers they provide. The coverage notes that only 26% of UK consumers answer calls from unknown mobile numbers. A separate report projects deepfake ID-document fraud to grow roughly 40-fold, with those fake identities then used to impersonate customers in contact centers.
Source: CXM roundup, Sept 25

Notable news & launches

  • Dreamforce follow-ups: Salesforce launched Casey, a ready-made Agentforce service agent, plus a Command Center that shows human and AI agents in one view and workforce planning that covers both. (Salesforce Ben, Sept 21)
  • Telstra is rolling Agentforce into live workflows (refunds, eligibility checks, restoring service) across 1,000+ reps. (CX Today)
  • Zoom bought Onyx, an open-source AI search tool that companies can host on their own servers, so sensitive data stays off outside clouds. (CXM)
  • Mutual of Omaha Mortgage uses voice AI to handle the first 30 seconds of calls before handing borrowers to a loan officer. (CX Dive, Sept 23)
  • Coming up: Customer Service Week is Oct 5–9. (Achievers)

Worth a read

One takeaway

The question for leaders has moved from "how much can AI handle?" to "can we prove it resolved the issue, and that customers could always reach a human when the stakes were high?"