Step Aside Chatbots, AI Customers Just Showed Up
This week on Funnel Frontier: An AI just called a bank, passed checks, and made a deal.
This week:
- AI customers have just entered customer service
- Salesforce’s Cimulate acquisition could kill keyword shopping
Stat of the Week
34% of businesses using generative AI report exceptional customer service compared to those that don’t. (CRM.org)
AI customers have just entered customer service
A major U.S. bank recently negotiated a debt settlement over the phone. Authentication passed. Terms agreed. Case closed.
Only after the call did someone ask: was that even a human?
It wasn’t.
And here’s the real problem: there was no policy for that scenario. The agent followed the script. The system logged the interaction. Operations had no idea how to classify it.
Contact centers spent the last three years deploying AI to serve customers. Now AI is showing up as the customer.
The architecture problem nobody planned for 📐
After the debt deal closed, the bank had a bigger issue than collections.
The AI had passed knowledge-based authentication, negotiated terms, and completed a transaction and there was zero internal policy for handling a non-human caller acting legally on someone’s behalf.
Wayne Kay from TTEC Digital says this post-call panic is spreading. At a recent CX summit, most leaders admitted they would simply hang up and label it fraud.
The catch? Services like Kickoff, Google’s AI shopping agents, and OpenAI’s Operator are legitimate delegated agents. They represent real customers with permission.
The real issue now isn’t AI detection, it’s control. Most contact centers can confirm identity, but they can’t verify scope. What exactly is this AI allowed to do? Negotiate? Transact? Change data? On whose authority?
That requires machine-level authentication, defined permission limits, and clear internal policy. Without it, you either block valid customers or approve actions you don’t fully understand. Neither is a strategy.
And while that’s happening… we’re about to get our first AI coworkers? 💼
While contact centers are figuring out how to handle AI customers, OpenAI is busy building AI coworkers.
The new Frontier platform is designed to help enterprises deploy agents that don’t just chat, they work. Think CRM updates, ticket resolution, workflow execution, cross-system reasoning. Not a sidebar assistant. A digital employee with defined permissions.
OpenAI’s pitch is simple: the models are ready. The bottleneck is enterprise plumbing.
Frontier connects CRM, ticketing, data warehouses, and internal tools into what it calls a shared “semantic layer,” so agents understand context before they act. Each agent gets its own identity, permissions, memory, and evaluation loop. In other words: onboarding, guardrails, supervision.
Early use cases aren’t cute demos. One manufacturer cut optimization work from six weeks to a day. A global investment firm freed up 90% more selling time by letting agents handle process-heavy steps.
💡 The takeaway
Here’s the uncomfortable truth: most CX stacks were built to verify people, not permissions.
Now you have two new actors:
- AI customers are calling in with delegated authority.
- AI coworkers executing workflows inside your systems.
Both force the same shift: from “Who are you?” to “What are you allowed to do, and who approved it?”
If you’re leading CX, here’s where to start:
- Fix authentication first. Move to token-based, machine-to-machine verification with clear delegation limits.
- Audit your permissions model. AI should get minimum necessary access. No overexposed CRM environments.
- Split the lanes. Humans need empathy. Machines need structured exchange. Build a transactional API path for AI agents and keep your human support path human.
- Write the policy now. If an AI customer calls tomorrow, who owns it? Ops? Fraud? Legal?
The Week @ CRM.org
Mobile CRM. Because workflow on the go is a deal-maker.
Best AI CRM Software for 2026. CRMs used to just store names. Now they predict churn, write your emails, and remind you who ghosted you.
🌸 Weekly Bloom 🌸
Balancing Customer Retention & Churn. Customer churn is like having aliens abduct your paying customers. It’s time to power up the retention forcefield!
Salesforce’s Cimulate acquisition could kill keyword shopping
Salesforce is acquiring Cimulate, and it’s a shot across the bow of every retailer still worshipping the sacred keyword.
Because shoppers don’t think in keywords anymore. They think in vibes.
Nobody wakes up and types “blue cotton midi dress under $150.” They think, “I have a beach wedding in May and I refuse to look like a tablecloth.”
That’s intent. And intent is messy, contextual, and very bad at fitting inside a search box built in 2012.
Search is dead. Long live intent. 🛍️
Salesforce didn’t buy Cimulate to make search slightly better. It bought it because keyword search is a relic.
Cimulate built an intent-aware context engine for retail. Instead of reacting to what a shopper types, it analyzes real and simulated shopper journey data to predict what they’re actually trying to do — compare, browse, gift, upgrade, splurge, save.
Once folded into Agentforce Commerce, that engine upgrades discovery from “here are 42 blue sweaters” to something closer to a digital sales associate. It can factor in behavior, context, and similar shopper patterns to surface smarter recommendations in real time.
For retailers, that means fewer dead-end searches and more conversions.
For Salesforce, it means Agentforce Commerce gets a brain at the top of the funnel, not just a checkout button at the bottom.
What this really means for you? 💡
If you’re a Salesforce Commerce customer, this is your cue: search is no longer a feature. If your catalog, customer data, and browsing behavior live in silos, no AI discovery layer will magically fix that. Intent engines only work when the data underneath them is clean and connected.
If you’re not on Salesforce? The competitive bar just moved anyway.
Your site needs to understand why someone is browsing, not just what they typed. Occasion. Budget. Style. History. Signals from marketing. Signals from service. All connected.
So here’s the practical play:
- Audit your product data. Is it structured well enough for AI to reason over?
- Map behavioral signals. Are you capturing intent or just clicks?
- Break the silos between commerce, marketing, and CRM. Context dies in isolation.
Because the next generation of e-commerce won’t be a better search bar. It’ll be a guided experience.
And if your stack can’t support that, someone else’s will.
Galactic Gourmet
CRM blips from around the web
Genesys Unveils Industry-First Agentic Virtual Agent for Enterprise CX. Powered by large action models (LAMs), the new Genesys Cloud Agentic Virtual Agent can autonomously resolve complex customer requests across front- and back-office systems.
Cisco Launches Webex Contact Center for ServiceNow. Webex Contact Center now integrates natively with ServiceNow, embedding voice and digital channels. The move streamlines agent workflows and accelerates issue resolution with built-in AI and unified data.
Mailchimp Rolls Out Advanced Ecommerce Marketing Tools. Intuit Mailchimp introduced new data-driven features to help ecommerce brands drive higher ROI across email and SMS. The updates unify customer data, expand global SMS coverage, and add AI-powered automation and analytics.
If you'd like weekly CRM news like this delivered to your inbox, subscribe to Funnel Frontier!