There’s a version of this conversation that agency owners had about websites in 2005. “Do we really need one? Our clients know where we are.” Then about email. Then about online quoting. Every time, the early movers captured market share while the late movers spent more to catch up.
AI in customer service is at that inflection point right now, and the data is no longer ambiguous.
The numbers
Google Cloud published a study in September 2025, conducted with National Research Group. It surveyed 3,466 senior executives at companies with 100+ employees and $10M+ revenue across 24 countries, with the fieldwork run between April and June. The findings:
- 63% of executives report their organization is using AI for customer service and field service. Not piloting. Using.
- 52% have deployed AI agents in production.
- 88% of early adopters are seeing positive ROI.
- 77% say their AI spending has increased year over year.
These aren’t projections or forecasts. This is what’s already happening at the companies your agency’s clients work with, buy from, and compare you to.
Financial services is leading
The study also broke the agent numbers down by industry: of the organizations already running AI agents, the share that put those agents on customer service and experience. Financial services, which includes insurance, came in at 57%, the highest of any industry in the study.
| Industry | AI adoption for CX |
|---|---|
| Financial services | 57% |
| Manufacturing | 56% |
| Public sector | 51% |
| Retail and CPG | 47% |
| Telecommunications | 45% |
| Media and entertainment | 44% |
| Healthcare and life sciences | 42% |
Your industry isn’t behind on AI. It’s actually ahead of most. The question is whether your agency has already put AI on the phones and the inbox, or is still deciding.
What “adoption” actually looks like at an agency
When the study says “63% use AI for customer service,” it doesn’t mean every one of those companies replaced their entire call center with a robot. Most are using AI for specific, well-defined tasks:
- Answering routine inquiries: “What’s my policy number?” “When does my coverage renew?” “Can you send a certificate?”
- Routing calls: AI figures out what the caller needs and gets them to the right person, with context.
- Documenting interactions: transcripts, summaries, AMS entries generated automatically.
- Sorting and prioritizing email: separating client requests from newsletters and carrier updates.
- Drafting responses: AI writes a reply, a human reviews and sends.
None of this is science fiction. It’s plumbing. Mechanical work that someone at your agency is doing manually right now, call by call, email by email.
The cost of waiting
The Google study found that early adopters, the companies that committed to AI agents early and embedded them across operations, are pulling ahead in ways that compound:
- 82% of early adopters have deployed 10+ AI agents, compared to 39% of all organizations.
- 78% have been running AI in production for over a year.
- Early adopters are allocating 50%+ of their future AI budget to agents.
- 39% of their total annual IT spend goes to AI, compared to a 26% industry average.
The gap between early and late adopters is widening, not narrowing. The early movers have a year of operational learning, a year of cost savings, and a year of data that late movers don’t have.
In insurance specifically, this matters because the agencies that adopt AI now are building something the ones that wait can’t easily replicate: a workflow that’s faster, cheaper, and captures intelligence from every customer interaction. Every call that gets transcribed and summarized is data. Every routing decision is a pattern. Every email that gets sorted is time saved. Over a year, that compounds.
What agencies are spending on AI, and where it comes from
The spending data is instructive:
- 77% of organizations report their AI spending has increased.
- 58% are funding AI with net new budget, not cutting other areas.
- 48% are reallocating budget from non-AI areas to fund AI investments.
For an agency, the “non-AI budget” being reallocated is usually the answering service line item, the extra CSR hire that was planned, or the overtime that built up from understaffing. The math often works out to: “We were going to spend this on more people doing the same work. Instead, we spend less on AI doing the mechanical parts, and the people we have do more selling.”
The security concern is real, and it’s your advantage
The study found that 37% of executives say data privacy and security is the number-one consideration when choosing an AI provider. Integration with existing systems was second at 28%. Cost was third at 27%.
Insurance agencies should weight security even higher. You handle PII, financial data, and on benefits lines, health information. A breach isn’t just embarrassing. In the states that have adopted the NAIC Insurance Data Security Model Law, it triggers a documented investigation and notice to your insurance commissioner within 72 hours, and what regulators actually judge you on afterwards is whether you had a real information security program in place beforehand.
This is actually where many generic AI tools fail the insurance test. Ask four questions in writing: where does our data physically sit, does our agency’s data share a database with anyone else’s, will you sign a data-processing agreement, and will you show us an independent security audit. A vendor who cannot answer those in writing is a vendor you cannot defend when a carrier, a regulator, or a commercial client’s risk manager asks.
If you’re evaluating AI for your agency, the security question shouldn’t be an afterthought. It should be the first question you ask. The second question should be: “Can I audit the answer?”
The competitive argument
Here’s the uncomfortable version of this conversation.
Your agency’s value proposition to clients is some combination of: expertise, relationships, service quality, and availability. AI doesn’t replace the first two. But it directly competes on the second two.
An agency running AI answers every call in under two seconds. Yours goes to voicemail after four rings. Their client gets a COI in three minutes. Yours gets it in four hours. Their producers spend 25 hours a week selling. Yours spend 13 hours selling and 12 hours on admin.
Same producers. Same market. Same clients. Different infrastructure.
The client doesn’t know the difference is “AI.” They just know one agency is faster, more responsive, and harder to leave.
The honest part
We build AI for insurance agencies, so we have a bias. And the Google study was funded by Google, who sells the underlying infrastructure, so they have one too.
But 3,466 executives across 24 countries is a serious sample. And the directional finding, that AI in customer service has moved from experimental to mainstream, and that the early adopters are pulling ahead, matches what we see in the industry. The agencies we talk to aren’t asking “should we use AI?” anymore. They’re asking “which AI, and how fast can we get it running?”
If your agency hasn’t started yet, the good news is you don’t need to overhaul everything at once. Start with one workflow: the one that wastes the most time and annoys your team the most. For most agencies, that’s phone handling and post-call documentation. Automate that, measure the result, and decide what’s next.
If you want to see what that first step looks like for your agency, book a 15-minute call. We’ll walk through what the AI handles, what it doesn’t, and how it fits into what you’re already doing.
Georgijus Korobkovas
Founder & CEO