Dirty Martini Volume 22

The Dirty Details

AI in customer service (in some form) is probably the most mature of any of the functional areas in a business. The early days of the chatbot were a form of AI. In our most recent AI Customer Service Survey (Retail + Healthcare leaders) study with Dialpad + Google Cloud, 77% of organizations are now piloting or deploying AI in customer service.

The ‘should we or shouldn’t we’ adoption argument isn’t really worth making any more. The critical component is execution. Execution means connecting AI to the systems, rules, and measurement discipline required to turn “AI in the building” into “AI resolving the issue.”

77%

of organizations are now piloting or deploying AI in customer service.

Key Takeaways:

The “AI that knows” vs. “AI that does” delta is massive.

80% have AI-enabled information retrieval but only 25% have autonomous resolution enabled.
Only 45% have AI directly connected to the action systems it needs to do the work, and 21% of AI-to-human handoffs happen because AI understood the request but lacked the technical ability to complete it. That’s the difference between “AI that can answer” and “AI that can act.”

What this means for B2B tech marketersPosition around outcomes + operating model, not “AI features.” With 77% already piloting/deployed, “we have AI” is not differentiation. Your message has to be “we connect to systems-of-action so AI can actually resolve.”

Cross-channel context is breaking, and voice is where the problem shows up most painfully.

Only 13% of respondents stated they fully preserve customer context when someone switches channels; 53% have partial context and 21% mostly lose it.
Only 33% say full cross-channel history is available at the point of interaction.
This is why “AI can retrieve info” is not the same as “AI can remember the customer.” Memory is architecture, not UX.

What this means for B2B tech marketersIf you can, make cross-channel memory a first-class claim. Only 13% fully preserve context across channel switches. If you can carry history from digital into voice (and vice versa), you can create your own moat around this.

Governance is the real key to autonomy.

Among Planning-stage organizations, 83% say undefined AI decision rules block progress; even for those Piloting, it’s 46% (vs. 25% for Deployed).
It’s important to note that decision rules aren’t something a company figures out after they deploy. Defining governance at the outset may take more time, but it will improve the chances for success.

What this means for B2B tech marketersSell decision rules and human operating models as part of implementation. The market is blocked by governance (83% at Planning). The winning vendors will package “how to define authority, escalation triggers, and human-in-the-loop design” — not just software.

Most teams claim ROI measurement but fewer than half can prove the ROI exists.

68% say AI metrics are tied to measurable ROI, but only 43% have enough data to state whether AI is paying off, a 25-point gap.
Worse, only 46% can distinguish “resolved” vs. “deflected.” This means a lot of teams can report containment (agent avoided) but can’t prove outcome (issue completed).

What this means for B2B tech marketersBe ruthless about measurement language. If prospects can’t distinguish resolved vs. deflected, stop letting them anchor on containment. Bring a measurement POV: resolution quality, customer effort, and “why handoffs happen.”

Incisiv · AI Customer Service

The capability cliff

AI can answer far more often than it can act.

100%

75%

50%

25%

↓ The capability cliff

80%

75%

69%

62%

43%

25%

Information
Retrieval Navigation
& Routing Real-Time
Agent Support Sentiment
& Intent Proactive
Engagement Autonomous
Resolution

The bottleneck

Only 1 in 4 organizations can resolve autonomously.

Source: Dialpad + Google Cloud AI Customer Service Survey (n=150 Retail + Healthcare leaders)

Dive deeper into our cross-industry research and content.

If the customer service AI study has you thinking about execution (systems + governance), these two snapshots widen the lens on what “operationalizing AI” looks like when the problem is activation speed and content capacity — not model quality. This was part of a Five-Part Series with Adobe and Microsoft

Consumer Goods: "The Moment Is the Market"

Brands have poured money into AI and personalization, but response systems still run on old planning cycles instead of the consumer's actual decision moment. This will often leave the investment mostly unrealized.

Speed is now the differentiator:

real-time AI can spot a signal (a basket add, a purchase lapse), but only 8% of the customer journey is personalized beyond a surface level because responses arrive days after the moment has passed.

Content production is the bottleneck:

personalization engines can target dozens of segments, but content teams can typically only produce for a handful. Brands using GenAI in production workflows are seeing 47% higher output and 8% higher conversion.

Data is fragmented across channels:

only 3% of Consumer Goods brands have fully integrated, accessible consumer data, so a shopper who buys in-store, browses online, and calls service gets treated as three different people.

Media & Entertainment: "From Signal to Screen"

Streaming, gaming, and broadcast platforms have the richest audience data of any industry. However, their promotional and activation systems weren't built to match that precision. Viewers still get generic experiences despite highly specific recommendation engines.

Churn is predictable but under-acted-on:

session drop-off and failed searches signal disengagement days before cancellation, and 94% of executives say viewers switch platforms over content quality or discovery friction.

Recommendation depth outpaces creative production:

engines can identify thousands of audience micro-segments, but marketing teams often produce only a handful of promotional assets, leaving most personalization unrealized (only 50% of the M&E journey is personalized end-to-end, with 95% of data still siloed).

Cross-device continuity drives loyalty:

90% of executives say viewers want seamless cross-device experiences, yet fragmented profiles (streaming, gaming, ad, service data all separate) mean subscribers frequently have to "start over" at every screen.

Both reports land on the same core argument, which is that the technology to detect intent exists, but most organizations haven't rebuilt the operational systems. Content production, data integration, and workflow need to act on it in real time.

Straight from the Shaker: The New Value Unlock of PR

Get practical insights and best practices straight from our industry experts as we shake up and serve up our knowledge to help you improve your go-to-market strategies. We share tips each month to help you stay ahead of the game.

We’ve watched the evolution of Public Relations (PR) in B2B tech over the last decade here at Incisiv and longer in other stages of our careers. There was a time when mentions and impressions were meaningful metrics in a marketing strategy. Then the board/CEO/ELT started asking a different question: “How much pipeline did it create?” At that point, PR got demoted from “strategy” to “nice-to-have,” and a lot of comms teams got told to go find a number they were never built to own.

Here’s the twist: In the age of AI, we’re swinging back. Why? Because the metrics are changing. Mentions are becoming a distribution surface again, because LLMs are turning third-party citations into discovery.

The mechanism is simple: in an AI answer world, you’re either referenced, or you’re invisible. Multiple analyses have found AI brand mentions correlate most strongly with branded web mentions and broader third-party visibility (Ahrefs; Seer Interactive).

It’s not the prettiest landing page that wins. It’s the brand that shows up across the open web, in enough credible places, often enough, that the models learn to pull it forward.

Here are some points to consider when putting together your strategies:

Stop treating PR as “awareness.” Treat it as retrieval.

Ahrefs analyzed 75,000 brands and found “branded web mentions” strongly correlate with AI visibility (roughly 0.66–0.71), and YouTube mentions are even higher (~0.737).

PR now compounds SEO/AEO.

Lily Ray analyzed 11 sites hit by a January 2026 Google visibility drop and found every one also saw AI search citation declines, averaging -22.5% across LLMs, with ChatGPT down -27.8%.

You need “mention architecture,”

not one big splash. Seer Interactive’s study (10,000 questions) found page-one Google rankings correlate strongly with LLM mentions (~0.65).

Measure the new outcome: AI-influenced demand.

Similarweb (via Search Engine Land) found users were 2.5× more likely to visit an AI-recommended brand than a competitor, and those visitors consumed more content (12 pages / 11.8 minutes vs. 6.5 pages / 5.6).

With the migration of search and discovery to LLMs, PR (and its ability to get coverage in third-party citations) has again become critical to a marketing team's strategy.

Second Round - Incisiv on the Road…..

On Stage, our Chief Insights Officer, Gaurav Pant, - Wednesday, September 23, 9:00 AM - 9:40 AM ‘Tapping Data Analytics and AI for Productivity and Profitability’:

AI is driving measurable productivity and efficiency gains across retail and CPG—boosting on-shelf availability and streamlining workflows. This session opens with a deep dive on how retailers and brands are embedding AI into operations, followed by a panel of retail and brand executives on:

Where AI agents cut administrative burden and drive the most impact
Using data and technology to unlock productivity and profitability
Guardrails for safe, scaled agentic AI

The Speakeasy at Groceryshop— An Executive Dinner: We are headed for the Four Seasons Las Vegas for a private dinner with grocery executives. Our exclusive sponsor, Scandit, will be joining us for the first time, and we’re super excited to share the latest trends and build great relationships.

The Happy Hour at Groceryshop - Incisiv clients and friends - reserve an hour to hang with the Incisiv gang at RiRa on September 22nd. We’ll have the registration page up next week, but ping us in the meantime if you’d like us to register you!

The Speakeasy at Shoptalk Fall 2026 — An Executive Dinner & Private Distillery Tour - We are diving into the Nashville experience and producing our dinner at Nelson’s Green Brier Distillery. At Shoptalk Fall 2025, we had over 20 high-profile retailers join us and anticipate a similar audience for 2026. We are partnering with Manhattan and Talkdesk but do have one sponsorship still available. Reach out to Mara Dosso (mara.dosso@incisiv.com) for more details.

New York Community Dinner Series. Our next Dinner is on September 17th in conjunction with Manhattan, and we are focusing the conversation on Bringing the Sexy Back to the Store. We have a GREAT group of retailers and brands registered and are looking forward to the evening.

New York Community Dinner Series 2027. We are building our calendar for our quarterly dinner series in 2027. If you'd like to discuss a date or theme that fits your schedule, now is the time to reach out! Email Mara Dosso (mara.dosso@incisiv.com) and we’ll schedule a call.

Believe it or not, NRF 2027 is coming! Our biggest Speakeasy of the year, The Speakeasy @ NRF, will take place Saturday, January 9th at Zuma. Over 50 retailers will join to kick off the 2027 NRF Big Show. For more information on how to get involved, email Mara Dosso (mara.dosso@incisiv.com) or Bill Little (bill.little@incisiv.com).