AI Agent Adoption Hits 79% While Productivity Proof Remains Missing

The Perception Gap

Dialpad now serves 70,000 companies across more than 50 countries, a figure Brian Peterson's Google Cloud Events interview found, a footprint built by the former Google Voice team on Google's backbone and Gemini. In 2025 alone, 79 percent of enterprises adopted AI agent technology, a figure Dialpad's March 2026 Business Wire release reported touting built-in ROI validation, no-code agent building, and governance. The adoption wave is real. What's missing is the productivity proof.

Enterprises are forming strategic partnerships and sponsorships with agentic AI platforms like Dialpad to enhance customer interactions, while investors fund AI agent startups and competitors strengthen their omnichannel CX offerings. That sentence is the nut of this story: the money, the partnerships, and the competitive response are all moving faster than the verified metrics.

Randstad, the world's largest staffing firm, illustrates what deployment at scale looks like. With 30,000 to 40,000 employees across 27 countries (and roughly half a million people staffed globally in any given week, a figure Ryan Henry cited in a Google Cloud Events interview), the company migrated off an on-premises Cisco system and legacy Avaya hardware in six months. The catalyst was COVID, which exposed how chained employees were to physical desks with no mobile app and no flexibility. After the move, employee satisfaction scores shot up, with specific mentions of freedom from the desk. Call performance improved: response rates rose, talk time fell, and transcripts let teams interact more precisely with customers. Stability (no longer worrying whether the server was up) removed a layer of cognitive load from every employee.

The technology still has sharp edges. One Randstad customer runs 200,000 password-reset phone conversations a month, a volume the company argues shouldn't be handled by humans. Dialpad's proprietary AI model measures caller sentiment with 90 percent accuracy, and the data shows a hard threshold: if a customer sits on hold more than five minutes, satisfaction drops by half. The old interactive voice response systems never worked well; callers learned to hammer "representative" until a human answered, only to repeat the whole story because context never transferred. Agentic AI changes that calculus because voice remains the most accessible channel — every person on earth has a phone number, and no custom software is required to use it.

Velocity is the problem nobody has solved. "What came out six weeks ago is already old," a Dialpad leader said in a July 2026 briefing. Regulations shift monthly across states and countries. Much internal AI adoption remains unsanctioned, and security teams are scrambling to rein it in. Telcos still weaponize number porting to lock in customers — a process so painful "God himself would struggle with it."

Hiring Data Reveals Where the Money Flows

Dialpad's job board tells a clearer story than any press release. As of the latest ingest, the company posted 12 new roles in seven days, bringing its live listing count to 35 positions. The salary bands reveal where the pressure sits:

Median band across all 35 roles: $125,000. Range: $56,000 to $270,250.

The freshest batch skews senior. Two legal hires signal the compliance load that comes with autonomous voice agents handling refunds, rebookings, and PII across voice and digital channels. The board data shows zero roles tagged to Vancouver or Kitchener in the current window. That absence is itself a data point: Dialpad's visible hiring momentum right now clusters around its San Ramon headquarters and a Denver satellite, not the Canadian corridors the broader narrative often cites.

Andreessen Horowitz shows a different velocity. Its board listings held at 13 roles with a $104,000–$405,000 band and a $258,000 median, but zero new postings in the past week. The open slots (Partner-level corporate controller, DevX and security automation staff engineers, a GTM executive talent lead) reflect a firm staffing its portfolio-support engine, not a direct AI hiring surge. The contrast matters: Dialpad is hiring operators to ship product; a16z is hiring partners to scale the bets it has already placed.

What the board data captures is the translation layer between "agentic AI" as a keynote slide and "agentic AI" as a payroll line. Every "execute complex workflows autonomously" claim on the marketing site maps to a Senior Manager of Agentic Sales Engineering who can demo that workflow to a Fortune 500 prospect, a Staff Product Designer who can make the handoff from bot to human feel invisible, and a privacy counsel who can argue to a regulator that the audit logs are immutable and the PII protection holds. The hiring is real, the salaries are posted, and the locations (for now) are San Ramon, Denver, San Diego. If Vancouver and Kitchener are the next nodes, the listings will lead the announcement.

Dialpad's own description, "AI-native, Agentic AI-powered omnichannel contact center and communications platform built to transform customer experience at scale," positions the company as a consumer of frontier model capabilities rather than a model builder. That distinction matters. Integrating large language models into a production contact-center stack requires latency optimization, compliance guardrails, and prompt-engineering tooling that differ fundamentally from training runs. The board's salary bands reflect a company staffing for productization, not research.

Rural Connectivity Becomes AI Infrastructure

Tractor Supply, the largest rural lifestyle retailer in the United States with over 2,200 stores and 200 Pet Sense locations nationwide, has made T-Mobile's 5G business internet the backbone of its store technology experience. The partnership addresses a fundamental constraint: rural connectivity. "In rural America, connectivity is a significant challenge," the company notes, and with 50,000 team members dependent on a strong technology ecosystem, the retailer needed a partner capable of scaling a unified internet solution across the continental U.S.

The deployment targets locations with little or no existing internet coverage. T-Mobile's 5G business internet delivers the high bandwidth and low latency necessary to power Tractor Supply's diverse digital solutions across all stores simultaneously, a requirement driven by the volume of concurrent transactions and the demands of real-time AI and IoT workloads. The retailer has already implemented a generative AI solution called Hey Guru and uses IoT technology branded as Tractor Vision to surface customer and product insights. Both depend on the consistent, cloud-configurable connectivity that 5G provides.

Store-level feedback underscores the operational impact. Team members describe the system as making daily work "so much more efficient and easier," with one noting that having the tool "in my hand at all times" makes it "quick and easy" to serve customers. The connectivity layer is not a passive utility; it is the enabling infrastructure for the retailer's digital transformation. As Tractor Supply frames it, reliable and secure connectivity sits at the core of how they innovate customer experiences and build brand loyalty.

This partnership illustrates a broader dynamic shaping the agentic AI field: the performance ceiling for autonomous voice and multimodal agents is increasingly defined by the network beneath them. Low-latency, high-throughput connectivity (especially in distributed or rural footprints) is becoming a prerequisite for enterprises that want to run AI agents at the edge, in-store, or across hybrid environments. T-Mobile's 5G business internet product is positioning itself as that prerequisite, and Tractor Supply's deployment across 2,400-plus locations represents one of the largest real-world validations to date.

The arrangement also highlights a competitive flank for carriers. AT&T, based in Dallas, continues to market fiber and wireless alternatives, offering 1 GIG fiber for $50 per month to new customers and up to $800 per line to switch, but T-Mobile's fixed wireless access play for enterprise has secured a flagship reference account in a vertical where uptime and geographic reach are non-negotiable. AT&T's stock has declined roughly one-sixth over the past year amid fiber infrastructure investment pressures, and its Q2 2026 earnings will test whether its strategy can reverse the trajectory.

For AI platform providers, the lesson is clear: agentic AI does not float above the network. Its responsiveness, reliability, and geographic viability are bound to the connectivity layer. Partnerships between carriers and enterprises deploying AI at scale (like Tractor Supply's with T-Mobile) are becoming the de facto reference architectures for the next wave of omnichannel customer experience.

Incumbents Counterpunch

The agentic AI wave Dialpad is riding has forced the two largest incumbents in cloud contact center software to accelerate their own generative AI roadmaps. Genesys and Five9 together control the bulk of the CCaaS market. Genesys serves 8,000-plus customers across more than 100 countries and processes roughly 7 billion conversations each quarter. Five9 just earned its eighth consecutive Leader placement in the Gartner Magic Quadrant for CCaaS. Both are now repositioning around the same autonomous-agent narrative that Dialpad has made central to its brand.

Genesys moved first on the acquisition front. In January 2024 it bought Radarr Technologies, an AI-powered data analytics company, to layer real-time sentiment and intent signals onto its existing Genesys Cloud CX and Genesys DX platforms. The combined stack now markets itself as "the agentic AI CX platform for the enterprise," promising AI agents that "reason, act and adapt across experiences." The company backs that claim with third-party validation: in the 2025 Gartner Critical Capabilities for CCaaS, Genesys was the only vendor ranked number one in three of five use cases; IDC's 2025 MarketScape for general-purpose conversational AI platforms named it a Leader; and Forrester's Q2 2025 Wave for CCaaS placed it in the Leaders quadrant. Those awards arrived alongside self-reported metrics — 60 percent of calls resolved by AI, 20 percent lower costs, a 25-point CSAT lift — that Genesys publishes without independent audit but that signal the KPI framework buyers now expect.

Five9's public response has been quieter but deliberate. Its eighth straight Gartner Leader designation cited "Ability to Execute and Completeness of Vision" — language that in recent cycles has come to mean demonstrable generative AI features in production, not roadmap slides. The company has not disclosed a Radarr-scale acquisition, but its partner ecosystem (Salesforce, ServiceNow, and a deep bench of system integrators) mirrors Genesys's integration strategy, giving enterprises a low-friction path to embed Five9's AI agents into existing CRM and workflow tools.

The competitive dynamic is now a feature race rather than a platform race. Genesys emphasizes "purpose-built for CX" AI that connects customer intent to people, systems, and workflows; Five9 leans on its Gartner momentum and established enterprise install base. Both are chasing the same metric Dialpad introduced to the conversation: the percentage of interactions an autonomous agent can close without human handoff. Genesys says 60 percent; Dialpad's public case studies cite similar ranges. Five9 has not published a comparable figure, a silence that analysts read as either competitive caution or a gap yet to close.

For buyers, the convergence means RFPs now ask for agentic capabilities by default — reasoning, multi-step task execution, cross-channel memory — rather than treating them as experimental add-ons. The vendors' hiring patterns confirm the shift: Genesys's engineering hubs have posted applied scientist roles focused on LLM fine-tuning and low-latency voice inference, while Five9's recent requisitions cluster around prompt engineering and evaluation frameworks for generative dialogue. The talent war is the clearest signal that the feature gap is narrowing, and that the next differentiation layer will be deployment speed, compliance tooling, and the ability to prove ROI in regulated verticals.

Productivity Claims Meet User Reality

The marketing claims are specific. Dialpad's own materials promise "instant, autonomous resolution" for refunds and rebookings, "unlimited scale" across voice, chat, SMS, and email, and a platform that "executes complex workflows autonomously" while delivering "AI insights and automations" that help reps "close more deals." The company positions its agentic AI as a productivity multiplier (automatic note-taking, real-time coaching, rich contact profiles pulled from CRM integrations) all designed to keep agents productive whether they're at a desk or switching to a mobile device mid-call.

What the research does not contain is a single verified, third-party productivity metric from an early enterprise adopter. No case study cites a percentage reduction in average handle time. No customer quotes a measurable lift in first-contact resolution. The closest thing to field data are the app-store reviews, and they tell a different story. One user reported "constant spam calls" (three per hour despite filters) and a denied refund after annual cancellation. Another described the Android app ringing endlessly without an answer option, forcing a phone restart. A third detailed "dozens of hours wasted" and 80-plus emails over a disconnected-call issue that support blamed on the customer's network for weeks. Billing complaints recur: extra licenses charged, fees piled on, tickets opened and unresolved for months. A positive review exists (a rural property manager citing reliable calls and improving quality) but it reads like a small-business use case, not an enterprise contact center deployment.

The workforce signal is clearer in the hiring data. Dialpad added 12 roles in the past seven days alone. The newest postings cluster around product, compliance, and a newly explicit "Agentic Sales Engineering" function. That hiring velocity suggests the company is staffing for platform depth and enterprise-grade governance, not just feature velocity.

Competitors are moving in parallel. Genesys and Five9 have both launched generative AI features to defend their installed bases. The competitive pressure is visible in the hiring: agentic-specific engineering and sales-engineering roles are appearing across the category, not just at Dialpad.

The tension is straightforward. The platform vendors are building for autonomous workflows at scale (refunds, rebookings, appointment scheduling, order processing) and hiring the teams to deliver them. Early user feedback, where it exists publicly, flags reliability, support responsiveness, and billing transparency as live problems. Productivity gains at the enterprise level remain unquantified in the public record. The next quarter of deployments will either produce the first hard numbers or deepen the gap between the demo and the floor.

Regulation Arrives Before Standards

The European Union's AI Act entered into force in August 2024 and begins its phased enforcement in 2025, establishing the world's first comprehensive regulatory framework for artificial intelligence. For agentic AI platforms — systems that can autonomously plan, execute, and adapt across multi-step tasks — the Act's risk-based classification creates immediate compliance questions that U.S.-based vendors like Dialpad, Genesys, and Five9 are only beginning to address in their European go-to-market strategies.

The Act categorizes AI systems into four risk tiers: unacceptable, high, limited, and minimal. Agentic AI deployed in contact centers likely falls under "high-risk" when it makes or materially influences decisions affecting individuals' access to essential services, creditworthiness, or employment, scenarios common in banking, insurance, and telecom support. High-risk systems face requirements for risk management systems, data governance, technical documentation, human oversight, accuracy and robustness testing, and post-market monitoring. They also require conformity assessments before market placement and registration in an EU database.

Article 14 mandates a human-oversight architecture for high-risk systems — a requirement that cuts to the core of what makes an agent "agentic." Vendors must demonstrate conformity assessment readiness and technical documentation for high-risk use cases before they can sell into Europe. The Act's extraterritorial scope means that even U.S.-hosted platforms serving EU customers must comply, a reality that complicates the "single platform, global deployment" model these vendors have historically sold.

Dialpad's own hiring signals suggest the company is treating regulatory readiness as a priority. The board lists a Corporate Counsel for Telecommunications & Regulatory in San Ramon alongside a Senior Product and Privacy Counsel in San Diego. These roles sit alongside the Sr. Manager, Agentic Sales Engineering position in Denver, indicating that regulatory and product teams are scaling in parallel.

Genesys and Five9 face the same regulatory surface area. Both operate substantial European footprints. Neither company's public filings have detailed specific EU AI Act compliance roadmaps for their generative AI agent features (Genesys Cloud AI and Five9 Genius AI) though both have emphasized data residency and GDPR alignment in prior communications.

The research available for this section does not include direct statements from Dialpad, Genesys, or Five9 on their EU AI Act preparation, nor does it contain analyst reports specific to agentic AI classification under the Act. The hiring evidence shows Dialpad investing in regulatory legal capacity at the same time it scales agentic sales engineering, a signal that compliance is being treated as a product enabler, not an afterthought.

For enterprises evaluating agentic AI vendors, the regulatory horizon introduces a new procurement criterion: can the vendor demonstrate conformity assessment readiness, technical documentation for high-risk use cases, and a human-oversight architecture that satisfies Article 14? Vendors that cannot may find their European pipelines constrained regardless of technical capability.

The Desk That Moved

Randstad's employees used to be chained to physical desks. Now half a million people a week work from wherever the signal reaches. Tractor Supply's rural store managers pull up Hey Guru on a handheld device in an aisle where fiber never ran. The connectivity layer — T-Mobile's 5G, Dialpad's Gemini backbone, the carrier partnerships underneath every agentic platform — is the new desk. It moves. It scales. It fails when the network fails.

The 79 percent adoption figure will climb. The hiring boards will fill. The regulators will write the rules that the July 2026 briefing warned were shifting monthly. But the picture that holds is the one Randstad's satisfaction scores captured: freedom from the desk, provided the infrastructure doesn't break.

Working in frontier tech? Zero G Talent tracks the openings: see every open Dialpad role, browse frontier tech jobs, openings at Andreessen Horowitz, and the people building the field.