Insider Report August 2025

Curated Enterprise Communications News and Insights from August 2025

Featured

GPT-5 Schizophrenia: GPT-5 reviews show significant variation, attributed to a substantial price increase and the model’s inconsistent performance across different applications. The new $200-per-month fee for ChatGPT, a tenfold increase from the Plus version, has sparked considerable user reaction. This price point surpasses the cost of many established software suites and typically exceeds individual software budgets. Despite widespread concern regarding the price, some users are reportedly seeking higher tiers with greater token access, suggesting a specialized need for the model’s capabilities.

The diverse user experiences reflect this inconsistency. GPT-5 is simultaneously described as “revolutionary” and “useless,” indicating its utility highly depends on the specific task. While it scores highly on intelligence tests, it struggles with basic functions like maintaining a consistent conversation persona. It excels at complex analytical tasks but underperforms in creative writing.

This variability signifies a shift from the expectation of uniformly improved AI models with each iteration. GPT-5 is characterized by a “jagged” performance profile, demonstrating strengths in certain areas and weaknesses in others. This uneven capability is not a defect but rather an indication of future AI development, where increases in parameters and capabilities do not guarantee universal improvement. The traditional assumption of linear AI progression no longer applies.

GPT-5 Pro utilizes a distinct architecture involving parallel reasoning chains that self-correct to arrive at a single answer, an approach that prioritizes accuracy and differs from previous AI models. This architectural choice marks a fundamental change in AI development, emphasizing the importance of selecting the appropriate AI for specific tasks rather than relying on a single “best” model. Competitive advantage will increasingly stem from architectural design choices, and the most advanced model may not always be the most suitable for general work. Companies that recognize AI as a collection of specialized tools, rather than a singular capability, will gain a strategic advantage.

The New Prime Directive: The new prime directive in CX is to reduce employees. Automation (and AI) and improved self-service are two primary ways to accomplish this. AI is getting all the attention, but improving self-service is an equal partner. 

Not long ago, brands intentionally limited self-service capabilities to protect the customer. By limiting what customers can do on their own, you limit how much trouble they can cause. For example, changing or canceling flights should be easy via self-service, but canceling just the outbound flight of a round trip could be a mistake, so it requires a call. That way of thinking is out. Digital native customers are demanding more self-service capabilities. Today’s CX requires supercharging AI automation and increased self-service capabilities — regardless of channel. Password resets, replacement credit cards, returns, status inquiries, etc., are now more than likely to be possible without human interaction. This prime directive is new and only whispered but has significant implications across CX. 

General News

You’ve Got Less: AOL will discontinue dial-up internet service on Sept. 30, 2025, ending a 34-year chapter that began in 1991. The shutdown also affects associated legacy software, including the AOL Dialer and AOL Shield browser. While only a modest number of users remain — often in rural areas still underserved by modern broadband — the move highlights the fading viability of dial-up as a fallback connection. The sunsetting of AOL’s dial-up service marks a technological ending and a cultural milestone — the last mainstream link to the early internet era falling away, even as broadband finally becomes ubiquitous. Note: AOL lives on; the dial-up service is what’s going away. 

We the People Own Intel! The US Government is now the largest shareholder of Intel (10% of Intel for $8.9B — no board seat). That’s on top of the $2.3B the government has paid Intel under the CHIPS and Science Act. Now every American has a stake in this bellwether that missed ARM, AI/GPU, advanced manufacturing (10 and 7 nm), and mobile (it turned down Apple to make its iPhone chip). President Trump and CEO Lip-Bu Tan announced the deal on Aug. 22. On Aug. 7, Trump had demanded the Intel CEO resign because he’s “conflicted” (with US and Chinese alliances), adding “there is no other solution to this problem.” 

No Western chip company can compete with TSMC. The Taiwanese company has a lock on state-of-the-art chip production. Practically, a lack of competition means higher prices. Strategically, China believes it owns Taiwan, so it could act on it by attempting to restrict chip exports to the US, as the US does to China (including chips made in Taiwan). China got several international brands, including the NBA and Marriott, to apologize for suggesting Taiwan is an independent country. 

No Discount for You: Microsoft is ending the discount on its online services products purchased through volume licensing programs, effective Nov. 1, 2025. Historically, businesses could obtain volume licenses for products like Microsoft 365, Azure, and Dynamics 365 at varying price points across different price levels. The recent announcement means all Online Services sold under an Enterprise Agreement or Products and Services Agreement will have a single, consistent price across each tier, publicly available on Microsoft’s website. This change will prevent some customers from securing the same discounts they had in previous contracts, though standardized prices may be lower for some. While license cost discounts are being removed, concessions on multi-year terms and value-added services may still be sought. This is expected to present a significant price increase for many customers, likely to fund Microsoft’s substantial AI investments, and consequently, it presents a considerable opportunity for alternative providers of Office and email services.

MS also made a larger-than-usual number of other Teams announcements this month, the highlights including Teams Copilot defaulting to not transcribing meetings where it’s permitted — improving privacy and control; Copilot Chat in Outlook summarizing long threads without a full Copilot license; Microsoft Places gaining a web management portal, more desk-status modes, and upcoming map-based booking; Teams Rooms on Android dynamically resizing the in-room video tile in October based on attendee count; and Teams desk phones adding AI prompts, transfer summaries, and real-time transcription.

Google Working to Secure Android: As we recently predicted, Google will begin to require all Android apps — including those installed via sideloading or third-party stores — to be tied to a verified developer, starting in some geographies in 2026 and everywhere in 2027. Verification will require real identity details for individuals and organizations (legal name, address, email, phone, and for companies, a verified website and D-U-N-S number). Google frames the shift as targeting higher malware rates from internet-sideloaded sources while keeping distribution “open.” In the context of Microsoft’s MDEP release, this is Google tightening the certified Android perimeter — taking beginning steps that could lead to its own “hardened” version of Android by baking a first-party trust layer into the platform; enterprises and banks will welcome it. No impact to OEM forks.

Cool Like a GlassHole: Google Glass launched in 2013 for $1,500. It was ahead of its time, but in hindsight, the launch was a precursor to an inevitable wave. We called Google’s early adopters “Glassholes.” Snap glasses were cool, but Meta’s Ray-Bans crossed over 2M units sold. In Q4 2024, Meta shipped nearly 900K units, and sales continue to grow. Like all good tech, Meta’s frames disappear. Glassholes turned the users into the Borg; Meta offers style and function: a 12MP camera translates signs, describes landmarks, and even guides the blind. Apple and Google also have their sights on glasses. 

AI News

OpenAI Calling: OpenAI has introduced a Realtime API with features for building voice agents. The API now includes support for remote MCP servers, image inputs, and phone calling via SIP. Additionally, OpenAI released gpt-realtime, an advanced speech-to-speech model that can follow complex instructions, use tools, and produce natural-sounding speech. It can also interpret system messages and developer prompts more effectively, including reading disclaimers, repeating alphanumerics, and switching languages mid-sentence. Two new voices, Cedar and Marin, are exclusively available in the Realtime API. This single-model API reduces latency, preserves nuance in speech, and produces more natural, expressive responses. The API also supports SIP to connect apps directly to the PSTN, calling systems, desk phones, and other endpoints.

The comms sector loves what Generative AI has done for it; now, we shift to what generative AI will do to it. This announcement commoditizes the voice interface for AI assistants. Anyone can now provision a phone number to forward to the OpenAI SIP URI. How many prompts does it take to become a unicorn? If Sam Altman is right, this will accelerate the complete automation of customer service. He’s wrong, because automation is the sector’s newest “last mile” problem. The sector needs to remain focused on the outcomes ball.

The Report That Moved Markets: The State of AI in Business 2025 reveals that most enterprise generative AI initiatives are not achieving rapid revenue growth. Only about 5% of AI pilot programs succeed in accelerating revenue, with the majority having little to no measurable impact on profit and loss. The data also indicates that over half of generative AI budgets are allocated to sales and marketing tools, despite MIT finding the highest ROI in back-office automation. The report further highlights the widespread use of (unsanctioned) “shadow AI” and the ongoing difficulty in quantifying AI’s impact on productivity and profit. The primary issue is a learning gap for both the tools and organizations. The headlines were that AI is failing, but the report highlights flawed implementations. 

No-Pilot VFR: Microsoft has announced that it will not proceed, for now, with the planned feature enabling Copilot to analyze on-screen content in recorded and transcribed Teams meetings (Message ID MC998512). Initially scheduled for rollout beginning mid-March 2025 in targeted releases and late April for general availability, this capability would have allowed Copilot to parse visual content in addition to transcripts and chat. The feature is now paused as of Aug. 5, 2025, with no new timeline provided, and Microsoft has said updates will come via Message Center. This delay highlights the challenges of adding AI to sensitive visual data and suggests Microsoft may be taking extra time to address privacy, accuracy, and reliability concerns. It also allows Danto to stress one of his key advice points to buyers: Never buy solutions based on “the roadmap,” because they may never come.

ChatGPT 5 Reset: Recently, OpenAI unveiled GPT-5, an AI model that has generated significant user dissatisfaction, particularly among ChatGPT Plus subscribers. Many users express a preference for previous models, such as 4o and 4.1, citing a perceived limitation in functionality with the new release. GPT-5 Thinking is now limited to 200 messages per week for Plus subscribers, who also report losing access to various AI models previously available, as OpenAI asserts GPT-5 can reason as needed.

The considerable backlash stems partly from Sam Altman’s pre-release hype, which suggested a revolutionary advancement. Despite GPT-5 outperforming its predecessors in benchmarks, it represents an incremental upgrade rather than the monumental shift some expected, similar to the initial launch of ChatGPT. Users report that GPT-5 is performing worse than 4o, and while this requires further testing, paid subscribers feel disadvantaged by the new release. OpenAI faces pressure to resolve launch issues like slow and poor responses to retain its user base.

This situation hints at an impending “AI correction,” where expectations meet the reality of slower, incremental progress. Historically, significant AI milestones have been measured in decades, not months, as evidenced by Deep Blue beating Kasparov in 1996 and Watson winning Jeopardy in 2011. While continuous model improvements are anticipated, the revolutionary AI many envision is still years away.

Surprise! Meta at the Center of Ethics Issue: A leaked internal Meta document revealed policies allowing its AI chatbots to use romantic language with children and to create statements that demean people based on protected characteristics. Following a Reuters report, Meta confirmed the document’s authenticity, then revised and removed parts of it, claiming it was a mistake. This incident raises questions about Meta’s ethical standards and AI ethics at our private AI vendors in general. While Meta is bolting on some ethics to its models, other companies, like Anthropic, integrate ethics into their model training. That seems like a stronger approach, but we don’t really know the underlying AI ethics that power the training. 

While We’re on Meta: Meta has paused hiring in its AI division — described internally as “basic organizational planning” — following a leadership reshuffle. The move comes after an expensive talent sprint with outsized stock offers that raised questions about sustainability and dilution. It also lands during a broader AI selloff, as investors reassess stretched valuations and demand more evident proof of real-world returns. Net effect: Meta looks to be shifting from land-grab to ROI discipline, signaling a phase of consolidation where execution matters more than headline hiring. 

Also, Meta signed a deal to license Midjourney’s text-to-image “aesthetic technology” for future models and products, with collaboration between the companies’ research teams. The move aims to elevate visual quality across Meta’s AI tools as it races OpenAI and Google. Third-party licensing should speed feature rollout and reduce content-creation costs, but it also spotlights gaps in Meta’s in-house stack and adds dependency and IP-risk trade-offs.

AI Slope of Enlightenment: A tough month for AI. Sam Altman even said we’re in an AI bubble. Meta freezes AI hiring after a lavish spending spree. MIT finds 95% of enterprise AI projects are failing. It is not doom, it’s the slope of enlightenment. Let’s play back the prophets of doom more realistically. 

• GPT-5’s Flawed Launch: Sam Altman admits to “totally screwing up” the GPT-5 launch, which users found cold, impersonal, and inferior to GPT-4, mainly due to exaggerated hype.
• Meta AI’s Ill-Timed Restructuring: Meta’s AI division restructured into four units and froze hiring after months of high compensation. Meta’s restructuring isn’t a retreat; it’s preparation for the next phase. Their new chief AI officer’s internal memo, which leaked yesterday, doesn’t sound like capitulation: “Superintelligence is coming, and to take it seriously, we need to organize around the key areas that will be critical to reach it.”
• AI Pilot Failures: An MIT study reveals 95% of corporate generative AI pilots fail to deliver meaningful returns, attributing failures to implementation, not AI limitations. The report also references startups jumping from zero to $20M in revenue within a year, and “exponential gains” in productivity from established companies. Chatbots are easy; implementations with +ROI are complex. 

The Chat experience is maturing, but AI isn’t. Agentic solutions are just getting started. The biggest bottleneck today is chip supply; tomorrow, power. OpenAI has built smarter models that they literally cannot deploy because they don’t have the chips for new and current models. If AI interest were cooling, we would have a chip surplus.

The easy stuff is over. The chatbot gave everyone, technical and not, access. We don’t see pharmaceutical companies accelerating their designs or logistics companies optimizing routes. The slope of enlightenment changes the narrative from “AI will change everything tomorrow” to “AI will change everything, but it’s going to be messy and take work.” It’s time for the slope. 

Microsoft LLM: Microsoft AI (MAI) is developing its own LLM (instead of relying solely on OpenAI). MAI has its own foundation, team, infrastructure and purpose-built models. It recently previewed MAI-Voice-1, a natural speech generation model available in Copilot Daily and Podcasts, as well as a new Copilot Labs experience. Additionally, MAI has begun public testing of MAI-1-preview, its first end-to-end trained foundation model, on LMArena. Crude comparisons impressively place MAI about 6 months behind the leaders. 

Meetings and AV 

David Danto is the primary author of this section. 

RTO Hype: Despite the flood of recent RTO headlines, hybrid work has held steady, and major city office tower vacancies are growing. As an example, the 50-story Worldwide Plaza in Manhattan is about 40% vacant today and might jump over 70% if Nomura fails to renew (likely). That kind of hole is already pushing even top-rated bond tranches toward losses and forcing painful resets for aging Midtown assets. The splashy wins at a few new A-plus towers are real — but they mask a bifurcated market of legacy buildings with debt sliding into distress. RTO press releases <> real estate recovery. 

Teams Certification on Cisco Devices: Cisco has reached a milestone with its Teams Rooms portfolio: Every Cisco device is now recertified to run on Android 13 with the release of RoomOS 11.29. The certification is valid through August 2027, with software support extending until 2029. This update also brings the latest security patches and unifies the entire portfolio on a single OS version. While not a flashy announcement, it underscores Cisco’s commitment to long-term device support and security, a key differentiator in enterprise collaboration where longevity and consistency are critical.

Google Adds Hands-Free Check-In: Google introduced Meet automatic room check-in that uses ultrasound proximity to detect when a user is in a Meet-equipped room. When you join via Companion mode, Meet can recognize the room and check you in automatically, reducing echo and double-join mistakes. The capability is tied to Meet hardware and Chrome, with admin controls to enable or disable it. It’s a small — but meaningful — UX improvement that nudges customers toward Google’s integrated room stack while trimming help-desk friction.

Logi Sight Adds Meet: Logitech’s COT solution — Logitech Sight — added certification by Google for its Meet platform, making the Sight the only COT product on the market certified for Teams, Zoom, and Meet. The other brands work but have not achieved certification at this point.

Rebates on Sennheiser Team Connect Bars: Sennheiser is running a limited-time US rebate: $150 back on TeamConnect Bar S and $300 back on TC Bar M for purchases made Aug. 1-Oct. 31, 2025, with claims and receipts due by Nov. 30 via email. The TC Bars (which we reviewed here) are well-suited for smaller rooms but may struggle as the target locations get longer. This is the first cash-back deal in videoconferencing that I can remember, in an industry that has previously supported trade-ups and competitive trade-ins. I’m not sure how many large enterprises would be interested in the belief that they didn’t get the best deal up front, so this is likely targeted to the SMB market. 

New-ish DTEN Videobar: DTEN announced its Bar Connect — a BYOD version of its DTEN Bar that shifts from a native rooms appliance to a plug-in USB-C bar for Zoom, Teams, or Meet. Unlike the original Bar that runs Zoom Rooms or MTRoA, Bar Connect moves compute to the laptop/PC while the bar provides camera, mics, speakers, and switching managed through the DTEN app. DTEN is the latest manufacturer to repackage an integrated bar-with-codec into a BYOD SKU to capture the growing installed-PC and bring-your-own-laptop segment. At first glance, other than the removal of the internal codec, the device is identical to the earlier version of the DTEN bar, which we reviewed here.

Lenovo Partners With Jabra: Jabra and Lenovo just unveiled the second-generation PanaCast 50 Room System for Microsoft Teams Rooms, pairing the existing 180° 4K PanaCast 50 bar with Lenovo’s ThinkSmart Core Gen 2 and offering either USB or IP touch controllers. The update focuses on heavier AI workloads and deployment flexibility — adding USB-C BYOD, improved cable management, and global availability — with an MSRP of $4,899 (USB controller) or $5,799 (IP controller). The bar hardware itself is unchanged; the bigger story is a more capable compute and controller choice to meet AI-enhanced collaboration features without complicating installs. The broader signal: Lenovo is leaning into partnership over ownership — a contrast to the HP-Poly experience many view as a cautionary tale — suggesting that co-developed bundles with best-of-breed partners may be a smarter, lower-risk way to grow share in meeting rooms. 

Steelcase Sold: HNI — an Iowa-based manufacturer of workplace furnishings with brands like HON, Allsteel, and Gunlocke — has agreed to acquire Steelcase in a $2.2B cash-and-stock deal expected to close by year-end. For many years, Steelcase has pushed to make furniture part of office technology — experimenting with collaboration tables and embedded display concepts — with limited success. Lower demand for office furniture may be driving sector consolidation. Data from WFH Research’s Survey of Working Arrangements and Attitudes showed the remote vs. in-office mix has held steady for years. 

Largest North American Digital Sign: Assembly Atlanta installed a 140-foot-tall, 22-foot-wide freestanding LED tower built by Nanolumens — described as North America’s largest of its kind — at the complex’s entrance. Beyond the spectacle, this is a solid reference build for dvLED at architectural scale.

First Australian Zoom ISV Partner: Spacera, a company specializing in Meeting Room as a Service (MRaaS), has become the first Australian partner in Zoom’s ISV Exchange Program. This partnership integrates Spacera’s meeting room management platform with Zoom Rooms, aiming to enhance the reliability of collaboration spaces. The integration includes AI-powered room-readiness checks, automated issue resolution, occupancy analytics, and support for various AV ecosystems. This move signifies a growing trend in the industry toward more robust and automated management of meeting room infrastructure.

Customer Experience

A Tale of Two MQs: Gartner published a Magic Quadrant for CPaaS last month. This month, Gartner published an MQ for Conversational AI (CAIP). The common thread is CX. 

CPaaS MQ Quotes: CPaaS allows businesses to build communication workflows that digitally operationalize CX. By 2029, 95% of global enterprises will leverage CPaaS to operationalize CX. CPaaS standard features include messaging channel APIs, conversational bots, APIs for payments, and customer success programs, to name a few. CPaaS providers are prioritizing growth through the deployment of more integrated, seamless solutions infused with AI to deliver value-based outcomes …. Key investment areas … include expanding conversational and omnichannel capabilities and advancing rich conversational experiences through enhanced messaging platforms, such as RCS and WhatsApp. They are also forming partnerships and enabling integrations with leading AI vendors, such as Amazon, Anthropic, Google, Hugging Face, Meta, Microsoft, and Mistral. Several vendors are evaluating agentic AI possibilities. 

CAIP MQ Quotes: CAIPs are defined as SaaS products that enable the development of chatbots, virtual assistants, and CAI agents. Typical use cases for CAIPs include customer interaction automation, employee assistance, and sales and marketing automation. CAIPs provide multiple coding options: low-code, no-code, GenAI-assisted, and pro-code. CAIPs should provide customizable techniques for natural language processing, workflow building, and integrations with back-end systems and data. Standard features include voice experiences using telephony and multichannel connectivity. Many vendors are repositioning as AI agent platforms.`

CCaaS, CPaaS, and CAIP are separate tools with meaningful overlap in the CX space. We are seeing significant partnerships and M&A across all three areas. The emergence of a CX MQ is inevitable; however, not all vendors represented in these three reports will make it into a consolidated report. That’s just three MQs; more exist around CRM, ticketing systems, customer engagement centers, and customer data platforms. The 2025 MQ for CCaaS is expected soon. 

What’s sad about the CAIP MQ is the low bar for Leaders. Here are some cautions from CAIP MQ Leaders: Light investment in research. Less composable. Complex portfolio that’s difficult to navigate. Does not hold patents, non-diverse customers, generic strategy. Immature solution. Also, the inclusion criteria stipulate that all providers offer a standalone solution, while most CCaaS providers position CAI as an add-on. The only dots that are also active in CCaaS are Google and Sprinklr (and NiCE after it completes its acquisition of Cognigy). 

For the Children: NiCE and RingCentral renewed their long-term agreement for RingCentral Contact Center to be powered by NiCE CXone Mpower. This extension continues a partnership established in 2015, focusing on go-to-market strategies, sales, onboarding, and post-sales engagement. It combines industry-leading UCaaS and CCaaS. A key aspect of the renewed partnership will be re-energizing the channel partner ecosystem. These companies would likely love to get a divorce, but seemingly agree that the partnership has value to both providers. The leadership at NiCE, in particular, seems determined to expand reach through partnerships. The arrangement inherently positions RingCentral’s RingCX as an entry-level solution, which is reasonable compared to CXone, but likely infuriates team RC. Note: RingCentral’s branding survives in most of its partnerships. 

Twilio Finally Launches RCS: Twilio announced the general availability of RCS messaging. This allows brands to send branded, interactive communications to customers. Existing customers can upgrade from SMS without code changes. Twilio’s RCS manages device compatibility and carrier onboarding, includes SMS fallback, and offers verified business branding. It is available in over 20 countries and 55 carriers. It took Twilio longer than expected, but it’s still early. RCS will rapidly become a critical component of omnichannel CX. RCS deployments are slowly opening up to more companies.

Avaya Voluntary RIF: CXToday reported that Avaya has offered a voluntary exit package to all employees, following a series of layoffs around the globe that began in late 2024. This move is likely aimed at prioritizing profitability over growth. These actions raise questions about Avaya’s long-term future and its strategy, particularly as its priority is designed to protect its existing customers rather than foster growth. Avaya has not confirmed the news. CEO Patrick Dennis responded on social, saying the author was a “pleeb.” Dennis did not respond to a request for additional information. 

Separately, Avaya announced that it has resolved its lawsuit with Cameron Weeks, Bracken Fields, and Back Ventures LLC over its acquisition of Edify. The terms of the settlement are not public. Weeks and Fields co-founded Edify (and Sharpen) but were not part of the acquisition by Avaya. The two are expected to launch a new communications company this fall. 

UC and Messaging

RC Fresh AIR: RingCentral recently announced enhancements to its AI Receptionist (AIR) product. A new standalone version, AIR Everywhere, can now integrate with any phone system. The updated AIR also offers appointment booking capabilities with Google Calendar and Microsoft Outlook, and supports additional languages, including British English, Australian English, Spanish, and French. RingCentral AIR will be released in the UK and Australia by the end of September. These developments position AIR as a significant addition to UCaaS solutions, extending AI’s impact beyond the contact center. Also, AT&T expanded its Office@Hand Portfolio by adding RingCX CCaaS and RingSense Conversational Intelligence Solutions. 

MULTIVERGE: ComTel Spa signed an agreement with NEC Europe Ltd to acquire NEC Italia Spa and NEC Nederland BV (which also owns a branch in the UAE). The companies generated revenues of EUR49.7M last year. 

This was the final step in NEC’s exit of its UNIVERGE UC portfolio, a process that started just over a year ago. The NEC UNIVERGE global brand of PBX and key systems went to several buyers. At one time, in terms of ports, NEC was the largest global PBX provider, but like AT&T, Siemens, Alcatel, Toshiba, and more, it did not successfully transition to UCaaS. 

Intermedia was quick to acquire the UNIVERGE Blue UCaaS and CCaaS offerings. Intermedia was the underlying supplier of these and other NEC cloud services and continues to operate them in North America and EMEA. It is phasing out the UNIVERGE branding but retaining the three partner go-to-market modes (agent, reseller, and a hybrid version called Co-Op). In partnership with Intermedia, NEC continues to offer UCaaS in Australia and Japan. There is no cloud offering in APAC or LATAM. 

Two NEC partners acquired major parts of the premises-based business. Forerunner (US) acquired it in the Americas (excluding Brazil). Forerunner also acquired worldwide rights to the 3C IP platform. Comtel acquired the EMEA businesses (and sources 3C licenses from Forerunner). Metodo acquired NEC UNIVERGE in Brazil. NEC retained the business in Australia and Japan.

BenchMark

Dmitry Netis (LI, email) of The Benchmark Company (now part of StoneX Group) provides Insider subscribers with financial insights quarterly. Dmitry leads M&A advisory and serves as a Managing Director and Co-Head of Technology Investment Banking at Benchmark. Mr. Netis has been immersed in enterprise communications for over 15 years as one of the first Wall Street research analysts covering this space, serving institutional and private equity clients. Benchmark offers capital markets, private placement, research, and M&A advisory services.

Every quarter we aim at capturing big trends related to macroeconomic activity and their impact to the overall cloud communications/CX sector, investor sentiment, vendor moves, and M&A. Depending on the news flow of each, each quarter may have a different tone to it. Macro is single-handedly the most important factor. There are policy shifts and economic uncertainty in the market each day to make people anxious and wondering what’s next. For businesses, uncertainty is always a negative as it slows investment and spending decisions. More than half way through the year, the U.S. economy is toying with higher levels of inflation (due to tariffs) and slowing labor force participation. European markets are faced with weaker growth, while a deflationary cycle is now hitting Asia. China GDP projected at 4% or below.

The Fed’s pivot to a more dovish stance boosted markets. Last week, the Fed held its annual summit in Jackson Hole, where Chairman Powell discussed the near-term strategic policy framework. The takeaways are as follows. First, the Fed seems closer to cutting rates in September than the last time Fed Chair Jerome Powell spoke publicly. He said that “risks to inflation remain tilted to the upside, while risks to employment are tilted to the downside, " creating a "challenging situation" for the Fed's dual mandate of stable prices and maximum employment. 

Moreover, the effects of tariffs on consumer prices remain clearly visible, albeit their uncertainty is now more priced in. The GDP growth has slowed in the first half of 2025 to 1.3%, while the labor market is in a "curious kind of balance." The Fed is preparing to cut rates mainly due to tariffs’ less damaging impact on inflation than expected. This has given a boost to indices like the Russell 2000 & S&P 500, which have subsequently surged, and bonds have rallied, signaling that we are potentially entering the bull market again following April’s trough. The new cycle can typically last 1.5-2 years before the Fed turns hawkish again, should the economy recover more strongly than is currently expected or inflation reaccelerates. The minor correction in stocks, given the high likelihood of a cut that is now priced, could be possible if the Fed decides not to cut rates after all, because either GDP growth is better or inflation is higher than expected; however, it may give rise to investors to buy the dip.

Unpacking the downside risk to the labor market first suggests that the July employment report told the Fed that perhaps there is more weakness in the labor market now than they thought, creating a shift in policy to put in a few rate cuts to shore up the labor market. Second, the Fed is now saying that it won’t tolerate inflation above or below its target rate of 2%, signaling it is prepared to move swiftly to tighten or loosen the economy should inflation remain above or below its target. Wall Street economists expect two cuts of 25 basis points before the year's end. In the baseline case, where the Fed is not assuming a recession, quarterly rate cuts each quarter are plausible, leading to the terminal target Fed funds range of 2.75-3.00% (versus 4.25-4.50% as of late August).

AI trade cooling off? Recent headlines suggest that the pace of cutting-edge AI gains may be flattening. Meta has delayed Llama 4, and OpenAI’s GPT-5 fell short of its hype, with AI executives pointing to investors growing overly excited about the technology’s growth trajectory. An MIT study finds 95% of custom generative-AI pilots fail, reflecting widespread corporate skepticism toward bespoke tools. In other words, “only 5% of customer enterprise AI tools reach production”. Executives still fear data leaks, hallucinations, and misaligned workflows, so most firms rely on off-the-shelf products. The perceived slowdown makes AI feel less like a moving target, which could drive businesses to embed AI into everyday operations. This is likely to remain a multi-decade transformation for enterprise leadership teams. Not accounted in the MIT survey is a “shadow economy”, a trend where workers use chatbots and other AI tools on their own, often without the blessing of their companies. This is what currently drives AI’s transformative impact on the economy and across enterprise workflows, although not through official channels yet.

AI revenues reach $18.5 billion. Talking about the impact on the economy, The Information reports that in less than three years, OpenAI, Anthropic, Anysphere (Cursor), xAI, and 14 other AI startups have gone from almost no revenue to collectively generating more than $18.5 billion in annualized revenue. Leading the money trail are OpenAI at ~$12 billion (65% market share) and Anthropic at $4 billion (22%), dwarfing the rest of the pack, which is the remaining $2.5 billion with at least $50 million in annual revenue each. These native AI apps are among the world’s biggest cloud customers—OpenAI and Anthropic will burn over $10 billion in Capex this year for developing and running their AI models—boosting the performance of cloud providers like Amazon, Microsoft, Google, and Oracle. Interestingly, Microsoft is seeing an added benefit, likely generating $1 billion a year of revenue from selling models from OpenAI and other providers to its enterprise customers, while its own productivity apps, like Microsoft Copilot, which have struggled through launch, are still coming of age.

AI push drives enterprises to unite HR and IT departments under one roof. An interesting survey, conducted by Nexthink, finds 64% of senior IT leaders at big companies expect their HR and IT departments to merge within five years as AI rewrites workflows. Several companies like Moderna, Covisian, and Bunq have already placed both functions under a single executive overseeing people and technology. Moderna’s chief people and digital technology officer now runs 5,000 employees, company-wide AI training, and an OpenAI partnership. A call center operator, Covisian, merged its 27,000-person HR and IT teams in April 2023, saying that this lets one leader tweak HR or tech processes instantly, driving better employee and customer efficiency. Bunq claims that merged departments remove turf wars and speed AI rollout, aiming for 90% operational automation by 2025 without layoffs, albeit highlighting that a clear trade-off between rapid execution and depth of skill still exists.

Is Cognigy worth 26x revenue to NICE? On the UC/CC news front, one could not help to overlook NICE’s acquisition of Cognigy, and the price it paid for it ($955 million), which in our estimation represents ~26x multiple of 2025 revenue and what is likely an unprofitable operation. That’s a departure from NICE’s typical 2-4x range for its prior bolt-ons (recent acquisition of LiveVox at $350 million, for instance, registering at ~2.6x). Notwithstanding the premium it paid, the move is justified, giving NICE access to bespoke AI technology and GenAI wrappers geared for CX workloads. It is reminiscent of the bold move NICE made when it entered the CCaaS market via the acquisition of inContact in 2016 (ironically, the same year when Cognigy was born) for which it dished out $940 million, or ~4.0x revenue multiple. With this acquisition, NICE has become an instantaneous Conversational AI and Generative AI CX leader, owning its technology stack and customer mindshare at the top of the enterprise pyramid (i.e., 500 concurrent agents and above), also paving the way for its two main competitors, Genesys and Five9, as well as potentially others in the down-market, to make similar moves. The remaining group of sizeable (by revenue standard) AI players that are tied to CX workloads comprises mainly of 3 players: Kore AI, Omelia, and Amelia, which we suspect NICE tried to buy in the past but opted against it. It will be interesting to see what happens to them, especially when NICE was the primary revenue source for some of them (e.g., Omelia). Other vendors in Gartner’s Conversational AI Magic Quadrant (comprising 22 vendors, no less) will likely grow in strategic importance as well. 

While inContact has proven to be a highly successful acquisition under prior leadership, the new CEO at NICE, eager to make a big move, will need to prove he can execute this strategy while keeping the Cognigy team in place. After the acquisition news hit the tape, the stock, after a brief knee-jerk cheer from investors, has withered, trading now at a 3.1x trailing price-to-sales multiple and 9.6x EV-to-EBITDA. The earnings in the most recent quarter ending in June, while delivering a beat (of $0.02 on EPS and $14.5 million on revenue), fell short of expectations on the full-year revenue outlook. The company reaffirmed its 2025 revenue to be $2,918-$2,938 million (excludes Cognigy, which is expected to close by the end of this year), representing a lackluster 7% YoY growth at the midpoint. The company maintained its growth outlook for cloud revenue of 12% for the full year, consistent with the prior quarter’s, which investors took to imply a lack of acceleration in growth coming on the heels of cloud revenue growth of 24% in 2024. After its fourth-quarter print in February, the stock took a nosedive and remained largely unchanged, coming down ~22% from its highs.

Verint Calabrio merger. Another standout deal, architected by an enterprise software PE, Thoma Bravo, has surfaced this quarter. After much speculation, Verint is going private at a $2 billion valuation, representing an EV/sales multiple of 2.2x and EV/EBITDA of 16x on a TTM basis (2x and 7.6x, respectively, based on forward numbers). The PE firm’s plan is to merge it with Calabrio, a WFO/WFM company it acquired for $1 billion in 2021. After 4 years under Thoma Bravo’s ownership, the entire executive leadership team has been replaced several times—the company has a new head of product, head of marketing, head of HR, new CRO, new head of services and customer success, and has cycled through at least 3 CFOs. All that says to us they are still figuring the growth out, i.e., what the proper sales motion is, along with how to slow down the churn (from Cisco enterprise accounts and the like, and faced by “full stack” players like NICE and Genesys). The sector has been under pressure from GenAI, drastically changing customer conversations. As for the Verint part of it, beyond obvious cost synergies, which makes this deal even more attractive, Thoma gets a hold of a strong enterprise customer base on the one hand and conversation intelligence and various bots, getting them into analytics and real-time agent assist, on the other. Marrying that conversational intelligence and WFM will be critical to their combined success, helping them fend off the competition (NICE has made its move to address exactly that, and we suspect Genesys will be next). Given that Verint’s open CCaaS approach has worked so far (i.e., staying highly relevant in the food chain without owning a CCaaS platform), we see the merged entity strengthening its move to AI-based digital interactions (versus moving backwards to owning a CCaaS stack) as a key component of its long-term strategy. Given that the market is currently spending at least 10x more on human agents vs. virtual agents, this move lends itself well to more accurate and optimal allocation of agent resources as the most prized possession in the CX WFM stack (i.e., accurately forecasting and scheduling agents based on agent performance and not just agent skillset). Yet over time, it is not inconceivable to think that an acquisition of a CCaaS vendor could help them close the gap in the mid and lower ends of the market, where one throat-to-choke systems are more commonplace. With several pure-play WFO vendors being taken out recently (Playvox, Cogito, Virtual Observer/CSI, Verint/Monet), we would expect a strong follow-through reaction from the remaining market leaders, mainly Genesys and Cisco. To wrap up, Genesys’ recent $1.5B raise likely leaves them out of the IPO race in the intermediate future, though should enable them with ample firepower to make decisive M&A moves to close the gap (particularly with key corporate investors like Salesforce and ServiceNow as backers, though one can’t help but wonder if all three are chasing the same AI targets).

Five9’s stellar quarter marred by CEO retirement. Five9 took out all the stops in the quarter, which was marred by executive rotation, delivering solid upside to estimates and guiding revenue growth at 10-15% through 2027 (with potential upside from AI). In the second quarter, Five9 posted 12% top-line growth and 16% growth in subscription revenue (representing 81% of revenue). Adjusted gross margins were 63% with a medium-term guidance of 66%-68% based on margin expansion from subscriptions. Adjusted EBITDA at 24% margin delivered a record performance and was raised to 25-30% over the mid-term. Enterprise AI revenue showed 42% growth YoY with bookings tripling YoY in the quarter, while total ACV bookings were the highest in 2 years. With such a stellar backdrop, CEO Mike Burkland has replaced his CRO, CAO, and Legal Officer, and announced his own retirement, which comes on the heels of thirteen-year CFO Barry Zwarenstein’s retirement last quarter. With this many changes, investors are likely to take a wait-and-see approach until the new CEO steps in and they can gain confidence in the new management team. With all these moves, the stock is off 25%+ over the last six months. Moreover, Five9 will likely need to bulk up in GenAI to catch up with NICE, as both leaders and former CEOs of Inference and Acqueon (CAI/IVA and outbound customer engagement/predictive analytics vendors it acquired back in 2020 and 2024, respectively) have left the company.

Twilio’s falling from grace is all due to gross margins. We wrote extensively in the last quarter newsletter about Twilio’s restructuring efforts (and activist-influenced Board), making the right moves to right-size the business for profitability, cash flow optimization, and churn mitigation. This earned them 3x return from the trough low-50s/share last summer to a peak of near $150s in February. The shares are off roughly 30% from peak, mainly due to margin pressure—a recurring issue since 2018, when telecom carriers first introduced a surcharge (fee) on 10DLC number routes (for A2P messaging). That surcharge was supposed to be passed through entirely to consumers, investors were led to believe, but it hasn’t quite materialized in such a way. In addition, the carriers kept raising their fee (A2P messaging price) every year, which it is now a significant portion of Twilio’s messaging revenue. That, plus a greater mix of international messaging (lower margin), continued challenges in Segment (i.e., lack of growth and churn in Segment CDP), and a second-half reduction in FCF, has sent shares to lower levels after the 2Q print. Non-GAAP gross margin for Comms BU edged 60 bps lower to 49.2% in 2Q from 49.8% in 1Q and 51.6% for the full year of 2024, with likely pressure going forward. Another trend to pay attention to is the growth of new authentication methods (such as passkeys and biometrics), putting negative pressure on conventional 2FA messaging, which is just starting to grow in adoption.

UCaaS bland. Not much striking news out of the UCaaS world, with key public vendors showing unimpressive top-line growth of ~5% at one side of the spectrum and flat to negative on the other. For privates, the story is slightly different. Dialpad, for example, surpassed the $300 million mark in ARR late last year, claiming a growth rate of over 50% in 2024, and has continued a 30% CAGR for several years now. Most of this success is due to its AI-powered CCaaS strategy and cloud-to-cloud migrations on the UCaaS side, which it executed flawlessly. However, with heavy investments in sales and AI product development (signaling at least $50 million dedicated to AI R&D since 2023), the company likely remains shy of break-even, as it pushed its IPO plan further into 2026. From other privates, we heard that Intermedia is growing on the UCaaS side at an above-20% growth rate. 

The Business Section

Verint Acquired: Thoma Bravo has acquired Verint in a $2B valuation. This is likely a great exit for CEO/Founder Dan Bodner, but not so great for its customers. Verint is two companies: an old-school WFO/WEM player and a rising superstar in AI automation. Bodner has successfully navigated several pivots, but Bravo likely wants the old company. 

There was a time when Verint and NiCE were toe-to-toe competitors. NiCE pivoted from WFO/WEM in 2016 when it acquired inContact. Verint went CC neutral, which has been very lucrative, but has a small problem: most CCaaS providers (and Verint partners), have their own native WFO/WEM. This is why Verint’s pivot to AI Automation is so brilliant. It’s timely, and the old business fuels the cash, customers, and customer data for success. 

Bravo seems excited about the WFO/WEM business. This is a PE firm, and it’s far more likely interested in profits than growth. Bravo will cut costs by consolidating its WFO/WEM acquisitions (Verint now and Calabrio and Medalia acquired in 2021), and profit from the fact that these brands will deliver revenue for years, without the need for R&D. We are seeing something similar play out at Avaya. 

WFO/WEM remains critical, and not all CCaaS providers have developed or acquired a solution. I would rather have seen any or all of these firms go to an established CCaaS provider, like Cisco, where the strategy would include growth. 

While Bravo positions itself as an innovation machine, the innovation pressure will be on the CCaaS sector to stop positioning their native WEMs as WEM-Light. That won’t be easy: Every ounce of CX R&D is going into ways to eliminate agents — not schedule them. 

Not Bravo’s Dollars: Thoma Bravo secured permission from its backers to allocate portions of buyout deals to other investment vehicles. This move aligns with a growing trend of allowing affluent individuals to invest in PE plays, and it is expected to accelerate following Trump’s executive order opening the $9T retirement market to alternative investments. Thoma Bravo has implemented these terms across its latest European fund, a new $24B global fund, and its latest mid-market fund, all of which closed this year.

Yorktel + Kinly: Yorktel announced it will merge with Amsterdam-based Kinly to create a combined AV and UC integration firm serving more than 2,500 clients across 27 locations, supported by 1,600 employees and over 900 specialist certifications. The companies say the move will accelerate global growth and expand their managed services and systems integration portfolios, pending regulatory approval. The deal also reflects a broader trend in the AV market toward consolidation (or collapse, depending on your perspective) as larger players buy smaller ones to expand capabilities and geographic reach in an increasingly competitive landscape. (See our interview with Yorktel CEO Ken Scaturro.) —Danto

Max Cap: Capacity has secured over $90M in funding, bringing its total to $155M. The company has acquired 12 customer experience firms since 2023, including Call Criteria, Verbio, YouCanBookMe, CereProc, SmartAction, Envision, Lucy AI, Linc, LumenVox, Denim Social, and Textel. These acquisitions span QM, voice agents, scheduling, TTS, virtual agents, coaching, knowledge management, conversational commerce, speech processing, social campaigns, and SMS engagement. The CX sector is overcrowded. Capacity is doing a rollup and grabbing non-distressed assets.

Anthropic $10B: Anthropic is close to securing $13B (Series F) of new funding with the round led by Iconiq Capital, which values the company around $83B (up from $61.5B earlier this year). Anthropic’s annual recurring revenue has surged to $5B, with projections of reaching $9B by year-end. The funding, including investments from Qatar Investment Authority, GIC, and Amazon, will fuel Anthropic’s competition. OpenAI and xAI have both raised billions. Despite previous concerns about taking money from authoritarian countries, Anthropic CEO Dario Amodei acknowledged the necessity of raising funds from Middle East countries.

Pylon 31M: San Francisco-based Pylon has raised $31M, co-led by Andreessen Horowitz and Bain Capital Ventures. Pylon differentiates itself by targeting B2B businesses, handling customer interactions across channels like Slack, WhatsApp, Discord, and email. The 50-person company serves over 760 business customers. Pylon also aims to boost sales by alerting salespeople to potential upsell opportunities or product managers about new feature needs. Pylon was founded in 2022 and expects to reach $10M in annual subscription revenue over the next 12 months. The company aims to replace systems like Zendesk and Salesforce’s Service Cloud.

PS I Love UJET: Onix, Google Cloud Partner of the Year, has acquired the professional services (PS) unit of UJET, aiming to enhance its AI-driven customer engagement solutions. This acquisition integrates UJET’s Google Cloud-based platform and its professional services team, which has expanded significantly in the last two years due to its Google Cloud partnership. Onix will leverage this team’s expertise in deploying Google’s Customer Engagement AI Suite globally, strengthening its position as an implementation partner for UJET’s CCaaS platform. The move allows UJET to focus on its core software innovation while ensuring customer continuity, as the same professional services team will remain. No terms were disclosed. 

Leadership Changes

Colby Nish is the new CRO at Dialpad. Previously led sales at Zoom and Cisco. Ulrik Hvid is no longer the VP of Global Sales & Marketing at EPOS. Mike McCarthy has left Huddly and is now VP of North America sales with Kramer. Laura Bassett (known from NiCE and Sharpen) joined Joulica as CMO. UJET named Ian Peters-Campbell (WEX) its CTO and Kristin King (Windstream) its CRO. 

On the analyst/AR side: Jared Oliva has joined the AI team at Zoom. Josh Blalock has moved from Jabra to Shure. Dan Root left Barco and joined Netspeek as VP of Sales and Partnerships.

Meet Us

Dave Michels is the founder and chief protagonist of TalkingPointz and the editor-in-chief of the Insider Report. [email protected]. David Danto is AV Aficionado at TalkingPointz. [email protected]. 

Goodreads

1. Microsoft admits it ’cannot guarantee’ data sovereignty — Under oath in the French Senate, exec says it would be compelled to pass local customer info to US admin.
2. AI Is a Mass-Delusion Event — Three years in, one of AI’s enduring impacts is to make people feel like they’re losing it. 
3. Companies Are Pouring Billions Into A.I. It Has Yet to Pay Off. — Corporate spending on artificial intelligence is surging as executives bank on major efficiency gains. So far, they report little effect on the bottom line.
4. SoftBank to invest $2 billion in Intel — Cash from public and private sources flows into Intel.
5. Yorktel to Merge with Kinly, Expanding Global AV and Managed Services Reach — M&A creating another survivor in the consolidation of AV companies.
6. Microsoft’s August security updates break Windows recovery, reset — More untested patches that cause damage from MS.
7. F-35 pilot held 50-minute airborne conference call with engineers before crashing fighter jet Ever have the desire to abort a customer service interaction? 
8. Why We’re Leaving the Office Earlier but Still Showing Up on Time — We may be RTO, but we don’t stay long.
9. When Free Isn’t Free: The Hidden Dangers of WhatsApp in the Workplace — A decent helping of FUD.
10. Google announced the next step in its nuclear energy [for AI] plans — The tech giant inked a deal to support a next-generation nuclear energy reactor being built in the shadow of the Manhattan Project.
11. A Better Way to Think About AI — Artificial intelligence is ready to collaborate. Why fixate on automation?
12. All the announcements from Google’s Pixel 10/Watch 4 event — The heat rises in mobile devices’ use of AI. 
13. Bondholders of NYC’s Worldwide Plaza Face $488 Million in Losses — Don’t buy into the recent RTO hype.
14. YouTube’s Sneaky AI ’Experiment’ — The video platform is quietly using AI to “improve clarity” in uploaded content. Why?
15. Amazon quietly blocks AI bots from Meta, Google, Huawei and more — Amazon is escalating efforts to keep artificial intelligence companies from scraping its e-commerce data.
16. Google Risks Making Android Users Furious Over App Sideloading Policy Change — As we predicted, Google is taking steps to harden its version of Android in the face of MDEP.
17. Bankers learn of firings via premature email to hand back their laptops — Australian bank ANZ apologizes and offers counseling to more than 100 recent former employees.
18. Ursula von der Leyen’s plane hit by suspected Russian GPS interference — We may be at the beginning of the end of GPS. European Commission president forced to land in Bulgaria using paper maps.

Other Recent Stuff

Effective in 2025, we moved Insider Lite to Substack. You can view the most recent edition here. This is free content, and Substack makes things easier for everyone, but you will need to subscribe separately here. Insider Lites typically publish 1-3x a month.

• How NiCE Won a 40,000-Agent Contact Center Deal with DWP — Genesys fumbled with MultiCloud.
• On the Road Again — Video: Dave and David discuss September event plans.
• Something New, Something Blew — The long, problematic tail of upgrades.
• The CEO Is Now an AI Agent — How far do we take trust in AI?
• 8x8, The Power of Dave, and Lions — On 8x8’s new AI branding.
• Two Factor Follies — Let’s just admit we no longer carry phones, they carry us.
• TalkingPointz Chatz — what makes a good headset — Video
• AI Isn’t Coming for Your Job — It’s Coming for Google’s Job — The power of search is no longer pointing you to the answers, it’s simply giving them to you.
• Welcome to Taco Bell AV — Consolidation continues.
• How to Retire in Business Without Actually Leaving — The saga of the boomerang CEO.
• The Triple Peak Workday Needs to Go — Can AI shorten the workday? 
• Michels’ Thoughts on Genesys and Five9 News
• Michels on TR Regarding Cognigy Acquisition 
• Reach Out and Text Someone — Why nobody answers the phone anymore.
• First or Stickiest: Remember the Diamond Rio? — Is Zoom losing its edge in video?
• Who Is Watching Whom? — Thoughts on ownership of your biometric/digital data. 
• An AV Rose by Any Other Name — Product naming in AV.
• Prem Is Back, Baby
• Zoom Delivered What Microsoft Inspired, and a discussion about it
• A Closer Look at the Sennheiser TC Bar S — Product Review and Video
• The TalkingPointz Whitepaper on Video Interoperability is now publicly available

Recent Insider Reports (only new Insider Reports are behind the paywall) 

1. Insider July 2024
2. Insider August 2024
3. Insider September 2024
4. Insider October 2024
5. Insider November 2024
6. Insider December 2024
7. Insider January 2025
8. Insider February 2025
9. Insider March 2025
10. Insider April 2025
11. Insider May 2025
12. Insider June 2025
13. Insider July 2025

TalkingPointz Insider Reports are available through a subscription service at TalkingPointz.com.

Any Questions?

Your paid enterprise subscription entitles you to a conversation on any of the topics covered above. 

About

Insider Reports offer timely, curated, and opinionated news. The Insider Report is published in the first week of every month. Enterprise subscriptions also include TalkingPointz research notes. These reports are geared toward the industry itself, though some enterprises subscribe, too. Core subscribers are vendors, providers, consultants, and (financial and industry) analysts. TalkingPointz offers enterprise vendor, channel, consultant, and personal subscriptions.