Nobody’s Bingo Card Survived the Summer

Seven weeks to CASA26 — and the industry has already rewritten itself since spring.

Three earnings reports in seventy-two hours, and the most interesting thing about them wasn’t the numbers.

Twilio reported revenue up 22% year-over-year — accelerating from an already three-year-high quarter — with net retention at 116% and guidance above consensus. The stock jumped 16%. NICE also beat, also raised its full-year EPS guidance, and grew AI ARR 52% to $362 million. The stock fell 7%. Five9 beat as well, with subscription revenue up 14% and AI revenue up 78%, and raised its full-year AI growth outlook from 40% to at least 60%.

Three beats. Three completely different verdicts. The market has stopped paying for a good quarter. It is paying for the rate at which AI becomes the revenue.

Which is a strangely precise restatement of something a Dutchman said about football.

Last month in New Jersey, Spain won the World Cup playing the total football Johan Cruyff exported from Amsterdam to the world. Argentina defended deep, played rough, and finished the final with two shots, neither on target. Spain moved, combined, saw the whole pitch, and won. Cruyff’s old line: don’t run to where the ball is. Run to where it’s going.

That is what got priced this week. Not the best quarter. The company that had already moved.

Seven weeks from now, the people trying to work out where the ball is going for our industry will be in Cruyff’s hometown, at the Tobacco Theater in Amsterdam, for CASA26.

And here’s the thing about this year: nobody’s bingo card survived the summer. If you filled one in last January, be honest about what wasn’t on it.

1. The C-suite carousel is spinning everywhere

Ericsson changes CEO after nine years — Per Narvinger takes over October 1. Weeks later, Vonage elevates its API chief, Christophe Van de Weyer, to CEO and Ericsson’s Executive Team: the clearest signal yet that network programmability is the growth engine, not the side bet. Sinch is in transition, with Jonas Dahlberg as acting CEO.

And this is not just a CPaaS story — the contact center world is doing the same thing, faster. Five9 is the clearest case in the industry: a new CEO at the start of the year, and by June a rebuilt executive bench — new CTO, new chief sales officer, new head of transformation and strategy — essentially a new leadership team inside six months, assembled to run an AI-first playbook. Nobody had “Five9 replaces its entire C-suite” on their January card. But that’s the point: boards across CPaaS, CCaaS, and infrastructure have concluded that the team that built the last era isn’t automatically the team for this one.

And the rebuild is already showing up in the results you just read — two months from a new bench to a Fortune 100 financial services win worth around $100 million in total contract value.

2. The contact center land grab

Salesforce launched Agentforce Contact Center at Enterprise Connect — a native CCaaS, generally available, from the company that spent a decade partnering with the contact center industry instead. Zoom now positions its platform as an orchestration layer for human and AI work. The pattern: everyone wants to own the communications layer, because the communications layer is where the intelligence meets the conversation. CX without the pipe is a dashboard; the pipe without intelligence is a commodity. The race is to own both.

3. The per-seat model is being dismantled in public

Here’s the shift that gets least airtime and matters most, because it changes how the money actually arrives. At 8×8, usage-based revenue went from 14% of the business in fiscal 2025 to 23% in fiscal 2026 — CPaaS APIs, digital channels, AI. CEO Sam Wilson has been unusually blunt about why: CFOs are tired of paying for unused seats, the walls between UC, contact center, CPaaS and CRM are coming down, and what customers buy is changing along with how they pay for it. Per-seat pricing made sense when the seat was the unit of work. An AI agent is not a seat.

RingCentral is telling the same story from the other end. Paid AI products now sit at 13% of ARR, doubled year-over-year, with more than 16,000 AI Receptionist customers — up 400% — carrying higher ARPU and net retention above 100%. The stock rose 25% on it. But run the subtraction: if AI ARR doubled to 13% while total revenue grew 5.9%, the other 87% is close to flat. The AI line isn’t being added to a growing base. It’s replacing a stalling one.

And there’s a number underneath all this that the industry is not discussing nearly enough. 8×8’s gross margin fell from 66% to 61% year-over-year. Consumption revenue and AI delivery do not carry seat-license economics. Everyone is talking about AI as a revenue opportunity. Almost nobody is talking about what it does to the margin structure you built the last decade on. Both questions get answered in the same P&L, and only one of them is on the conference agenda.

4. The developer was the moat. The agent doesn’t read the docs.

Vonage just integrated its APIs into AWS’s agentic development environment, Kiro — so AI coding agents can wire up communications without a developer ever opening the documentation. Tools like Cognition’s Devin push the same way: integration cycles collapsing from quarters to days.

Sit with what that does to a go-to-market. CPaaS was built on developer love — win the developer, get pulled into the enterprise from below. Every evangelism programme, every hackathon, every beautifully written doc set was an investment in that moat. An agent has no loyalty, attends no conferences, and cannot be evangelised. It selects on price, latency, and whether your API is machine-legible. The route to market that built this industry is being automated away by the same wave everyone is selling into.

Which explains something that otherwise looks like fashion: why the Network API world went all-in on MCP almost overnight. CAMARA published a white paper in January on exposing network capabilities to AI systems through the Model Context Protocol, aligned with GSMA Open Gateway and TM Forum. For five years this industry marketed Network APIs to developers who largely didn’t turn up — fragmented across operators, inconsistent, slow to procure, awkward to build on. As a developer product, they were mediocre. As an agent product, they are excellent.

Because the thing an agent cannot do is verify. It infers. It guesses where the user is, whether the number really belongs to them, whether the line will carry video. The network doesn’t guess — it knows. Verified identity, real-time location, network state, regulatory standing: precisely the facts an agent cannot reason its way to and has to be told. That is the collateral telcos have been sitting on while the industry argued about API counts. MCP is just the connector that finally makes it legible. The shop window was built for a customer who never came. The customer that did come doesn’t have eyes.

5. Capital is running at two speeds

Paris-based Gradium hit $100 million in seed funding seven months after launch, with NVIDIA joining the round. ElevenLabs raised at an $11 billion valuation in February — and is reportedly in talks for a secondary at roughly double that. NVIDIA put $1 billion into Nokia last October; the stock has more than doubled since as AI-RAN went from concept to a ten-operator customer list.

On the public side, I’ve already given away the punchline. But there is a second number buried in these results worth sitting with: AI now sits at 13% of ARR at RingCentral and roughly 15% of the cloud and subscription base at both NICE and Five9. Three companies, arriving there independently, landing in the same band. Eighteen months ago that line barely existed. The pilots converted — and the market has started pricing companies on how fast.

Sinch, meanwhile, is running Twilio’s play two years behind: restructuring done, the CFO who executed it now in the CEO seat, shares up roughly a third over the past year. The revealing number: Twilio and Sinch are less than 2x apart on revenue and more than 10x apart on market value. That gap is the market’s price for one thing — proving you’re fit for the AI era. Capital, private and public, has already decided where the ball is going: toward whoever owns the intelligence in the conversation.

6. AI is moving into the network itself

At MWC in March, Deutsche Telekom premiered the Magenta AI Call Assistant — real-time translation, summaries, and contextual help embedded directly in the network, activated mid-call with “Hey Magenta.” No app, no device, world-first network-based AI in the call itself, rolling out in Germany this year with up to 50 languages planned. The voice AI comes from ElevenLabs — yes, the same one the capital markets are chasing — and the network-level integration was carried out by Radisys, arguably the earliest to make in-network AI voice work in production. Radisys is running the same play with Rakuten Mobile, which is embedding its Engage Digital Platform directly into the core to spin up AI services from a single platform. Mavenir is on the same path — embedding AI into the mobile core with Turkcell, building an AI innovation hub with Telefónica. Multiple vendors, multiple operators, one direction: after twenty years of value migrating out of the network into over-the-top layers, intelligence is pulling it back in — new services, born in the core, monetized by the operator. That’s not a roadmap slide anymore. It’s shipping.

The planning horizon has collapsed

There’s a pattern underneath the surprises, and it’s the real story. You can’t extrapolate your way to 2028 from last year’s survey data. Understanding where the ball is going requires understanding the forces at work — the capital flows, the technology inflections, the strategic repositioning happening in real time — and no single company sees all of it alone.

That isn’t a claim from a desk. Over the past six months we’ve been on the road with our partners at Sandbox — more than a hundred conversations with operators, cloud communications players and innovators across the ecosystem. The consistent finding: the technology questions are largely settled. What people want to know now is who to build with, on what terms, and with whose capital behind it. Those conversations are where the CASA26 program came from, and what we heard goes on the stage in September rather than staying in a deck.

That’s exactly what CASA26 is built for.

Four days to re-strategize for ’27 and ’28

From September 20 to 23, around 150 senior leaders from telcos, CPaaS and CCaaS platforms, AI innovators, and the capital behind them come together in Amsterdam — invite-only, practitioner-level, no vendor theatre. Sunday opens the week, two days of main program follow, and Wednesday closes with the GSMA telco day. All of it structured around the questions that actually matter: Intelligent Engagement, Sovereign AI & Trusted Infrastructure, Network APIs & Telecom Platforms, and Innovation, Ecosystems & Real Outcomes — and the main program closes on ecosystem and capital, where the next layer of value gets built and where investment actually flows, with our partners at Sandbox.

This is not a look back at the year. It’s the room where the industry compares notes on the storm — and works out, together, where the ball is going before the next cycle starts.

Because if the last six months are any guide, the seven weeks between now and September will bring surprises nobody has on their card today. The people who navigate them best won’t be the ones scanning the landscape alone.

They’ll be the ones in the room.

CASA26 — September 20–23, Tobacco Theater, Amsterdam. By invitation.