At its annual user conference, Xperience 2026, Genesys unveiled an ambitious product vision that marks the beginning of its latest act, staking out a new market it intends to pursue.

After CTI, routing, omnichannel, and cloud contact centers, this fifth act harnesses agentic AI to move Genesys beyond interactions into resolution workflows.

The timing is notable. Genesys Cloud continues to grow at more than twice the market rate. Yet Genesys is making a significant bet on what comes next, expanding its ambition from managing customer interactions to orchestrating customer requests and desired outcomes.

The company has the financial strength to make that bet. Genesys Cloud ARR reached nearly $2.9B, up 30%; AI ARR hit $400M, up 60%; and first-half revenue grew 20%. Genesys also reported new financial metrics, including a 30% GAAP operating margin and more than $500M in free cash flow for the same period.

At the event, Genesys introduced several new products and presented a restructured product portfolio to execute on this vision.

Genesys Navigator is the first new product. It is a customer-facing agent that acts as a business front door, pulling together the relevant context, understanding customer intent, and routing the inquiry to the right “actor” to handle it, human or AI.

Genesys has purposefully divided the roles between Navigator and actors. Navigator uses a general-purpose model to understand why the customer is reaching out. Actors such as Genesys Agentic Virtual Agents (AVA) are focused on resolving the request. AVA uses Scaled Cognition’s new APT-2 Large Language Model (LAM), which Genesys says reduces hallucinations. It now also benefits from Pinkfish declarative workflows and MCP integrations that significantly broaden the scope of actions it can take.

Genesys Orchestrator is the second, and arguably most significant, new product announced. Orchestrator is context-aware and manages multi-step journeys toward desired outcomes. Steps are delegated to an actor: a human interaction or work item, outbound outreach, or any of the automations Genesys supports, including Guides, Workflows (Pinkfish), first-party agents such as AVA, or third-party agents via A2A.

Orchestrator is fundamentally a decisioning engine, not just a workflow and journey orchestration tool. It uses a newly introduced, proprietary specialized model, Kaleidoscope, which Genesys explained was created to make sense of policies, context, intents, and events to determine the desired outcome and the next best delegation towards that goal, strictly within the policies, guardrails, and scope defined by the enterprise. Genesys resolutely positions Orchestrator as autonomous. The ambition to rely entirely on autonomous decisions in a world of customer service where predictable and repeatable processes have traditionally reigned supreme is audacious.

Navigator is expected to reach general availability in Q4 FY27, followed by Orchestrator in Q1 FY28.

This ambitious undertaking rests on two essential capabilities: context and control. Genesys AI Control Plane centralizes governance and provides observability for these autonomous steps, while Genesys Contextual Intelligence maintains persistent context by connecting interaction signals, customer identity, history, business events, and journey data.

It felt like Genesys wants to break away from the traditional CCaaS market, almost as if it were willing to burn bridges with it.

Why make such a sweeping move now, when Genesys Cloud continues to perform so strongly?

My read is that Genesys is anticipating slower CCaaS market growth and intensifying competition from hyperscalers, CRM providers, and Conversational AI specialists, and is moving on the offensive to expand its addressable market.

What Genesys unveiled at Xperience 2026 is a bold move. Following NiCE last year, Genesys is now setting its sights on resolution workflows, an increasingly contested market with CRM and other technology providers.

CCaaS players are all striving to redefine and expand their market. Many have adopted CX platform terminology without clearly defining how their newly minted CX platform differs from CCaaS. Genesys has taken a serious stab at answering that question.

At its Investor Forum in May, Atlassian revealed that its customer service business, delivered through its Service Collection bundle, had surpassed $1B in ARR and was growing 30% year over year.

It has been adopted by 65,000+ businesses using Jira for internal use cases, such as employee help desks and engineering work management, as well as external ones for customer service and support. Atlassian is seeing particularly strong traction with product- and technology-centric businesses, where tightly connecting customer support with product and engineering is central to the customer experience.

What I hadn’t appreciated is that more than 75% of Jira Service Management users are non-developers. That is notable because Jira has long been perceived as a service desk for developers and product teams.

With 150,000+ customers using Jira Cloud, Atlassian also has a substantial installed base to cross-sell Service Collection into, giving it a long runway for growth.

Atlassian introduced its Customer Service Management solution in April 2025. Built on Jira, it includes a customer-facing help center with a built-in knowledge base. Customers can submit service requests via email, chat, embedded product widgets, or forms on a support site, which Atlassian calls omnichannel intake.

Because Customer Service Management runs on the same platform as Jira Service Management, support teams can seamlessly escalate customer issues to product and engineering teams. Issues can also be connected to internal conversations across Slack, Microsoft Teams, and Zoom, bringing additional context into the resolution workflow.

This fits into a broader shift in Atlassian’s strategy. The company built its remarkable product-led growth around standalone products that were easy to adopt. In 2025, it pivoted toward Collections as a new go-to-market model. Collections bundle complementary apps and AI agents around a business function, using a common platform and shared Teamwork Graph to connect teams and provide richer context.

In October 2025, Atlassian introduced Service Collection, combining employee- and customer-facing service management, asset management, and Rovo to build AI agents that automate support workflows and assist humans. AI-assisted resolutions are priced at $1 each.

At Team ‘26, its annual user conference held in May, the company expanded that vision with several significant announcements:

• It announced support for messaging channels such as WhatsApp and SMS, as well as live chat agents with handoff to human agents.

• For voice, which has been supported via Amazon Connect, Atlassian announced native voice support with an AI agent front end and escalation to human agents.

• The company also unveiled Optimize, its solution for self-improving support. It will focus first on identifying knowledge gaps and updates and generating articles from human-resolved conversations, with Atlassian envisioning a future of proactive service that spots emerging issues before customers file a ticket.

Separately, Twilio highlighted Atlassian as a major customer on its 2Q26 earnings call, indicating that Atlassian will use its Flex SDK and Super Network to build and deliver omnichannel support within its customer service management app.

Atlassian is quietly assembling a serious customer service and contact center challenger. With $1B+ in Service Collection ARR, a large Jira Cloud installed base, and rapidly expanding omnichannel and AI capabilities, it has the distribution, platform, and capabilities to become a force the market needs to reckon with.

Earlier this month, I attended the KeyBanc Technology Leadership Forum.

What I love about this event is that it provides a broad perspective across the entire technology landscape.

One topic dominated this year’s agenda: data centers.

The scale of the buildout is extraordinary. The 14 largest publicly traded data-center operators are expected to spend nearly $750B in CapEx in 2026. But the buildout is increasingly constrained by five forces: the 5 Ps:

• Planet: climate and environmental constraints

• Policies: regulation and permitting

• Power: electricity and grid availability

• People: availability of skilled construction workers

• Pushback: growing resistance from local communities

What I found even more interesting was that these constraints are prompting a rethink of not just how quickly we can build computing capacity, but how much we actually need.

Furthermore, token economics, Chinese models, and open-weight models are forcing a closer look at the economics of inference and the architecture required to run AI.

I left the conference with five ideas to explore:

1. Deployed GPU capacity may be underutilized for inference.

2. Memory is becoming the constraint, forcing a rethink of how AI systems are architected.

3. Networks are becoming the critical connecting layer between storage and compute.

4. Frontier models aren’t always necessary. SLMs and open-weight models may often deliver the best economics and performance for specific tasks.

5. Today, we hear about agentic pilots using reasoning every step of the way, but many enterprise processes need repeatability more than reasoning, pointing toward hybrid workflows that combine deterministic automation with agentic reasoning where needed.

We are still figuring out what AI actually needs to scale. AI infrastructure constraints are forcing a rethink of the assumptions underlying the current AI scaling narrative.

Crescendo has introduced its Customer Experience Platform (CXP).

It is a major milestone for the company and, in many ways, a complete business pivot.

Crescendo is a relatively new entrant in CX, founded in late 2023 by industry veterans including Andy Lee, founder of Alorica; Anand Chandrasekaran (ex-Five9); Slava Zhakov (ex-Genesys); and Tod Famous (ex-Cisco, ex-Genesys).

Andy played a central role from the outset, particularly in shaping the vision of disrupting the BPO industry with a 100% outcome-based business model.

Crescendo initially took off with a concierge use case, combining human agents and AI to drive self-service improvement and adoption. In my first briefing with the company in 2024, Tod Famous explained that the outcome-based model drove the company to take ownership of knowledge management and build a closed-loop feedback mechanism with agents.

In October 2024, Crescendo acquired PartnerHero, a small BPO that brought its ARR to $50M. In April, the company announced it had crossed the $100M ARR mark.

What I learned in my briefing earlier this month is that the outcome-based model also pushed Crescendo to expand beyond self-service, develop a Forward Deployment Engineer (FDE)-based go-to-market model, and build a comprehensive CX stack. The objective: address disconnected technology stacks and drive continuous, proactive improvement.

Today’s announcement is a major one for the company and helps explain why Andy stepped into the CEO role last month.

It feels like Crescendo 2.0.

Crescendo Customer Experience Platform combines CCaaS, ticketing, workforce management, self-service, knowledge management, quality management, conversation intelligence, and Voice of the Customer (VoC) functionality, packaged as a set of specialized agents.

Two agents particularly stood out to me:

• Applied Insights works across Crescendo’s data estate to generate proactive recommendations.

• Optimization Agent translates those recommendations into system setup and configuration changes.

Together, they form the cornerstone of what Crescendo calls its governed self-improvement loop, designed to enable continuous improvement.

Crescendo has already started deploying its technology with customers, initially targeting the 50-500 seat market.

The company will continue serving its existing outcome-based customer base, leveraging Alorica’s talent pool.

In a crowded CX market, this is a bold move. It will be interesting to see how the strategy plays out.

TTEC announced last week that it is exploring strategic options for its TTEC Digital business.

The company intends to capitalize on the current valuation premium associated with AI, a business it has successfully grown. The decline in its BPO business this quarter and its debt level were also likely key factors behind the strategic review.

With revenues approaching $500 million, TTEC Digital is one of the few CX-specialized mid-size system integrators to have emerged from the transformation of the North American contact center channel over the past decade.

Its evolution illustrates how fundamentally the channel has changed.

In the pre-cloud era, VARs were an essential channel for implementing contact center solutions. Deployments involved substantial system integration, creating a lucrative services opportunity alongside product sales. Professional services could represent $2 for every dollar of product sold.

The cloud changed those economics. Vendors increasingly took large enterprise deployments directly, while cloud and more holistic solutions reduced the amount of integration required for deployment. Professional services attached to product sales fell to less than 50 cents per dollar, a decline that market growth could not offset.

The result was a fundamental reformatting of the channel. Many VARs were acquired or transitioned into Technology Advisors (TAs) served through Technology Solutions Distributors (TSDs). Those that remained evolved into specialized SIs focused on a technology ecosystem. A few exceptions, including Miratech, Servion, and of course TTEC Digital, have developed multi-vendor practices.

AWS’s entry into the contact center market accelerated another part of this evolution, driving Global System Integrators (GSIs) to establish contact center practices. Initially focused on deploying Amazon Connect, they quickly expanded to other platforms and into integrating contact centers with the broader enterprise fabric.

This created a new opportunity for SIs. Rather than simply deploying the contact center, the opportunity increasingly became connecting it to CRM and other enterprise applications, and implementing AI, now a top priority in CX.

TTEC Digital is a product of this transformation. It started with a traditional contact center deployment business focused on Avaya and Cisco. Today, it has become a multivendor CX, CRM, and AI business, with sizeable practices around Amazon Web Services (AWS), Cisco, Genesys, Google, Microsoft, NiCE, Pega, and Salesforce. It built these capabilities through several acquisitions, including Serendebyte, VoiceFoundry, and Avtex. It has also built a portfolio of solution assets around these platforms to complement products and accelerate deployments, while significantly expanding its managed services business, which now represents roughly two-thirds of its revenue.

TTEC Digital generated $104 million in revenue last quarter, down 8.5%. The decline reflects the fading of one-time on-premises license and product transactions and declining revenue from its two legacy practices.

It will be interesting to see how TTEC’s strategic review unfolds, and what it ultimately reveals about the future role and economics of specialized CX system integrators.

Five9’s Q2 results offer another valuable perspective on the CX market, while providing an important six-month checkpoint for new CEO Amit Mathradas as he reshapes the company’s strategy, priorities, and go-to-market.

Revenue reached $312M, beating estimates and growing 10% YoY, marking the third consecutive quarter of accelerating growth.

CCaaS revenue grew 7%, in line with the previous quarter, while AI revenue jumped 78% YoY to $39M, now representing 15% of total subscription revenue. Together, these two growth engines drove subscription revenue up 14% YoY.

In a market where investors are concerned about seat contraction, Five9 indicated that its concurrent seat count continues to grow at the same pace as CCaaS revenue.

The company also highlighted its push to accelerate the ramp of AI deals. It drove an expansion of professional services to support faster deployments, putting pressure on GAAP gross margin, which declined to 53.4%.

One win is worth highlighting: a $100M total contract value over five years with a Fortune 100 financial services provider. Sold through the Google Cloud Marketplace, the deal builds on Five9’s major investment in Google Cloud earlier this year.

Amit discussed the executive changes announced in late June in the context of the four priorities he outlined last quarter: building a performance-driven culture, optimizing operations, strengthening the core business, and winning in AI-powered customer experience:

• Niranjan Vijayaragavan joined as Chief Technology Officer to bring together Product Engineering, Product Management, AI Automation, and Architecture.

• Rob Hornish joined as Chief Sales Officer, bringing expertise in translating complex technology into measurable business outcomes.

• Sven Linsmaier became EVP of Transformation and Strategy, a newly created role responsible for enterprise strategy, corporate development, and the company’s highest-priority initiatives.

• These appointments followed the hiring of Jay Lee as Chief Marketing and Growth Officer in April, bringing an analytical background to the expanded CMO role and its mandate to chart the company’s growth path forward.

The diverse, non-CX backgrounds of these new executives reflect Amit’s conviction that the next phase of customer experience will require much more than contact center technologies, alongside a greater focus on business outcomes in Five9’s evolving go-to-market.

Amit also shared the early direction for Five9’s positioning: a voice-led enterprise CX platform focused on complex, regulated enterprises such as financial services, healthcare, and insurance. The company believes humans and AI need to come together to deliver what it calls “humantic CX,” with voice AI deployed as part of a broader CX platform rather than as a point solution.

In just a few months, Amit has reshaped the leadership team, accelerated the AI business, and begun repositioning the company around a broader vision of enterprise CX. The pace of change is notable, and the next few quarters should provide an important read on how this new strategy shapes Five9’s next chapter.

Things continue to accelerate in the Conversational AI market.

Omilia is raising $67M in a Series B. This comes six years after its Series A in May 2020. During that period, its live ARR has grown more than 10x to over $60M, representing roughly 45% CAGR.

The long-standing player with deep roots in voice solutions has operated somewhat under the radar, despite being featured in the latest Forrester Wave, Gartner Magic Quadrant, and IDC MarketScape for Conversational AI.

Founded in 2002 as a Nuance value-added reseller, Omilia pivoted to a product business in 2008 and has since implemented several large-scale deployments at Capital One, Discover, RBC, Taco Bell, DWP, and PSEG.

The funding also comes with two notable hires: Nick Delis as Chief Revenue Officer and Ryan Kam as Chief Marketing Officer. That signals an ambition to step out of the shadows and compete more aggressively in a market that is heating up quickly.

Interesting development in the conversational AI space.

Yellow.ai announced it is going public via a SPAC merger with Bluerock Acquisition Corp.

As part of the process, it provided several interesting data points: revenue of $34M and enterprise value of $300M.

The most notable aspect is the company’s strategy.

Yellow.ai sees the BPO market as a major growth opportunity. The company plans to use the $200M raised through the transaction to acquire and roll up BPOs, transforming them into AI-native operations powered by its platform.

That ambition mirrors the direction companies like Crescendo have started to pursue.

At the same time, incumbent BPOs are not standing still and investing aggressively in building their own technology platforms. In its latest quarterly results, TP highlighted the growing traction of its TP.ai platform, with 1,100 AI projects now in production and 270 launched in the last quarter.

With Yellow.ai becoming public, we’ll gain much greater visibility into the transformation of the BPO market.

The investor presentation is worth a read: here

Verint held its annual customer and partner conference, Engage, in June.

The company was taken private in August last year and later combined with Calabrio. The merger closed in February, and Engage, held four months later, provided the first opportunity for the new Verint to unveil its refreshed visual identity and outline its direction.

Dave Rhodes addressed concerns from its two large workforce management (WFM) customer bases upfront, reassuring them that both Calabrio WFM and Verint WFM will continue to be supported, with no forced migrations.

At the same time, Verint is integrating Calabrio’s best-of-breed technology components into its CX automation platform.

The central message from the new Verint is its renewed commitment to an outcome-driven go-to-market (GTM) strategy. Verint continues to focus conversations on the business outcomes its 50 bots can deliver, rather than on technology or agentic capabilities. Its approach is incremental and allows enterprises to layer bots onto any existing contact center infrastructure. At a time when agentic narratives dominate, Verint is choosing to zig toward business outcomes while the market zags toward technology narratives.

To support this outcome-driven strategy, the company introduced a risk-free trial last year: if ROI is not demonstrated within the first 90 days, customers pay nothing. ROI calculators are embedded directly into the platform. Verint has also introduced workshops to demonstrate the art of the possible with its bots and identify priority use cases.

BT showcased a compelling example of this outcome-driven approach at the event. Anthony Cass detailed a phased deployment that scaled from 450 to 4,500 sales agents. Using three bots, Coaching, Wrap-up, and CX Scoring, BT reported improvements in real-time agent guidance, cross-sell and upsell, onboarding, and churn reduction. Cross-selling alone drove a $9M increase in annual revenue.

To reinforce this strategy, Verint is making substantial investments in GTM enablement and methodology. The goal is to equip teams to shift customer conversations from technology discussions toward measurable business outcomes.

On the product front, Verint Intelligent Virtual Assistant (IVA) deserves mention, as the company’s self-service offering is often overshadowed by its broader automation portfolio. It has now surpassed the 100-customer mark.

Among the product announcements, two stood out.

The first is Agent Factory, a control layer for its bots that enables customization and orchestration. It allows enterprises to orchestrate workflows across bots and humans, for example, combining post-interaction summarization with a proactive follow-up when customer sentiment indicates significant frustration.

Verint also unveiled a Desktop Intelligence product that uses AI to analyze screen captures, understanding both conversation outcomes and what actually happened during the interaction. This new layer of insight is a meaningful step toward unlocking the value of more than 20 years of interaction data. The solution can also be used to validate that what an agent says is aligned with what is entered into back-end systems.

After a somewhat quiet period, Verint reminded the market that it is a force to be reckoned with. The breadth of its automation portfolio, combined with a differentiated outcome-based go-to-market, gives Verint a distinctive position in a market increasingly centered around agentic AI.

Genesys announced it is acquiring Pinkfish.

Pinkfish is a relatively new, 30-month-old workflow orchestration platform. Despite its founders’ roots in customer service and contact centers, Pinkfish has so far targeted enterprise workflow automation broadly rather than CX specifically.

It can be viewed as a modern, AI-native evolution of automation platforms such as UiPath or ServiceNow, with a strong emphasis on simplifying and democratizing process automation through natural language and AI. Workflows are built from nodes/steps that can be deterministic, agentic, or involve tool calls. The company lists 500 integrations and 25,000 MCP tools.

Pinkfish applies agentic capabilities primarily at the step level to create workflows or agents using natural language, dynamically discover and select tools, and execute simple tasks. This enables a hybrid model where deterministic workflows can leverage AI-driven step execution. The ability to define workflows or steps through natural language is making process automation more accessible to business users.

Genesys has been aggressively expanding its agentic capabilities. It introduced its Agentic Virtual Agent earlier this year, featuring Large Action Models (LAMs) via its partnership with Scaled Cognition alongside traditional LLM-based approaches. LAMs aim to reduce hallucinations, in particular through curated training datasets, including synthetic data and simulated actions, which improve the accuracy of tool use and action execution. Pinkfish extends these capabilities by enabling self-service requests to be executed and fulfilled across enterprise systems.

Perhaps more significantly, this acquisition marks a meaningful step into back-office customer interaction fulfillment and end-to-end process execution. Genesys already had Work Item and Case Management capabilities, and the press announcement explicitly links the acquisition to the latter for long-running processes. The signal is clear: Genesys isn’t just orchestrating customer experiences anymore. It wants to own a larger share of CX workflow execution.

AI-related layoffs and workforce reductions continue to dominate headlines, with contact center employment often cited as being most exposed to AI-driven disruption.

Indeed, Agentic AI is reaching a level of capability where it can absorb a large share of first-level and transactional service interactions. Recent research highlights the scale of potential exposure: Anthropic found customer service roles to be among the most exposed to AI, with up to 70% of tasks potentially handled by AI. Similarly, Goldman Sachs estimated that a significant share of customer service roles are exposed to AI-driven automation.

The US Bureau of Labor Statistics (BLS) recently published its annual employment-by-occupation report, offering a timely opportunity to assess the state of the contact center and customer service workforce.

For the second year in a row, employment of Customer Service Representatives has declined by 5%. The label Customer Service Representatives is somewhat misleading, as it encompasses only a fraction of customer service roles. Another category, Computer Support Specialists, points in the opposite direction. Last year, employment grew 2% after two years of 1-2% decline. This was a surprise, given that customer support automation is one of the most mature use cases for AI.

Combining the two categories shows a 7.5% contraction over the past three years. While Gartner’s 2023 prediction that AI will drive a 20-30% reduction in customer service and support agents by 2026 has not materialized, these numbers signal a structural decline.

But they only capture a fraction — 75% and declining — of the broader customer-facing workforce using customer communication technologies across the customer lifecycle. They do not capture the expansion of customer engagement activities, both in volume and across a broader set of roles across the enterprise.

First, we often assume a stable workload of interactions. In reality, while AI is absorbing a substantial share of interactions, volume and complexity continue to surge, offsetting much of the efficiency gains from automation:

• The scope of activities handled virtually keeps expanding; COVID demonstrated that there is effectively no ceiling to what can be done remotely.

• Processes are becoming more complex, driven in particular by mounting regulatory and compliance requirements.

• Digitization is spreading across more processes, which generate exceptions, adding to customer service workload. In its Inner Circle Guide to Customer Interaction Analytics, ContactBabel found that 80% of complaints reaching contact centers are caused by back-office failures.

• In parallel, the contact center’s managed communication infrastructure has extended its scope beyond traditional service into adjacent functions, including sales and service delivery, such as financial advisory or healthcare interactions.

• Finally, the looming rise of agents acting on behalf of consumers will drive interaction volumes to new highs.

In its AI Automation vs. Human Effort internal study, Zendesk found that automation was actually driving a significant increase in interaction volumes, with larger increases occurring at higher levels of automation. The study reported up to a 200% increase in volume at a 60% automation rate, suggesting that Jevons Paradox is at play in customer service.

Second, the occupational definitions we rely on are increasingly misaligned with how customer-facing work is structured:

• The contact center represents a relatively recent addition to customer service. Its structure reflects a traditional pattern of decomposing customer service work across the enterprise, mirroring Taylorism in early 20th-century manufacturing.

• It is a carved-out “front-end layer” for customer service, optimized for rapid responses through large pools of agents handling high-frequency inquiries, while other tasks are offloaded to specialized departments. Over time, these functions have become even more specialized, creating fragmented customer service experiences, longer resolution times, and friction across organizational boundaries.

• AI is not only becoming the “front door” to service, delivering instant responses, it is also assisting customer-facing associates with specialized tasks, enabling a reaggregation of jobs. Instead of funneling customer communications through the contact center to dispatch the work, a larger pool of the workforce will interact directly with customers.

Taken together, these dynamics suggest the contact center, as a distinct organizational construct, is likely to dissolve into a broader customer-facing operating model.

Rather than framing this shift as AI-driven job elimination, the focus should be on how customer engagement work is redistributed between AI and humans across the enterprise, and how that redistribution reshapes customer-facing roles.

A corollary of this transition is that many entry-level customer service roles that have traditionally served as an entry point into the workforce will disappear, and organizations will need to rethink pathways for employees entering the workforce.

NiCE held its annual conference in Orlando earlier this month.

The event was rebranded as NiCE World, reflecting both the company brand identity introduced a year ago and its broader ambition in the CX space. The event was structured around a show floor designed to facilitate interactions among the 3,000 customers, partners, and employees in attendance.

Under Scott Russell, NiCE has re-centered its GTM around partners, especially global systems integrators (GSIs). Partners now drive ~70% of net-new enterprise logos, with GSIs contributing 14% of ACV, roughly 3x their contribution a year ago. In international markets, close to 90% of the company’s business is partner-led via resale or co-sell.

Scott Russell reasserted the company’s position in the CX AI space with his characteristic conviction, stressing that AI is built in, not bolt on, and emphasizing the importance of a unified platform and its ability to scale.

NiCE reaffirmed Cognigy’s contact center neutrality. NiCE’s strategy is to offer multiple entry points into its suite, CCaaS (contact center), workforce management, or Cognigy’s conversational AI platform.

Nine months after the Cognigy acquisition closed, the integration is effectively complete. Cognigy now has access to NiCE’s proprietary datasets, including vast volumes of historical conversational data. These datasets include highly valuable human-handled interactions, typically the most complex and business-critical conversations, providing Cognigy with a strategic differentiator over other pure-play conversational AI providers. Cognigy has also become the agentic and conversational AI layer of CXone, bringing its best-of-breed capabilities into the NiCE platform.

The event showcased learnings from large-scale AI deployments. One point that caught my attention was that several customers explained they chose to tackle the most complex problems first with AI, rather than starting with simpler use cases, as has traditionally been the adoption pattern. This suggests new levels of confidence in the technology.

Philipp Heltewig offered a compelling example of Cognigy handling over 80% of the 20 million customer conversations with Lufthansa during a one-week strike, including cancellations and flight re-arrangements. He also warned of the imminent rise of consumer-facing conversational agents, urging enterprises to prepare for a significant increase in AI-driven traffic. These examples illustrate the technology’s progress, including instances of AI outperforming humans or managing situations humans could not.

Finally, Jeff Comstock and the product team provided a private preview of several new products and capabilities worth discussing:

• Agentic Engagement Plane is a new layer that allows AI and human agents to operate as a unified workforce. It uses the concept of engagement paths to define the resources that can handle conversations across NiCE and third-party conversational AI and CCaaS platforms. It builds on a global view of the workforce and supports diverse human-in-the-loop models, including full takeover, silent assist, and mid-flight takeover.

• It is supported by the next evolution of Experience Memory, which maintains customer identification and authentication, conversation context, and journey history, including reasoning and decision traces.

• Experience Optimization is evolving into Agentic Analytics spanning conversation, workflow, and outcome data. It leverages a set of domain-specific AI agents to analyze this dataset to identify new automation opportunities and improve existing ones. It draws on a new Screen Intelligence capability to understand everything that happened.

• NiCE is working on a redesigned agent workspace to weave AI agents into the conversation stream, whether handling the conversation, providing recommendations, or requiring human validation.

• AgentForge is a new capability that generates AI agents from artifacts such as documents or specifications.

• Guardian AI is a new governance and control layer that monitors both human and AI agents to enforce guardrails and policies and surface risks.

These building blocks demonstrate NiCE’s continued advancement toward managing customer conversations and resolution workflows across the full span of customer engagement through a unified human and AI workforce.

Last week, Zendesk held its annual user conference, Relate, alongside an Analyst Day. The event marked the culmination of a broad set of organic and non-organic developments that have re-shaped Zendesk’s product footprint into a comprehensive service suite spanning both customer and employee service, with AI embedded natively across the platform rather than dependent on third-party capabilities.

To complete its suite and accelerate its roadmap, the company has become a serial acquirer, snapping up eight companies in the last three years: Cleverly (AI triage and assistance), Tymeshift (workforce management), Klaus (quality management), Ultimate (customer service automation), Local Measure (contact center), HyperArc (GenAI-powered analytics), Unleash (enterprise search), and Forethought (agentic AI).

At the event, Zendesk showcased its Resolution Platform, now organized around four core solutions, each serving as an entry point into the broader suite:

1. Customer Service: Zendesk’s historical core solution.

2. Employee Service: Formally launched as a standalone solution last year and now expanded with IT asset management (ITAM) capabilities.

3. Contact Center: Added through the Local Measure acquisition, the solution powered by Amazon Connect Customer is now fully integrated into the Zendesk platform. The offering includes Zendesk’s agent workspace and reporting capabilities, leveraging the HyperArc acquisition. Zendesk shared that it has added more than 100 new customers since the acquisition.

4. AI Agents: Built on the Forethought acquisition, the solution remains available for non-Zendesk customer service and contact center environments while being integrated into the Zendesk Resolution Platform.

Relate’s theme was the autonomous service workforce, underpinned by a vision of AI and human agents working together, with AI agents spanning both pre-built, highly specialized agents and custom-built agents created through its newly introduced Agent Builder.

Deon Nicholas and Sami Ghoche, co-founders of Forethought, detailed one of its key technologies, the self-improving loop, which now underpins Zendesk’s new Resolution Learning Loop. The loop begins with AI analyzing historical data to surface knowledge gaps and identify automation opportunities, detecting patterns that can be translated into either knowledge base content or agent-driven workflows. It then evaluates conversations across multiple dimensions using customer feedback signals, structured topics, and QA performance metrics such as churn risk. Based on these insights, the system generates and refines agents, which are validated through simulation-based test suites and then deployed via controlled A/B testing in production. Today, these stages are AI-driven with human approval gates for reviewing recommendations, validating test outcomes, and approving production rollouts. The new AI Studio is positioned as the control layer that will progressively automate this loop end-to-end, enabling increasingly autonomous agent improvement.

Zendesk also provided interesting details around its data platform. Unlike many CRM platforms that require custom objects and workflows to accommodate customer-specific business processes, Zendesk has been able to preserve a consistent core data model for the ticket lifecycle and ticket-to-contact relationship. This provides a much more readable context layer for AI and agents.

CEO Tom Eggemeier has emphasized Zendesk’s AI ambitions, sharing progress through AI annual recurring revenue (ARR), which he stresses includes only revenue from AI products. The company reached $250M last quarter and now expects to exit the current fiscal year, ending January 2027, at $400M.

In the backdrop of traditional product category boundaries coming down and with so many vendor options available for AI, differentiation was a recurring theme in discussions. These conversations are often centered on AI capabilities and the data that feeds them. What stood out is that when asked why they are with Zendesk on their AI journey, customers consistently pointed to its people, both in terms of expertise and a genuine commitment to making them successful.

I attended the TP Immersive conference earlier in May.

With Jorge Amar recently appointed as CEO, it was a timely opportunity to take the pulse of an industry at a crossroads. In February, he took the helm of the company founded 50 years ago by industry legend Daniel Julien, succeeding him.

Jorge came from McKinsey, where he was already working on TP’s strategic transformation before taking the CEO role. He stepped in with a clear conviction: TP sits at the center of the workforce transformation AI is about to unleash.

BPOs are first in line for AI disruption. A significant portion of their revenue still comes from the “butts and seats” model built around entry-level roles that AI agents are poised to replace. Under mounting pressure, BPOs are reshaping their business profiles by diversifying into higher-value services, moving up the value chain through more end-to-end process ownership, and shifting toward outcome-based models.

TP operates a large workforce and has to handle significant turnover. Over time, that has made it highly proficient at hiring, onboarding, and continuous training. This creates a foundation for transitioning workers into the new roles that Agentic AI will inevitably create.

Interestingly, a growing number of businesses struggling to realize the expected financial returns from their AI initiatives are turning to the BPO giant to take over functions outright, including the people, with the charter to optimize them.

BPOs are also “project factories.” They manage many projects and, with slim margins, profitability depends on disciplined execution, particularly around timing and delivery. TP is no exception. It worked on more than 500 AI projects last quarter alone, building deep expertise across both the technical and change management dimensions.

Last year, TP introduced TP.ai FAB, a platform designed to support its vision of humans and AI working together.

TP.ai FAB is built on an open framework that integrates with clients’ existing environments and incorporates technologies from selected partners, including Ema, Kore.ai, Parloa, and Sanas.

At the event, TP shared how it built five agentic solutions on top of this framework:

• FAB Assist delivers a complete work environment for human agents, combining AI assistance, KPI-driven coaching, and Sanas voice experience improvements

• FAB Connect provides an agentic customer self-service solution with escalation to human agents

• FAB Operate optimizes traffic orchestration across contact centers and agentic self-service based on customer intents and desired outcomes

• FAB Collect for collections and FAB Growth for sales are end-to-end solutions designed to maximize revenue outcomes through AI.

These solutions position TP to take on more outcome-driven engagements. One challenge, however, is that in multi-provider enterprise environments, the technology stack is often dictated by the customer.

Jorge explained that this transformation requires a fundamentally different conversation with customers. Success depends on the ability to deliver meaningful insight into what businesses should consider. Over the years, TP has built deep expertise across industries and business processes that it now needs to fully harness to earn that strategic seat at the table.

TP’s position at the intersection of technology, people, and large-scale delivery expertise gives it the right to participate in the emerging service market disruption. Key to its success will be the ability to realize definitional outcome-based projects within its customer base.

On the heels of its 1Q26 earnings call, Twilio hosted its Analyst Summit and Signal conference, where it unveiled new products alongside a new positioning.

Before getting there, it is worth pausing on the company’s remarkable turnaround. Once a market darling, Twilio exited 2022 with over $1B in losses, $1B in debt, and gross margins below 50%, entering a period of significant restructuring and strategic reset. Early 2024, Khozema Shipchandler, previously the company’s CFO, succeeded founder Jeff Lawson as CEO.

The financial turnaround was largely achieved in 2024, but it came at the cost of six consecutive quarters of stagnant growth, which led markets to question the durability of the long-term opportunity. In early 2025, Twilio shared a double-digit growth ambition. The latest quarterly results, 20% year-over-year growth after five consecutive quarters of sustained double-digit expansion, were rewarded with a sharp increase in market valuation, setting the stage for what the company is now positioning as its next chapter.

The company has spent the past couple of years unifying its portfolio, including its SendGrid (email) and Segment (CDP) acquisitions, an effort that culminated at Signal with the unveiling of a new Console giving developers a unified entry point across its products.

The next act for the company is to become the infrastructure for the agentic era, enabling continuous customer engagement with humans and AI. At Signal, it launched its conversation layer, adding three new building blocks to Conversation Relay:

Conversation Orchestrator enables continuous conversations across channels with both human agents and AI agents. It enables seamless channel switching, multimodal interactions, AI-human handoffs, and automated outreach for follow-up or notification.

Conversation Memory adds persistent context across interactions. It aggregates customer traits (profiles), identifiers, observations (signals extracted from conversations, such as behavior, sentiment, and preferences that can be manipulated semantically), and interaction summaries.

While leveraging Segment technologies, Conversation Memory is CDP-agnostic and can also connect to any CDPs, CRMs, and data warehouses. For AI agents, Twilio can further enrich context with knowledge through its Enterprise Knowledge ingestion module.

Conversation Intelligence lets you take actions in real-time on live conversation streams. It features a framework that uses language operators. These operators, either from Twilio or custom-built, include surfacing contextual knowledge recommendations, delivering real-time guidance, enabling intelligent routing and next-best actions, monitoring script adherence, and detecting sentiment and conversational signals.

Conversation Intelligence differs from Conversational Intelligence, Twilio’s post-conversation analytics solution, which is now evolving into a new product, Conversation Insights, that will combine the analysis of structured and unstructured data.

The conversation layer also includes Agent Connect, which integrates these building blocks into agentic platforms while abstracting away conversational complexities such as streaming and turn-taking. It supports OpenAI, Microsoft Azure, AWS Bedrock, Anthropic, and LangChain/LangGraph. It’s open-sourced and can be adapted to other frameworks.

Twilio has refocused on APIs and building blocks. That shift has reignited momentum with ISVs, a business that had been hindered by Twilio’s earlier ambitions toward packaged applications. In January 2025, Twilio disclosed that ISVs represented roughly a quarter of revenue, with the segment consistently outgrowing the broader business since then.

Serving both enterprises and ISVs across organizations ranging from SMBs to the world’s largest companies creates inherent positioning challenges. Twilio addresses the diversity of needs across these segments by marketing a collection of modules. While it makes it easy for builders to pick what they need, it can take some stepping back to see where the platform is heading.

Twilio is reshaping itself into an infrastructure provider for customer engagement in the AI world. Looking across its moves, three pillars are taking shape, each rooted in foundational requirements for customer engagement in the agentic era. First, conversation management with its newly unveiled layer. Second context, with Segment getting not only modularized but also embedded in other building blocks. Third identity and trust, with Twilio extending its authentication capabilities through its Stytch acquisition to support AI agents and expanding into compliance and fraud detection.

On April 23, AWS acquired NLX, and for a company with a strong build DNA, any acquisition is intriguing. Over 20 years, the cloud hyperscaler has made only eight that I could identify.

So, who is NLX?

Founded by two brothers, Andrei and Vlad Papancea, in 2018, NLX was born out of Andrei’s experience building conversational applications and the recurring effort required to navigate technology cycles, iterate on experiences, and expand the use cases these applications can support.

This led to the creation of Canvas, NLX’s conversation builder, designed to be to conversational application design what Figma is to UI design. The platform was built to insulate application creators from the complexities of channels, communication infrastructures, and AI technologies.

The obvious question: don’t all conversational AI platforms, including Amazon Connect, already offer a no-code builder?

While almost all conversational AI platforms feature no-code development environments, most remain fairly low-level and require significant technical expertise. NLX created a platform for business users that minimizes dependencies on developers.

Its building blocks abstract application logic across channels, enabling teams to compose multi-modal experiences that guide users to the right modality at each stage of a workflow.

NLX Canvas enables teams to rapidly iterate and expand use cases, a critical success factor for deploying and scaling conversational experiences that today often depends on forward-deployed engineers.

NLX has remained focused on large enterprises with complex requirements, spanning customer experience and other enterprise use cases. Its customer roster includes Comcast, Red Bull, Saks Fifth Avenue, and Toyota.

NLX’s vision aligns closely with Amazon Connect’s mission to create ready-to-use solutions. It already extends AWS Lex Flow Builder and Amazon Connect Flows.

Now part of Amazon Connect Customer, NLX will focus on customer service use cases on the hyperscaler CX platform, while support for other platforms becomes customer-driven.

At this stage of conversational AI adoption, many enterprises are prioritizing customer self-service and selecting best-of-breed platforms. NLX enhances Amazon Connect’s ability to compete in these self-service modernization initiatives.

At its What’s Next with AWS event, the cloud hyperscaler made several announcements, including a particularly ambitious one: the expansion of Amazon Connect into a family of agentic business solutions, which I want to unpack.

You knew Amazon Connect as contact center software. Reset that view: it now represents a family of agentic solutions that package AWS cloud services into ready-to-use offerings.

This announcement carries real strategic weight. A few years ago, a rumor circulated that a team inside AWS was making the case for acquiring HubSpot, sparking debate about the company’s developer DNA and its ability to move up the stack into business applications.

Today, AWS is providing a clear answer to that question. It is not only building business applications, but also using agentic AI to deliver a new class of solutions.

The infrastructure giant has built a $1B business with Amazon Connect, its CCaaS solution. Using Connect as the umbrella brand for its solution portfolio signals a clear intent: replicate the same playbook that drove Amazon Connect’s success.

Colleen Aubrey, Senior Vice President of Applied AI Solutions, outlined the AWS approach: it leverages Amazon’s operational expertise, targets enduring business problems its parent could not solve with commercial software, and infuses talent from Amazon to bring that expertise into AWS. Proprietary “secret sauce” built from these in-house solutions is packaged into AI agents that function as teammates.

In addition to Amazon Connect Customer, the renamed CX/CCaaS solution, the portfolio includes three additional offerings.

Amazon Connect Talent is an AI-powered hiring solution built on Amazon’s experience of hiring 250,000 associates each year for peak season. It turns job requirements into structured, skills-based assessments and AI-led voice interviews that candidates can take on their own terms. It applies science-backed scoring and evaluation models to rank candidates and improve matching quality at scale.

Amazon Connect Health builds on Amazon Pharmacy and One Medical. The goal is straightforward: reduce administrative and documentation work from providers, a huge issue in care delivery. It uses agentic capabilities across the patient journey, starting with scheduling and intake to handle appointment booking, insurance checks, and reminders to reduce no-shows. During the visit, it transcribes patient conversations in real time, surfaces clinical context and coding suggestions, and generates draft clinical documentation for provider review. Every recommendation is explainable, with supporting rationale available to clinicians. Post-visit workflows, including follow-ups and medication adherence tracking, are also automated.

Amazon Connect Decisions is the agentic evolution of AWS Supply Chain, addressing supply chain issues for manufactured goods. It features two AI agents, one handling the demand side and the other the supply side, that proactively surface issues and generate options. A third agent assists human decision-making, including the option to incorporate best practices from top-performing planners. The system is powered by proprietary prediction models developed from Amazon’s experience managing hundreds of thousands of SKUs across its retail and logistics footprint.

Discussing his December prediction that we will soon live in a world with over one billion agents, CEO Matt Garman highlighted that agentic AI is starting to enable things that weren’t possible before. Two in particular: the ability to be given a desired outcome and run the end-to-end workflow toward its realization, and the ability to free up time for humans. The Amazon Connect portfolio is aiming at delivering on that vision.

What I find most promising in AWS’s approach is that, unconstrained by any application heritage, it can tackle end-to-end processes without boundaries, exactly the scope agents are built for. It will be fascinating to watch it mature and grow.

SoundHound AI announced earlier this week its acquisition of LivePerson.

The price stands out. SoundHound AI is snapping up LivePerson for an equity value of $43M while gaining access to $74M of its cash. Including LivePerson’s remaining debt obligations brings the implied enterprise value to roughly $250M.

Not that long ago, LivePerson crossed the $500M revenue mark. For a company once valued at $4.7B, the road here is worth revisiting.

Founded in 1995 as a web chat pioneer, LivePerson successfully weathered the early 2000s internet bubble burst. It found renewed growth with proactive chat following its 2006 acquisition of Proficient Systems. At a time when most brands struggled to deliver effective chat experiences, the ability to engage high-value customers based on available agent capacity proved transformative. The company then expanded into a broader digital customer service platform, reaching $250M in revenue by 2015 before hitting a plateau.

Then came COVID, which boosted revenue 2x, pushing it past the $500M mark in 2022. LivePerson followed with an aggressive growth push, investing heavily in go-to-market and expanding into conversational AI through the acquisitions of BotCentral, Conversable, e-bot7, and VoiceBase, funding them by increasing debt significantly, from $200M to $750M.

2022 marked the peak. By then, contact center players had caught up on digital capabilities, and LivePerson’s digital channel edge had eroded. Its provocative market positioning, declaring the death of contact centers and aiming to eradicate 1-800 numbers, backfired: businesses still needed to support voice and increasingly valued integrated voice and digital experiences over standalone digital customer service solutions.

Revenues declined to $240M in 2025, driven by customer churn. Founder Robert LoCascio was replaced in January 2024 by John Sabino, who reduced debt to around $400M, still a high level relative to its revenues. Revenues continue to decline, expected to land around $200M for the full year. This exit should come as no surprise to industry observers.

In addition to an estimated 120 enterprise customers, SoundHound is getting a digital customer service solution with best-in-class messaging capabilities and a conversational AI platform that earned a Contender position in Forrester’s most recent Conversational AI Platforms for Customer Service Wave.

The press announcement offered little detail on SoundHound AI’s vision. We should expect it when the merger closes in the second half of 2026.

This marks SoundHound AI’s third large acquisition, with $500M as its next revenue milestone and a 2027 revenue target of at least $350M to $400M.

The real work lies ahead: streamlining four conversational AI technology streams and building a strategy that brings together its self-service and human-assisted messaging channel solutions.

2025 marked a pivotal shift for customer-facing conversational AI (CAI).

Last October, I published an industry landscape featuring 650 players and highlighted how difficult it was to break through. For years, the market was defined by the three hyperscalers, AWS, Google, and Microsoft, alongside a small cohort of challengers struggling to break through the $100M ceiling and a long tail of sub-$20M players. CRM and CCaaS providers either built solutions using hyperscaler building blocks or partnered with specialists, yet made limited inroads.

In 2025, the lay of the land was redrawn.

A handful of pure-play providers have now crossed the $100M threshold. Kore.ai was the first, based on my research, to break that ceiling, followed by three others:

• Uniphore, which indicated last October alongside its Series F that it was well past $200M, with scale accelerated by acquisitions.

• SoundHound AI, benefiting from the acquisitions of Amelia and Interactions, ended last year at $170M in revenue.

• Sierra reported reaching $150M in ARR last December.

2 players emerged from the customer support automation corner: Zendesk, which reached $100M in AI ARR last year, and Fin by Intercom, whose CEO Eoghan McCabe disclosed earlier this month that it was about to pass $100M ARR.

Closely behind, a solid cohort of followers above $50M is well-positioned to cross the $100M ARR mark soon, including Ada, Cresta, LivePerson, Observe.AI, Parloa, and Yellow.ai.

ElevenLabs, coming from the Voice AI space, entered the market. While it is difficult to isolate the CAI portion of its $330M ARR last year, it is likely material, and I would not be surprised to see it in this group.

CCaaS providers that relied on partners or hyperscalers are now building or acquiring their way into proprietary CAI offerings. They are likely the largest contributors to NICE AI and Self-Service ARR ($328M in 2025), Genesys Cloud AI ARR (estimated at around $300M in 2025), and Five9 AI ARR ($100M last year).

CRM players have also been aggressively putting together their own CAI offerings, positioning Oracle (Digital Assistant), Salesforce (Agentforce), SAP (CAI/Joule), and ServiceNow (Now Assist) as significant players. We should also include IBM (Watson Assistant) in this group.

Three engines are driving this growth.

First, the incredible pace of progress across three key technologies—GenAI, Agentic AI, and Voice AI.

Second, CAI is enabling new use cases that go beyond what was previously possible and are setting a new benchmark for customer experiences. Three are now mature solutions consistently delivering tangible results: customer support automation, concierge agents unifying sales and service into a continuous assisted experience, and autonomous agents qualifying and processing contact or demo requests, with others gaining traction in proactive service, nurturing and re-engagement, and back-office customer service fulfillment.

Third, businesses are ready to modernize their digital front door. Self-service has consistently topped the priority list for customer service organizations for over a year. A similar trend is emerging in sales to manage incoming appointment and demo requests.

Adoption is accelerating as technology redefines the art of the possible in customer experience. Large CAI players are beginning to establish themselves, coming from the various corners of this vibrant market.

I keep seeing posts about AI-driven layoffs that don’t match what I hear in direct conversations. The recent Citrini Research memo pushed me to write this.

I do not conduct quantitative research, but I review reputable studies. What they consistently show is that overall hiring has leveled off and that layoffs have remained limited, depending on the sectors. For example, retail jobs are declining and openings for white-collar roles, particularly in finance, have dropped.

On a recent podcast, Peter Cappelli, professor at Wharton, offered a grounded perspective. Gartner analyzed 1.15 million job eliminations reported between July 1, 2025, and December 31, 2025, and found that fewer than 1% were due to AI.

There is significant AI washing, with layoffs attributed to AI even when the true drivers are different. It is time to stop.

Zooming in on the technology sector, here is what I have been observing over the past few quarters:

• Many CFOs moved early in 2025 to curb hiring and boost productivity in response to rising economic uncertainty.

• Big Tech layoffs have been dominating the headlines and represent the closest casualty of AI. They are driven by three factors: post-pandemic overhiring corrections, growth slowdowns in large, mature product lines, and cost-cutting to fund their massive AI infrastructure investments.

• The restructuring of X under Elon Musk set a precedent. The company went from roughly 7,500 employees before the takeover to fewer than 3,000 (-60%). Revenues fell from $4.4 billion in 2022 to an estimated $2.9 billion (-34%). That reset reinforced a narrative that many software organizations could operate with far fewer employees.

• VCs are gravitating toward new startups with high revenue-per-employee potential, setting an aspirational $1M target.

• Beyond macroeconomic pressures, the SaaS sector faces sharper investor scrutiny. For years, it operated with deep pockets to grow, while efficiency took a back seat. As growth rates normalize and AI raises questions about long-term potential, investors now expect businesses to drive decisively toward profitability. This has driven, and continues to drive, workforce reductions.

Gartner and Forrester’s latest forecasts align:

• In October 2025, Gartner projected AI’s net impact on global jobs to remain neutral through 2026, emphasizing workforce transformation rather than outright displacement.

• Looking toward a longer time horizon, Forrester anticipates that automation and AI could account for about 6% of job losses in the United States, equating to roughly 10.4 million roles by 2030.

As 2026 unfolds, I am seeing another market shift. In 2025, the top expectation from AI was headcount reduction. Today, it is shifting to a secondary consideration as enterprises embrace AI as an inevitable way of operating.

AI will unquestionably impact the job market, at minimum driving a massive redefinition of roles. Mid- and long-term predictions remain uncertain, and AI washing only clouds the picture.