Agentic AI arms race: Zendesk and Forethought challenge Salesforce Agentforce and ServiceNow Moveworks

Zendesk, the privately held customer experience platform backed by Permira and Hellman and Friedman, has entered into a definitive agreement to acquire Forethought, an enterprise AI agent company, in what the company describes as its largest deal in nearly two decades. The proposed transaction will integrate Forethought’s self-improving AI infrastructure into the Zendesk Resolution Platform, adding autonomous workflow generation across voice, chat, email, and enterprise systems. Zendesk reported approximately $200 million in AI annual recurring revenue for 2025, with projections pointing toward $500 million in 2026, and the company expects autonomous AI agents to handle more customer service interactions than human agents before the year is out. The deal is expected to close by the end of March, subject to regulatory approvals, and Zendesk says it will push its product roadmap forward by more than twelve months.

What does Forethought bring to Zendesk that its existing AI agent platform could not already do on its own?

The acquisition is not primarily about adding AI to a platform that already has it. Zendesk’s existing AI agents already resolve more than 80 percent of interactions end-to-end across its customer base, and the company has assembled a meaningful portfolio of adjacent capabilities through a string of recent deals, including the December 2025 acquisition of Unleash, an enterprise search platform, and the earlier acquisition of Finnish automation provider Ultimate. What Forethought provides is a different architecture of learning. Where current Zendesk agents resolve interactions and flag gaps for manual review, Forethought’s platform is engineered to detect workflow deficiencies, generate new procedures to address them, and test those optimizations before deployment, all without human intervention.

This distinction matters strategically. The customer service AI market is bifurcating between platforms that automate well-defined, bounded queries and those that can genuinely handle ambiguous, multi-step service journeys. Forethought pushes Zendesk firmly into the second category. The combined capability, which Zendesk will brand as Forethought AI Agents by Zendesk, is designed to execute complex multi-step procedures autonomously, extend into voice channels, and reach into enterprise back-end systems even where APIs do not exist. That last capability, sometimes referred to in the industry as computer use, is particularly significant for enterprise accounts where legacy infrastructure often makes integration costly or impossible.

How does Zendesk’s M&A strategy over the past 18 months position it against Salesforce, ServiceNow, and AI-native challengers in 2026?

Zendesk has deployed close to half a billion dollars in acquisitions over the past eighteen months, a pace that sits in clear contrast to the consolidation posture adopted by several of its peers. The competitive context is instructive. Salesforce has invested heavily in agentic AI through acquisitions including Convergence.ai and Cimulate to power its Agentforce platform. ServiceNow spent approximately $2.85 billion to acquire Moveworks, an enterprise search and AI agent developer. NICE acquired Cognigy for around $955 million to bolster its customer experience platform. Each of these moves reflects the same fundamental bet: that the transition from conversation management to autonomous resolution is structural and irreversible, and that owning the AI layer in enterprise service workflows is a defensible long-term position.

Zendesk’s differentiation lies in its claim to resolution specificity. The company has built its recent strategy around the premise that its competitors are adding agentic AI features onto broader platforms, while Zendesk is building an entire platform architecture around the concept of resolution as a defined outcome. Whether that framing holds under pressure from Salesforce’s distribution advantages and ServiceNow’s enterprise penetration remains a live question, but the Forethought acquisition does tighten the thesis. It also introduces a noteworthy competitive nuance: Forethought will remain available to new customers who are not Zendesk users, a move that extends the addressable market while preserving optionality for enterprise accounts already invested in competing platforms.

The threat from AI-native startups is real and not trivial to dismiss. Sierra, one of the more prominent AI customer service challengers, reached $100 million in annual recurring revenue within two years. Decagon, Dolphin AI, and Yellow AI are also scaling. These companies carry none of the legacy infrastructure of an established helpdesk platform, which gives them architectural agility. Zendesk’s response is scale, distribution, and data volume. With roughly 20,000 customers actively deploying AI and an estimated five billion issues resolved annually across its platform, Zendesk has a proprietary training asset that pure-play startups cannot replicate in the near term.

What execution risks does Zendesk face when integrating Forethought and delivering on the self-improving AI roadmap it has outlined?

The strategic case for the acquisition is coherent. The execution risks are equally worth examining. Zendesk has made twelve acquisitions across its history, but the current acquisition pace, three deals in 2025 alone before this transaction, compresses the integration timeline considerably. Each deal introduces engineering dependencies, go-to-market alignment requirements, and product roadmap coordination challenges that accumulate. The claim that Forethought accelerates Zendesk’s roadmap by more than a year is credible in isolation, but delivering on that claim while simultaneously integrating Unleash, operationalizing voice AI agents, and sustaining commercial momentum across a customer base of over 100,000 organizations is a non-trivial operational challenge.

There is also the question of the self-improvement architecture itself. Zendesk’s Resolution Learning Loop promises continuous improvement by learning from every customer interaction without manual retraining. Forethought’s system extends this by adding autonomous procedure generation and pre-deployment testing. In controlled deployments these capabilities perform well. At enterprise scale, with the heterogeneity of data environments, compliance obligations, and domain-specific workflow requirements that large accounts carry, autonomous workflow generation introduces governance complexity that customers will need to manage carefully. Regulated industries in particular will require clear audit trails and override mechanisms before they allow self-modifying AI agents to operate without human checkpoints. This is not a reason to doubt the acquisition, but it is a deployment friction that sales and implementation teams will need to address proactively.

How does Zendesk’s pricing shift to resolution-based billing change the commercial dynamics for enterprise customers adopting AI agents?

Zendesk made a significant commercial architecture change in 2025 when it shifted billing away from interaction volume and toward effective resolution as the contractual unit. This outcome-based pricing model is strategically aligned with the resolution identity the company has adopted, and it creates a shared incentive structure between Zendesk and its customers. If the AI agent resolves the issue, the customer pays. If it does not, the cost structure shifts accordingly. In practice, this model creates transparency that per-seat or per-interaction pricing does not, and it positions Zendesk well against vendors whose pricing obscures the value delivered.

The commercial dynamics are not without tension, however. Resolution-based billing increases customer satisfaction by reducing friction, which in turn tends to increase the volume of interactions as customers engage more freely. That interaction growth increases Zendesk’s operational costs even as it raises customer lifetime value. Managing that tradeoff at scale will require the kind of operational intelligence that Forethought’s platform is specifically designed to provide, creating a reinforcing loop in which better AI performance supports the commercial model and better commercial outcomes fund further AI development. How Zendesk prices the expanded Forethought capabilities for existing versus new customers will be a key variable to watch after the deal closes.

What does Zendesk’s Forethought acquisition signal about where the broader enterprise service software market is heading in 2026 and beyond?

The transaction is a datapoint in a pattern that is now well established. Over 50 agentic AI acquisitions have been announced globally across the past two years, reflecting a broad consensus among enterprise software vendors that autonomous resolution capability is the next defensible moat in service software. What distinguishes this deal from several of the others is the specificity of the learning architecture Forethought brings. Most agentic AI platforms can handle the resolution of routine queries with sufficient training. Forethought’s contribution is the ability to diagnose what the agent cannot yet resolve, generate a procedure to address it, test that procedure, and deploy it autonomously. That closing of the feedback loop, without requiring human intervention at each step, is the technological distinction Zendesk is paying for.

The broader implication for the enterprise service market is that the productivity argument for AI agents is no longer the primary commercial driver. The architecture argument is. Enterprise buyers are no longer asking whether AI can automate a percentage of their support volume. They are asking which platforms can improve autonomously, integrate into systems that predate the API era, and operate across voice, chat, and digital channels without requiring a separate integration layer for each. Zendesk’s resolution-centric platform, reinforced by Forethought’s self-learning capability and Unleash’s enterprise search, is a credible answer to that question. Whether the integration delivers on the timeline promised will determine whether Zendesk can hold its position as competitors with larger balance sheets and broader distribution close the product gap.

Key takeaways: What the Zendesk and Forethought acquisition means for CX technology, enterprise AI buyers, and competing platforms

• Zendesk’s proposed acquisition of Forethought is its largest in approximately twenty years and follows close to half a billion dollars in M&A spending over the past eighteen months, signaling a deliberate bet that autonomous resolution is the defining architectural shift in enterprise customer service software.
• Forethought adds self-improving AI architecture to Zendesk’s Resolution Platform, enabling agents to detect workflow gaps, generate new procedures, and test optimizations autonomously without manual retraining. This is a qualitatively different capability from the AI automation already present in Zendesk’s platform.
• Zendesk reported $200 million in AI annual recurring revenue for 2025 across approximately 20,000 AI-active customers. The company has projected AI ARR of up to $500 million for 2026, a 150 percent year-on-year increase that the Forethought capabilities are expected to support.
• The transaction is expected to push Zendesk’s product roadmap forward by more than a year, a meaningful accelerant in a market where competitors including Salesforce, ServiceNow, and NICE have each committed significant capital to agentic AI acquisitions of their own.
• Forethought will remain available to customers outside the Zendesk platform, extending reach into enterprise accounts currently using competing CX infrastructure without requiring platform migration.
• Zendesk’s shift to resolution-based billing aligns its commercial model directly with AI performance, creating shared incentives with customers but also introducing interaction-volume growth as a cost management challenge at scale.
• Governance and compliance complexity in regulated industries represents the primary adoption friction for autonomous workflow generation. Enterprise accounts will need clear audit and override frameworks before deploying self-modifying AI agents at scale.
• The competitive pressure from AI-native startups including Sierra, Decagon, and Yellow AI is real, but Zendesk’s data asset, approximately five billion issues resolved annually across more than 100,000 customers, provides a proprietary training advantage that early-stage competitors cannot replicate quickly.
• The deal is expected to close by end of March 2026. Integration execution across three major acquisitions in under twelve months, Forethought, Unleash, and Ultimate, will be the near-term operational test of whether Zendesk can deliver on its accelerated roadmap commitments.
• For enterprise software buyers, the Zendesk and Forethought combination represents one of the more complete end-to-end resolution architectures in the market. Procurement decisions over the next twelve months will benefit from close evaluation of integration depth, governance controls, and pricing transparency as the combined product matures.