Perplexity wants enterprise AI on its terms. Convincing CISOs is the harder problem

• 
  
  Perplexity enterprise AI has arrived, aimed at Microsoft, Salesforce, and OpenAI.
  
  


• 
  
  The orchestration may be smart, but issues for remain.
  
  

Perplexity’s push into Perplexity enterprise AI is not subtle. At its inaugural Ask 2026 developer conference, held in a former church in San Francisco’s North Beach on March 12, CEO Aravind Srinivas made the company’s

ambition

plain: “A traditional operating system takes instructions; an AI operating system takes objectives.”

The venue felt deliberate. So did the message.

Two weeks after Perplexity Computer launched for consumers and triggered what the company described as a viral moment – users demonstrating it by building Bloomberg Terminal-style dashboards and replacing six-figure marketing tool stacks over a single weekend – Perplexity brought the platform to enterprise customers.

More than 100 organisations reportedly messaged the company in one weekend demanding access before the formal launch arrived. The product has substance. Computer for Enterprise orchestrates up to 20 frontier AI models inside isolated Firecracker virtual machines, the same microVM technology AWS developed for its Lambda serverless platform, ensuring that session data cannot cross between users or reach production infrastructure.

Enterprise customers get native connectors for Snowflake, Salesforce, Datadog, SharePoint, and HubSpot, Slack integration via direct message or shared channel, SOC 2 Type II certification, SAML single sign-on, SCIM provisioning, granular admin controls, and a zero data retention option.

Pre-built workflow templates cover legal contract review, finance audit support, and customer support ticket triage. Perplexity has also expanded its finance data tools to pull from more than 40 live sources, including SEC filings, FactSet, S&P Global, and LSEG.

The model-agnostic bet at the centre of everything

The architectural choice that underpins all of this is model agnosticism. Perplexity does not build its own frontier models. Computer routes tasks in whichever model is best suited for each job. The company’s own internal use data makes the case for why: in January 2025, 90% of enterprise queries were routed to just two AI models.

By December 2025, no single model accounted for more than 25% of use. Enterprises, in practice, are already running multi-model workflows. Perplexity is building the harness that manages them. That is also the central tension in its commercial position. As Axios observed, Perplexity’s core challenge is convincing customers to pay a monthly fee to a company that does not build its own models, when they could go directly to OpenAI, Anthropic, or Google.

The orchestration layer is the answer, but it requires enterprises to believe that Perplexity’s routing intelligence justifies the intermediary cost and the risk of routing sensitive data through a startup that is still, by enterprise standards, very young.

Where the trust question actually sits

The technology case is credible. The trust case is where Perplexity’s enterprise ambitions meet real friction. The company is three years old and is asking CISOs to route Snowflake data, legal contracts, and proprietary business intelligence through its platform.

Dmitry Shevelenko, Perplexity’s chief business officer, addressed this directly at Ask 2026. “The story of Perplexity in the enterprise is that we have tens of thousands of enterprise customers and only six people on our enterprise go-to-market team.”

The intention is to signal product-led growth: customers are arriving without aggressive sales pressure. The risk that statement inadvertently signals is a support and accountability structure that most enterprise security teams would find thin for a platform handling production data at scale.

Perplexity has moved to close the credentials gap. Beyond SOC 2 Type II, the platform claims GDPR and HIPAA compliance and states that enterprise customer data is not used to train or fine-tune its models. More than 7,000 Enterprise Pro customers are already on the platform, including NVIDIA, Databricks, and Stripe. The US Anti-Doping Agency, where data privacy is a regulatory requirement not a preference, is also among them.

But past incidents are in the record. In 2024, Perplexity faced scrutiny over web scraping practices that allegedly ignored the Robots Exclusion Protocol, pulling content from publishers that had attempted to block access. Security researchers also identified prompt injection vulnerabilities in its browser product.

The company disputes the severity of those findings, but for enterprise buyers evaluating a platform to handle sensitive workflows, the history sits with the certifications in any due diligence process.

The revenue gap that explains the urgency

The financial stakes clarify why Perplexity is moving at this pace. Research firm Sacra estimated the company reached approximately US$148 million in annualised revenue by mid-2025. Its own internal projections target US$656 million in annual recurring revenue by the end of 2026, a figure that requires roughly 230% growth in twelve months. Computer for Enterprise, priced to scale in entire organisations, is the primary vehicle for closing that gap. The company has raised approximately $1.5 billion at a $20 billion valuation.

The market timing is not wrong. Fortune Business Insights projects the global agentic AI market will grow from US$9.14 billion in 2026 to $139 billion by 2034. Gartner forecasts that 40% of enterprise applications will include task-specific AI agents this year, up from less than 5% in 2025. Perplexity is entering this market at the exact moment demand is sharpest.

A third position in a market that thought it had two

Perplexity’s arrival adds a third credible architecture to a field largely defined by two. OpenAI’s Frontier and Anthropic’s Claude Cowork operate as intelligence layers above existing enterprise systems. Salesforce and ServiceNow sit embedded in them. Perplexity is staking out the orchestration layer as a distinct category: not a model provider, not a system of record, but the routing intelligence that decides which model handles which task in all of the above.

Whether that position holds under pressure from companies with larger enterprise sales organisations, deeper institutional relationships, and their own model abilities is genuinely uncertain. What is not uncertain is that the enterprise AI agent market in 2026 is not a two-horse race, and CIOs evaluating their agent strategy now have a more complicated decision than they did at the start of the year.

(Image source:

Perplexity

)

Want to experience the full spectrum of enterprise technology innovation?

Join

TechEx

in Amsterdam, California, and London. Covering AI, Big Data, Cyber Security, IoT, Digital Transformation, Intelligent Automation, Edge Computing, and Data Centres, TechEx brings together global leaders to share real-world use cases and in-depth insights. Click

here

for more information.

TechHQ is powered by

TechForge Media

. Explore other upcoming enterprise technology events and webinars

here

.