Five9 Plans to Acquire Level AI for $325 Million: Customer Service AI Begins to Compete Across the Entire Operational Chain

The value of customer service AI is often boiled down to one question: How many human agents can it replace? The proposed deal between Five9 and Level AI brings another question into sharper focus: Once a large volume of customer conversations has been generated, who can turn them into actionable operational improvements?

On November 5, 2026, Five9 announced that it had signed a definitive agreement to acquire Level AI for approximately $325 million, to be paid in cash. The company expects to close the transaction in the fourth quarter of 2026, subject to customary conditions. Therefore, this is a signed-but-not-yet-completed transaction and should not be treated as a completed merger at this time.Official Acquisition Announcement,Five9 Quarterly Earnings Announcement

The ability to be acquired, covering both before and after the conversation

Five9 provides a cloud-based contact center platform. Level AI’s applications cover agent performance improvement, automation of voice and digital channels, and customer conversation analysis. The two companies have been collaborating since 2022 and already have application integrations in place. The official announcement also stated that, following the acquisition, Level AI’s applications will continue to be offered as standalone solutions while being integrated into the AI application layer of the Five9 Genius CX Platform.Official Trading Guidelines

From a business logic perspective, this gives Five9 the opportunity to address two types of needs: one group of customers is willing to purchase a complete contact center platform, while another group is not yet ready to migrate their existing systems but wants to add AI capabilities first. The latter approach is particularly noteworthy, as it separates “adopting AI” and “replacing the core platform” into two distinct decisions.

From the purchaser’s perspective, separating the two changes may reduce the initial scope of migration. However, this is only a potential advantage and does not automatically guarantee a simple deployment. Standalone applications still need to integrate with recording systems, text data, customer records, and business rules, as well as handle identity management, data retention, and access permissions. While software can be purchased separately, data governance is difficult to fully separate.

The administrative work behind customer service is where the real market opportunity lies.

In its transaction statement, Five9 emphasized that behind-the-scenes work—such as quality assurance, coaching, research, and analysis—requires extracting information from customer conversations; it also noted that both companies will continue to operate as independent entities until the transaction closes, and specific product integration plans will be announced as the process moves forward.Five9 Official Blog

This approach addresses a common operational challenge: While companies interact with customers on a daily basis, identifying common issues from these interactions and then getting product, training, or process managers to take action is often more difficult than generating a summary of a conversation.

For example, if a certain type of post-sales inquiry continues to increase, what managers really need is to distinguish whether the issue stems from the customer not understanding the instructions, inconsistent promises made by the sales channel, or a problem with the product itself. Different causes require different handlers and different evidence requirements. If AI simply condenses hundreds of call segments into a single sentence—such as “the customer was dissatisfied with the experience”—the amount of information is reduced, but the quality of decision-making may not necessarily improve.

Therefore, the long-term value of this transaction should be assessed based on whether “information can be integrated into the workflow”: Are issues accurately categorized? Can they be traced back to specific conversations for verification? Are there designated individuals responsible for implementing improvements? And can changes be observed after implementation? A visually appealing analytics dashboard is just one part of the equation.

Integration can bring convenience, but it may also increase dependence.

When a platform is integrated with specialized AI applications, it typically reduces some of the repetitive configuration work and makes it easier to incorporate analytics into the interfaces that agents and supervisors use on a daily basis. For companies without dedicated integration teams, this is a real draw.

However, clients also need to evaluate new dependencies. If models, data analysis, quality control rules, and workflows are gradually consolidated into a single supply system, future migrations will no longer involve just audio files, but will also include tagging systems, evaluation criteria, training data, and established operational practices.

At present, the roadmap for standalone applications offers some flexibility but cannot replace specific contracts. Companies should continue to ask whether data can be exported, how historical scores will be retained, whether existing integrations will continue to be supported, and how they will be notified of price adjustments. The strategic direction outlined in the announcement and the final products that can be purchased and deployed remain two distinct levels of information.

Don't base acceptance solely on “how much call time was saved”

If average handling time is set as the sole metric, the system may encourage ending conversations as quickly as possible without actually resolving the issue. The same applies to AI quality assurance: scoring all conversations increases coverage, but whether the scoring criteria are accurate and fair still requires separate verification.

A prudent pilot program should simultaneously monitor the first-call resolution rate, the repeat contact rate, the volume of manual corrections, and customer feedback, while also conducting spot checks of the original content corresponding to the conclusions. When agent performance is involved, manual review should be permitted in particular. Factors such as accent, language, the complexity of the interaction, and the customer’s emotional state can all influence the model’s judgment; therefore, a single automated score should not be the sole basis for evaluating an agent’s performance.

It is also important to calculate the total cost comprehensively: in addition to the application costs, this includes integration, data cleansing, rule maintenance, and manual review. Automation only truly improves operations—rather than simply shifting labor from the front lines to the system maintenance team—when the benefits outweigh these ongoing costs.

What Companies Should Ask Their Suppliers Now

For customers currently using these services, there is no need to rush into a migration simply because of a transaction announcement. A more practical approach is to clearly outline the issues related to the next renewal or pilot phase: who supports the current features, who responds to outages, whether existing data will be processed at a different location, whether new features will be billed separately, and which integrations are still only in the planning stages.

At the same time, suppliers should be asked to demonstrate a complete workflow, rather than just showing a demo clip. You can start with a set of historical conversations with known outcomes to examine how the system identifies issues, draws conclusions, and generates follow-up tasks; business leaders can then determine whether the recommendations are truly actionable. If the dialogue analysis cannot provide a clear basis for its conclusions, it should not be directly incorporated into management decisions.

The evaluation period should not be limited to the first week after launch. Updates to the knowledge base, changes in business policies, promotional campaigns, or product changes can all render old decision rules obsolete. Companies need to determine who will be responsible for ongoing maintenance and when to resample and recalibrate. Clearly defining these responsibilities in advance often has a greater impact on long-term effectiveness than adding a few more automated features.

Five9’s decision to acquire additional capabilities in this instance reflects its commitment to customer-facing AI. Whether the $325 million investment will yield sustained returns depends on the closing of the deal, product integration, and customer adoption. For enterprise users, a more valuable metric is whether each conversation helps improve the next service interaction—and whether this improvement process can be explained, verified, and managed.