Signals Inbox·July 20, 2026·AI Customer Support

Sierra vs Decagon: who is winning AI customer support?

Sierra is winning the bigger race in AI customer support, with several times Decagon’s commercial scale and a deeper hold on major enterprises. Decagon, though, may already have the sharper operating product for the teams running these agents every day.

Q1Why does everyone compare Sierra and Decagon?

Sierra and Decagon keep getting compared because they are chasing the same prize: becoming the main AI layer between large companies and their customers.

Both were founded in 2023. Both build agents that handle chat, email and voice. Both want those agents to complete real work, such as replacing a card, issuing a refund, changing a subscription or updating an account, rather than answering a few questions before handing the conversation to a person.

The overlap has become much harder to ignore lately. Decagon raised $250 million at a $4.5 billion valuation in January 2026. Sierra raised $950 million at a $15.8 billion valuation a few months later. Decagon added more than 100 global enterprise customers during its latest fiscal year, while Sierra says it now works with more than 40% of the Fortune 50.

They are increasingly showing up in the same enterprise evaluations, competing for the same customer-support budgets and trying to replace the same older software stacks.

Q2Why is it so hard to say who is winning?

Sierra is much larger today, while Decagon sometimes looks stronger when we examine how support teams actually build and improve their agents.

Sierra leads on revenue, funding, valuation and penetration among very large regulated companies. Decagon has published more detailed examples of quick launches, lower service costs and agents that customer-experience teams can modify without relying heavily on engineers.

Their ambitions are also starting to separate. Decagon remains centered on the daily operation of AI customer support. Sierra increasingly wants to automate longer business processes, including mortgages, insurance claims and healthcare administration.

There are really two races. One is about commercial scale and control of the largest enterprise accounts. The other is about which platform support teams find easier to deploy, manage and improve.

Q3Which company is actually bigger today?

Sierra is currently operating at several times Decagon’s commercial scale.

Sierra said in February 2026 that it had passed $150 million in annual recurring revenue. It had reached $100 million in seven quarters, then added another $50 million during the following quarter.

Decagon does not disclose revenue. Sacra estimated that it reached roughly $35 million in annualized revenue in October 2025, compared with around $10 million at the end of 2024. The figure is an external estimate and is older than Sierra’s disclosure, so it should be treated as directional.

Even allowing for further Decagon growth, the gap remains large. Comparing the latest available figures puts Sierra at roughly 4.3 times Decagon’s scale.

Funding and valuation tell a similar story. Sierra has raised around $1.6 billion, versus approximately $481 million for Decagon. Sierra’s $15.8 billion valuation is about 3.5 times higher.

Sierra and Decagon compared by commercial scale

Q4Is Sierra growing faster too?

Sierra is adding more revenue in absolute dollars, although Decagon may still be growing nearly as fast in percentage terms.

Decagon’s estimated annualized revenue climbed from around $10 million at the end of 2024 to $35 million ten months later. That represents growth of about 250%. The company also said both GAAP revenue and ARR more than tripled year over year during the third quarter of 2025.

Sierra’s rise has continued from a much larger base. It moved from approximately $20 million to $100 million in ARR over a year, according to Sacra, then added at least $50 million during its next reported quarter.

That single quarterly increase was larger than Decagon’s entire last estimated revenue base. The telling part is that Sierra’s growth remained extremely fast after it had already reached nine figures.

Decagon can still close the percentage-growth gap as newer enterprise contracts expand. For now, Sierra appears to be adding distance rather than simply protecting an early lead.

Q5Who has the stronger enterprise customers and global reach?

Sierra is clearly ahead at the very top of the enterprise market, especially in regulated industries and large international accounts.

Sierra says one quarter of its customers generate more than $10 billion in annual revenue, while half generate more than $1 billion. Its named customers include Cigna, ADT, DIRECTV, Gap, Rivian, SiriusXM, Rocket Mortgage, SoFi and Sutter Health.

The company also reports a wide operating footprint: its agents touch more than 95% of US shoppers, half of US families through healthcare deployments and roughly one quarter of European banking. These are Sierra’s own reach calculations, but the customer list supports the broader pattern.

Sierra has also opened operations in markets including the United Kingdom, France and Japan. SoftBank Vision Fund 2 joined its Japanese expansion, giving Sierra useful local relationships in a market where enterprise software sales usually require a strong domestic presence.

Decagon has become a serious global enterprise vendor in its own right. Deutsche Telekom, Mercado Libre, Avis Budget Group, Block, Chime, Duolingo, Hertz, Notion, Rippling and Hunter Douglas all appear among its customers. Hunter Douglas is rolling out localized agents across 11 countries.

Decagon’s customer acquisition pace may be higher. Sierra’s accounts, however, are more concentrated among the world’s largest and most regulated companies, where contracts can expand across many departments, countries and channels.

Q6Did Decagon take Chime away from Sierra?

Decagon appears to have won a major share of Chime’s AI support operation, although the public record does not confirm a complete replacement of Sierra.

Sierra previously published a Chime case study showing that its agents improved resolution from 50% to more than 70%. The deployment covered common issues involving direct deposits, passwords and account access.

Decagon now presents Chime as one of its flagship customers. Chime says it evaluated several vendors before selecting Decagon for chat and voice. Decagon’s agents reportedly resolve close to 70% of interactions, automate hundreds of thousands of chat messages and handle more than one million voice calls each month.

That is a substantial production deployment. Chime also uses a shared knowledge system across chat and voice, while its support teams modify agent behavior through Decagon’s Agent Operating Procedures.

We cannot tell whether Sierra still handles selected workflows behind the scenes. Large companies often split channels, retain several vendors during migrations or run overlapping systems. What we can say is that Decagon won a large, strategically important deployment after a competitive evaluation.

Q7Who is delivering better customer outcomes?

Decagon currently has the more impressive collection of detailed customer results, particularly when we look at implementation speed, cost reduction and revenue generated by agents.

Chime reports close to 70% resolution across chat and voice. Duolingo’s English Test launched Decagon chat in about one month and immediately reached 80% deflection. Noom increased deflection from 59% to 68% within 30 days.

The financial outcomes are even more striking. ClassPass says Decagon reduced certain support costs by 95% after winning an evaluation against 12 other AI products. Hunter Douglas attributes more than $1 million in revenue to conversations completed entirely by Decagon agents. Customers who interacted with those agents had an 85% higher average order value.

Sierra also has strong cases. SoFi’s agent handles more than 50,000 conversations each week, contains 61% of them and improved NPS by 33 points in three months. Ramp reports 90% automated resolution. Thrive Market recorded more than a 50% improvement in case resolution and close to 90% customer satisfaction for AI-handled interactions.

The measures do not line up perfectly. Companies define resolution, containment and deflection differently, and vendors naturally publish their strongest deployments. Even so, Decagon gives us more evidence connecting its product to cost savings and direct revenue.

Pricing reinforces the difference in positioning. Sierra usually charges for completed outcomes, which aligns its economics with successful resolutions. Decagon offers both per-conversation and per-resolution arrangements, giving buyers more flexibility when they still want human involvement or cannot define a clean end result.

Selected customer outcomes reported by Sierra and Decagon

Q8Which company launches faster?

Decagon has the stronger record for getting focused customer-support deployments live quickly.

Duolingo launched chat in roughly one month after its previous provider had spent a year failing to do so. Hunter Douglas moved from 10% of traffic to full volume in a week after the initial build. Noom improved its deflection rate by nine percentage points within its first 30 days.

Decagon’s operating model helps explain that speed. Customer-experience teams define workflows through Agent Operating Procedures, then use simulations, monitoring and experiments to test changes without rebuilding long decision trees.

Sierra has also moved surprisingly fast inside very large companies. One major healthcare organization reportedly launched in seven weeks. British retailer Next introduced chat, voice and WhatsApp in six weeks. SoFi reached meaningful production volume within three months.

For a focused support rollout, Decagon looks faster today. Sierra’s timelines become more competitive when the project involves regulated data, several backend systems and approval from a large enterprise.

Q9Which platform gives support teams more control?

Decagon currently gives customer-experience teams the clearest way to control agent behavior without waiting for engineering support.

Its central interface is the Agent Operating Procedure, or AOP. Teams describe policies and multi-step workflows in structured natural language. They can update refund rules, cancellation handling, account recovery or exceptions without rebuilding a traditional workflow tree.

Decagon has built the rest of the platform around that interface. Simulations test changes before launch. Experiments compare different behaviors. Watchtower evaluates live conversations. QA Hub combines automated and human review. Duet finds recurring failures and proposes improvements.

Duet Autopilot pushes this further by turning production problems into suggested agent updates. During beta testing, Decagon said customers accepted 93% of its proposed workspaces after review. The result comes from Decagon’s own evaluation, but it shows where the product is heading: fewer manual audits and faster corrections.

Sierra narrowed the gap with Ghostwriter. Customers can upload procedures, transcripts or recordings, describe the outcome they want and let the system create integrations, context blocks, simulations and agent behavior.

Ghostwriter may eventually require less manual setup than editing AOPs. Decagon still offers the more mature operating environment for support leaders who want to inspect and adjust the agent themselves.

Q10Who is ahead in voice AI?

Sierra has a narrow lead in voice infrastructure, while Decagon has one of the strongest disclosed customer deployments.

Decagon’s work with Chime shows that it can run voice at meaningful scale. The agents share knowledge with chat, which makes it easier to correct policies across both channels. Hunter Douglas has also moved one Australian brand to 100% voice traffic and is expanding the setup into additional markets.

Sierra’s voice operation appears broader. One year after launch, the company said it was powering hundreds of millions of voice conversations. It supports more than 70 languages and uses company-specific context to improve transcription accuracy.

Sierra has also released τ-voice, an open benchmark covering 278 customer-service tasks across retail, telecom and airlines. The tests introduce accents, interruptions, background noise and telephone compression, which are among the main reasons voice agents break in production.

Decagon’s Chime deployment already exceeds one million monthly calls. Sierra still wins this section through its wider platform scale, multilingual coverage and deeper public work on voice reliability.

Q11Who is solving the hardest problem in AI customer support?

Decagon is slightly ahead at improving agents after launch, while Sierra is better positioned for complex work that stretches across several interactions.

The hardest issue these days is keeping an agent reliable as company policies, products, customer behavior and underlying models keep changing. A strong launch means little when accuracy deteriorates three months later.

Decagon has created a tight improvement loop. AOPs define the intended behavior. Simulations test it. Watchtower and QA Hub identify production failures. Duet then suggests changes and measures whether they worked.

Sierra approaches the problem through more sophisticated infrastructure. Its agents can use several models, divide tasks into smaller components and rely on supervisors to verify execution. Ghostwriter builds and optimizes the context given to those models.

Decagon’s system feels more complete for teams managing thousands of daily support conversations. Sierra becomes more compelling when the agent must remember context, coordinate several systems and continue working over days or weeks.

Q12Which company has better technology?

There is still no credible head-to-head benchmark, but Decagon has stronger evidence from customer evaluations and Sierra has stronger public research.

ClassPass tested Decagon against 12 other AI support products before choosing it. Chime also conducted a broad evaluation across chat and voice. Duolingo had already tried another vendor and described Decagon’s deployment and maintenance process as far better.

Those customer decisions carry weight because they involved working products. The losing vendors are rarely named, however, so we cannot assume Sierra was directly involved in every comparison.

Sierra’s strongest technical contribution is τ-bench, an open framework for evaluating agents that must talk to users, use tools and change database states correctly. It later extended the work into knowledge retrieval and voice.

Decagon created DuetBench to measure whether an agent can improve itself. The benchmark supports the company’s Autopilot story, though it remains more closely tied to Decagon’s own product.

Today, Decagon has the stronger procurement evidence. Sierra has the more credible open research program. Neither side has proved that its underlying agent is consistently smarter across real enterprise workloads.

Q13Which company is moving beyond customer support faster?

Sierra is moving into broader business operations much faster.

The company already supports activities such as refinancing homes, originating mortgages, processing insurance claims and guiding customers through financial products. These workflows involve documents, approvals, follow-ups and several internal systems.

Its latest Horizon launch extends agents across days or weeks. Sierra describes use cases such as loan origination and healthcare prior authorization, where an agent must remember previous interactions and keep pursuing a goal over time.

That expands Sierra’s potential revenue far beyond contact-center budgets. A company may pay much more for an agent that completes a mortgage or obtains treatment approval than for one that answers a billing question.

Decagon is broadening too. Its Proactive Agents can initiate conversations for onboarding, retention and account updates. Hunter Douglas has already shown that Decagon agents can generate meaningful sales.

The split is getting clearer. Decagon is stretching customer support across the full customer journey. Sierra is trying to automate entire customer-facing operations.

Q14Who is shipping faster right now?

Decagon is releasing useful operator tools more frequently, while Sierra is making fewer, larger bets.

During the first half of 2026, Decagon introduced Proactive Agents, user memory, outbound voice, Agent Workbench, Duet, automatic optimization, root-cause analysis, QA Hub and Duet Autopilot. It also upgraded its simulation system to generate and maintain tests from production conversations.

These releases fit together. Each one reduces the work required to build, evaluate or improve an agent.

Sierra’s major launches have included Ghostwriter, new context-engineering and voice infrastructure, τ-voice and Horizon. Ghostwriter changes how an agent gets built. Horizon changes how long it can work and what kind of process it can own.

Decagon’s recent pace is more useful to a support team choosing software now. Sierra’s launches have greater potential to reshape the company over the next few years.

Q15Which company has the stronger moat?

Sierra currently has the stronger long-term moat, mainly because it is becoming deeply connected to high-value enterprise workflows.

Both companies use several third-party foundation models. A small improvement in raw language quality will therefore be difficult to protect for long. Competitors can often access the same models within weeks or months.

Defensibility comes from integrations, evaluation data, company policies, customer memory, security approvals and the cost of moving workflows elsewhere.

Decagon becomes harder to replace once a company has encoded hundreds of AOPs, connected its internal systems, created large test libraries and trained its support team on the platform. Its monitoring tools should also improve as they process more conversations.

Sierra’s integrations can reach further into the business. Mortgage origination, healthcare administration and insurance claims involve more systems, more approvals and more sensitive data than a typical support conversation. Replacing that infrastructure would be painful.

Its larger base of regulated enterprise customers also gives Sierra more experience with difficult implementations. Decagon may remain the better daily product for support operators, but Sierra’s deployments are creating heavier switching costs.

Q16Does Sierra’s funding advantage really matter?

Sierra’s larger cash reserve gives it more room to pursue expensive enterprise contracts and tolerate slow deployments.

The company has raised around $1.6 billion and reportedly had more than $1 billion of available capital after its latest financing. It can fund research, international offices, enterprise sales, voice infrastructure and large implementation teams at the same time.

Decagon is also extremely well financed. Its $250 million Series D brought total funding to roughly $481 million, which is enough to support aggressive expansion for several years.

The gap becomes relevant when a contract covers several countries, strict compliance requirements and dozens of integrations. Sierra can place more specialists on those projects and wait longer for them to become profitable.

Outcome-based pricing also exposes Sierra to more delivery risk because it may receive little or no payment when the agent fails to complete the agreed task. Its balance sheet makes that model easier to sustain.

Capital alone would not justify a lead. In Sierra’s case, the money sits alongside rapid revenue growth and a strong enterprise customer base, so it amplifies traction that is already visible.

Q17Are investors expecting too much?

Both companies are priced for extraordinary growth, and Decagon may carry the more demanding valuation relative to its last known revenue estimate.

Sierra’s $15.8 billion valuation equals roughly 105 times the $150 million ARR figure it disclosed earlier in 2026. Its revenue has probably increased since then, which would bring the current multiple down.

Decagon’s last independent revenue estimate was approximately $35 million. Comparing that figure with its $4.5 billion valuation produces a multiple near 129 times annualized revenue. Decagon likely grew between the estimate and the financing, so the real multiple should be lower.

The comparison is still useful: investors expect both companies to become major enterprise platforms, not specialized customer-service tools.

Sierra now needs to prove that Horizon can generate meaningful revenue from lending, healthcare, insurance and other long-running processes. Decagon must show that its rapidly expanding customer base can produce much larger contracts and sustained account growth.

Sierra carries the larger absolute valuation. Decagon has more scaling left to justify relative to the revenue figure we can currently observe.

Q18Are Sierra and Decagon leading the wider market?

Sierra appears to be the largest pure AI-native customer-agent company, though incumbents still have far stronger distribution.

Salesforce said Agentforce had reached approximately $800 million in ARR by January 2026, with 29,000 cumulative deals. That figure covers more than customer support, but it shows how quickly AI agents can spread through an existing CRM base.

Intercom’s Fin agent crossed $100 million in ARR, while the wider company passed $400 million. Zendesk is also adding advanced AI-agent capabilities across its installed customer base and strengthened its position by acquiring Forethought.

Sierra’s disclosed ARR places it above Fin and well ahead of Decagon’s last estimate among independent AI-native specialists. The broader market remains much larger than this two-company race.

The main threat to both startups is distribution. Salesforce, Zendesk and Intercom already control customer records, ticketing systems, workflows and enterprise buying relationships. Sierra and Decagon need to convince companies that the agent deserves its own strategic platform rather than becoming another feature inside existing support software.

Sierra has made more progress on that argument so far.

Q19Who is winning Sierra vs Decagon right now?

Sierra is clearly winning today, even though Decagon may have the better operating product for some customer-support teams.

The commercial gap decides the verdict. Sierra has several times more revenue, far more capital and a deeper position among the largest regulated enterprises. Its growth remained extremely fast after it crossed $100 million ARR, and it is already expanding into business processes worth much more than ordinary support tickets.

Decagon deserves a sharper conclusion than “promising challenger.” Its deployment speed, operator controls and customer outcomes are excellent. It has won difficult enterprise evaluations and shown that agents can lower costs, resolve complex requests and generate sales.

Those strengths have not yet translated into comparable scale. Sierra controls the bigger accounts, has more resources to expand globally and is building agents that can remain active across multi-week workflows.

Three criteria decide this: commercial scale, enterprise depth and expansion beyond support. Decagon’s product advantage matters, but Sierra has already cleared the harder company-building thresholds.

Sierra vs Decagon: the scorecard

Decagon can change the answer by showing a revenue base much closer to Sierra’s, expanding its largest accounts and producing repeated competitive wins where companies replace Sierra across several channels.

Sierra can lock in the lead by proving that Horizon works in production. Real loan originations, healthcare authorizations and insurance processes completed at scale would show that its move beyond support is already generating measurable value.

Right now, Sierra is winning the race to become the strategic enterprise platform for customer-facing agents. Decagon is closer to building the best operating system for teams that manage AI support every day. That second race matters, but the first is bigger today.