Customer service agents: which startup is ahead?
In our agentic AI market deck, you will find everything you need to understand the market
SUMMARY
Sierra is currently ahead in customer service agents, with Decagon as the closest challenger and Parloa in a clear third position.
Sierra’s lead rests on more than funding or a high valuation. It has the largest clearly disclosed software revenue base, unusually large enterprise customers and live agents handling difficult workflows in lending, insurance, healthcare, telecom and retail.
Decagon is the only challenger with a credible path to making the race close. Its customer additions and more than threefold growth are impressive, but the company still withholds the absolute revenue figure needed for a clean comparison with Sierra.
Parloa has already separated from the rest of the field. More than $50 million in annual recurring revenue, 150% net revenue retention and a strong position in regulated voice deployments give it a real enterprise platform, particularly in Europe.
Crescendo’s scale is substantial but harder to compare. Its business exceeds $100 million in annual recurring revenue, yet software, human support and managed operations are mixed together, so it should not be treated as a straightforward software peer.
PolyAI remains the strongest specialist for large voice deployments, while Ada has the deepest historical interaction record and a surprisingly strong recent rebound. Both are serious companies, though neither currently matches Sierra’s combination of breadth and disclosed commercial scale.
Published resolution rates are a poor shortcut for ranking these startups. A 90% rate on simple inquiries may be less impressive than a lower rate on regulated, multistep work that requires identity checks, policy decisions and completed actions in company systems.
The strongest moat is forming around workflow integration, company-specific policies, evaluation systems and deployment knowledge rather than access to any one foundation model. Better models help nearly every vendor at once.
The category is already moving beyond support. Sierra is furthest ahead in sales, retention, refinancing, insurance and shopping workflows, which could turn customer-service agents from a cost-saving product into a broader customer-operations layer.
Incumbents such as Salesforce, Zendesk, NICE, Genesys and ServiceNow can slow the startups through bundling and existing contracts. For now, however, Sierra has the clearest lead, Decagon has the best chance of catching it, and Parloa is the only other startup firmly inside the leading group.
This market map, featured in our agentic AI market deck, highlights top companies and startups in the agentic AI market
Which AI customer service startups belong in this comparison?
We are comparing eight independent startups that sell AI agents specifically for customer service, customer support or contact-center work.
The field includes Sierra, Decagon, Parloa, Ada, PolyAI, Crescendo, Maven AGI and Lorikeet. Each company sells a product that can speak directly with customers, access company systems and complete at least some service tasks without a human agent.
We exclude Salesforce, ServiceNow, Zendesk, Genesys and NICE because customer service represents one part of a much larger incumbent software business. Their products still affect the competitive picture, but comparing a young AI startup with Salesforce’s entire revenue base would tell us very little.
Intercom is also outside the main ranking. Fin is one of the strongest AI support products available, but Intercom was founded in 2011 and already had a large help-desk business before generative AI agents emerged. Forethought and Cognigy would previously have qualified, though both have since been acquired by larger customer-service platforms.
We also leave out general voice infrastructure providers such as Vapi and Bland. They can power customer-service applications, but customers normally need to build much more of the final system themselves.
Funding figures below represent publicly disclosed cumulative capital. Sierra’s total rose to approximately $1.6 billion after a recent $950 million financing. Decagon has raised $481 million, Parloa more than $560 million, PolyAI more than $200 million and Ada approximately $200 million. Maven AGI has raised $78 million, Crescendo $50 million and Lorikeet approximately $49 million.
| Startup | What it sells | Main competitive strength | Disclosed funding |
|---|---|---|---|
| Sierra | Omnichannel agents for service, sales and customer operations | Overall enterprise scale | ≈$1.6B |
| Decagon | AI concierges across chat, email and voice | Rapid enterprise growth | $481M |
| Parloa | Enterprise AI agent management with strong voice capabilities | European and regulated enterprises | >$560M |
| Ada | Omnichannel AI customer-service automation | Large installed base and operating history | ≈$200M |
| PolyAI | Enterprise voice and conversational agents | Global voice deployment | >$200M |
| Crescendo | AI-native contact center combining automation and managed services | Full-service delivery model | $50M |
| Maven AGI | Autonomous support agents connected to enterprise systems | Strong resolution results | $78M |
| Lorikeet | AI concierges for complex and regulated support | Fintech and healthtech specialization | ≈$49M |
Is there a clear leader in AI customer service agents today?
Sierra is the clear overall leader in AI customer service agents today, while Decagon and Parloa are the only challengers that have started to separate from the rest.
Sierra entered its third year with more than $150 million in annual recurring revenue. It later reported that over 40% of the Fortune 50 used its platform and that its agents were powering billions of customer interactions. The same company had started with four design partners only two years earlier. Even allowing for the fact that these figures come from Sierra, revenue, customer size and usage all point in the same direction.
Decagon has yet to reveal its absolute revenue, which limits the comparison. It did disclose that more than 100 new global enterprise customers joined during its latest fiscal year, alongside a $250 million funding round at a $4.5 billion valuation. Customers now include Block, Deutsche Telekom, Avis Budget Group, Chime and Mercado Libre.
Parloa passed $50 million in annual recurring revenue and reported 150% net revenue retention. Existing customers were therefore spending about 50% more than they had a year earlier, before adding any new accounts. Parloa also signed a multimillion-dollar contract with TP and raised $350 million at a $3 billion valuation.
Sierra is first for now. Decagon has the best chance of catching it, but without a disclosed revenue figure, we cannot yet call the race close.
If you want more recent data on this point, please see our latest agentic AI market report.
As this chart shows, and as featured in our agentic AI market deck, search interest in AI agents has been rising rapidly
Who has built the biggest AI customer service business?
Sierra has built the biggest clearly comparable AI customer service software business, while Crescendo has built a similarly large but more service-heavy company.
Sierra passed $100 million in annual recurring revenue in November 2025 and exceeded $150 million shortly afterward. The contracts usually last at least one year, are often multiyear and are billed annually upfront. That makes the number more reliable than an annualized calculation based on one unusually strong month of usage.
Crescendo has now confirmed that it exceeded $100 million in annual recurring revenue in under two years. It also reports more than 500 AI deployments and operations across six continents. The comparison is a bit messy because Crescendo combines software, human support, quality management and managed operations after acquiring PartnerHero. Some of its revenue therefore resembles outsourced customer service more than pure software subscriptions.
Parloa is the clearest pure-software comparator at more than $50 million in annual recurring revenue. Sierra was operating at over three times that disclosed scale when it last reported its figure.
Ada announced more than 100% year-over-year revenue growth, 108% growth in agentic-AI recurring revenue and 146% net revenue retention. Decagon disclosed more than threefold growth in both ARR and recognized revenue. Neither revealed an absolute revenue figure, so we cannot place them precisely against Sierra, Crescendo or Parloa.
| Startup | Best disclosed commercial evidence | How we interpret it |
|---|---|---|
| Sierra | More than $150M ARR | Largest comparable software business |
| Crescendo | More than $100M ARR | Very large business, though software and managed services are mixed |
| Parloa | More than $50M ARR and 150% NRR | Proven enterprise platform with strong expansion |
| Decagon | Revenue and ARR grew more than 3× | Fast growth from an undisclosed base |
| Ada | Revenue grew more than 100%; NRR reached 146% | Strong acceleration, but absolute scale remains private |
| PolyAI | More than 200 enterprise customers | Broad adoption without current revenue disclosure |
| Maven AGI | More than 50 customers | Credible early scale |
| Lorikeet | Rapid growth in regulated sectors | Still far smaller than the leading companies |
Which AI customer service startup is growing fastest now?
Sierra is adding the most revenue in absolute dollars, while Decagon and Ada may be growing just as quickly in percentage terms.
Sierra increased its annual recurring revenue from roughly $20 million to $100 million within about one year. It then added at least another $50 million within the following quarter. A fivefold increase is already rare for enterprise software; producing it while moving through the $100 million threshold is more impressive than showing the same rate from a small starting base.
Decagon disclosed more than threefold year-over-year growth in both ARR and recognized revenue. It also added more than 100 enterprises in one fiscal year. Those customer additions suggest that the increase came from a broad sales engine rather than one oversized contract. The unanswered question is how much each new customer is spending.
Ada’s recent rebound deserves more attention than it usually receives. Revenue more than doubled, agentic-AI ARR increased by 108%, voice ARR grew twelvefold and net revenue retention reached 146%. Ada is an older company than Sierra or Decagon, so doubling at this stage carries more weight than typical early-startup growth.
Parloa’s 150% retention rate shows that existing customers are expanding rapidly. Its US business has also begun growing faster than its European business, giving the company a second large market rather than relying mainly on its original regional base.
Sierra leads in absolute growth. Decagon and Ada are the strongest percentage-growth challengers.
This chart, included in our agentic AI market deck, illustrates yearly VC funding for agentic AI startups
Which AI customer service startups have the most mature products and handle the hardest problems?
Sierra currently has the strongest mix of production maturity and complex workflow execution, while Ada, Decagon, PolyAI and Parloa also run serious systems at scale.
Sierra has placed agents inside healthcare, telecommunications, banking, insurance, lending and retail companies. Cigna moved an agent into production in eight weeks and reported an 80% reduction in patient-authentication time. Singtel launched in ten weeks and reported resolution above 70%. Nordstrom introduced a voice agent in five weeks. These are selected company examples, though the short deployment times keep appearing across very different industries.
The tasks themselves are also unusually demanding. Sierra’s agents help customers refinance homes, originate mortgages, process insurance claims, authenticate healthcare patients, change subscriptions and complete retail returns. These jobs require the agent to collect information, follow company policy, call the correct systems and finish an action safely.
Decagon has shown similar maturity in regulated financial services. Chime reports approximately 70% resolution across chat and voice, a 60% decrease in support costs and a voice system handling more than one million calls each month. Decagon’s Agent Operating Procedures allow companies to describe detailed policies in natural language while keeping important rules under tighter control.
Ada has the deepest historical usage record. The company reports more than 550 enterprise AI agents, 6.4 billion interactions and nearly one billion conversations. Cebu Pacific runs Ada across 11 languages and says the system resolves 75% of interactions autonomously while customer satisfaction has increased by 50%.
PolyAI currently works with more than 200 enterprise customers across over 25 countries and 45 languages. Its largest deployments perform work equivalent to more than 1,000 full-time employees, according to the company. That operating history makes PolyAI one of the safest choices for a buyer mainly concerned with telephone service.
Parloa has production customers such as Allianz, Swiss Life, Booking.com and HealthEquity. Its strength is particularly visible in phone-based banking, insurance and healthcare interactions involving interruptions, accents, identity checks, emotional customers and older contact-center systems.
Maven AGI now serves more than 50 customers. Clio reports 80% autonomous resolution, 60% more tickets solved and support four times faster after replacing its previous chatbot. Crescendo has crossed 500 deployments, while Lorikeet is live across phone, chat, email, SMS and WhatsApp at fintech and healthtech companies.
Sierra takes the lead because it combines mature deployments with the widest range of high-risk, multistep workflows.
Who leads voice AI for customer service?
PolyAI remains the strongest specialist in voice AI customer service, with Parloa, Sierra and Decagon now close enough to challenge it on major contracts.
PolyAI works with more than 200 enterprises across 45 languages and over 25 countries. Customers include Marriott, Caesars Entertainment, PG&E and UniCredit. The company has spent years solving telephone-specific problems such as interruptions, latency, accents, call transfers and noisy connections.
Its Agent Development Kit now lets enterprise teams build and test agents through code, Git and continuous-deployment workflows. PolyAI has also added multilingual agents that can detect and switch languages during a conversation.
Parloa is the strongest voice-first challenger for European and regulated enterprises. Swiss Life reports 96% routing accuracy and calls reaching the right destination 60% faster. Sierra and Decagon are broader omnichannel platforms, but both already support large voice deployments, including more than one million monthly calls for Decagon at Chime.
PolyAI keeps first place because it has the broadest specialist voice record. Sierra may be the better choice when voice is only one part of a larger customer-service platform.
If you want more recent data on this point, please see our latest agentic AI market report.
This chart, included in our agentic AI market deck, shows how Cognition is positioned in agentic AI
Which AI customer service startup has won the best customers?
Sierra has the strongest collection of large and demanding enterprise customers, while Decagon has assembled the most impressive challenger portfolio.
Sierra publicly names Cigna, ADT, DIRECTV, Gap, Nordstrom, Ramp, Rivian, Rocket Mortgage, SiriusXM, Singtel, SoFi and Wayfair. More than one quarter of its customers generate over $10 billion in annual revenue, and half generate over $1 billion, according to Sierra’s year-two review.
Decagon’s customer list includes Block, Chime, Deutsche Telekom, Mercado Libre, Avis Budget Group, Affirm, Rippling, Notion, Duolingo, Eventbrite and Substack. The mix covers banking, travel, telecommunications, marketplaces and software.
Parloa’s public list is smaller but still strong, with Allianz, Swiss Life, Booking.com, HealthEquity and TP. PolyAI is well established in hospitality, utilities and banking, while Ada has hundreds of enterprise deployments across airlines, telecom and technology.
Sierra remains ahead on customer quality, with Decagon closing the gap fastest.
Which AI customer service startup can support the largest global deployments and keep scaling?
Sierra has the strongest overall scaling position because it combines large enterprise deployments with by far the biggest cash reserve, while PolyAI and Ada lead on international footprint and historical interaction volume.
PolyAI operates in more than 25 countries and supports 45 languages. The company recently increased its disclosed enterprise customer count from more than 100 to more than 200, meaning its customer base roughly doubled between two successive updates. Some of its largest implementations perform work equivalent to over 1,000 full-time employees.
Ada brings a different kind of scale. Its agents have processed 6.4 billion interactions and nearly one billion conversations. One airline deployment operates in 11 languages, and the company has recently expanded further across Europe and Asia-Pacific.
Sierra reports billions of customer interactions but gives fewer exact geographic figures. It has expanded into Japan with SoftBank support, entered France and opened operations in Canada. Its customer base also covers multinational companies in telecommunications, healthcare, finance and retail.
Sierra recently raised $950 million at a valuation above $15 billion, bringing cumulative funding to approximately $1.6 billion and cash on hand above $1 billion. That gives the company room to hire implementation teams, enter new countries, subsidize difficult deployments and acquire smaller technology providers.
Decagon has raised $481 million, including $250 million in its latest round. Parloa has raised more than $560 million and closed its latest financing only seven months after the previous one. Both companies can spend heavily on enterprise sales, product engineering and international expansion.
Crescendo can deliver across six continents through its combination of software and service teams. It also appears unusually capital-efficient, having raised $50 million while exceeding $100 million in recurring revenue. The comparison flatters Crescendo because acquired and service revenue make its business different from a normal software startup.
PolyAI leads on disclosed countries and languages, Ada on cumulative interaction history and Sierra on the overall ability to finance and support very large enterprise deployments.
This chart, included in our agentic AI market deck, illustrates yearly funding for agentic AI startups
Which AI customer service startup offers customers the best economics?
PolyAI publishes the strongest structured return-on-investment evidence, while Sierra and Crescendo have built the largest businesses around outcome-based pricing.
A Forrester Consulting study commissioned by PolyAI estimated a 391% return over three years, average savings of $10.3 million and a payback period below six months for a composite customer. PolyAI funded the study, so we would not treat it as a neutral market benchmark.
Sierra usually charges for completed outcomes rather than employee seats. Crescendo also prices around measurable outcomes but combines software with managed services, making its savings harder to separate.
Decagon customer Chime reported a 60% decrease in support costs. Maven AGI customer Papaya Pay reported a 50% reduction in cost per ticket, while Clio says support became four times faster.
No public evidence lets us calculate a fair cost per correctly resolved problem across every vendor. PolyAI leads on formal ROI evidence, while Sierra has built the largest software business around outcome pricing.
If you want more recent data on this point, please see our latest agentic AI market report.
Can we trust the resolution rates published by AI customer service startups?
Published resolution rates are useful inside one deployment, but they are too inconsistent to rank AI customer service startups directly.
Decagon reports roughly 70% resolution at Chime and more than 90% at Substack. Ada says its platform can resolve over 80% of support inquiries, while Cebu Pacific reports 75% autonomous resolution. Maven AGI customer examples range from approximately 50% to above 90%. These figures can all be accurate because the underlying problems are different.
Companies also use different definitions. Some count a conversation as resolved when no human takes over. Others ask the customer whether the issue was solved, check whether the customer returns later, or verify that a business action succeeded in the underlying system.
A contained conversation may still fail quietly. The customer can leave, open another ticket or call the company later. A stricter measure should therefore combine completed actions, customer confirmation, repeat-contact rates and satisfaction.
Recent research from Nubank shows why evaluation quality changes the outcome. Researchers studying five production deployments across a customer base exceeding 100 million users found that stronger evaluation and context-engineering processes improved both development speed and production results. In one card-delivery use case, the newer system improved AI transactional NPS by 37 percentage points and self-service by 29 points.
We give more weight to repeat customer expansion, large production volumes and contracted revenue than to one isolated resolution percentage. A vendor with lower published automation may be handling much harder inquiries.
This chart, included in our agentic AI market deck, compares the main business model options for autonomous AI agent platforms
Which AI customer service startup has the strongest moat?
Sierra currently has the strongest overall moat, built from enterprise workflows, customer trust, deployment knowledge and the data produced by billions of interactions.
Most companies in this market use models supplied by OpenAI, Anthropic, Google or several providers at once. Better foundation models therefore help many competitors simultaneously.
Sierra’s stronger assets sit around the model. Its agents connect with internal systems, follow company-specific policies, complete regulated actions and improve through evaluation data from live deployments. Ghostwriter may also reduce the time needed to create and optimize new agents.
Decagon is building a similar advantage through Agent Operating Procedures, customer memory, cross-channel context and monitoring. PolyAI has years of voice data involving accents, interruptions and routing, while Parloa adds European compliance and regulated-enterprise experience. Ada brings billions of historical interactions and long customer relationships.
Sierra currently has the best combination of workflow depth, enterprise trust and operating scale.
Which startup is moving beyond customer support into sales and retention?
Sierra is currently furthest ahead in expanding AI customer service agents into sales, account growth, retention and other revenue-producing work.
Sierra’s agents now support mortgage applications, refinancing, insurance sales and claims, banking offers, healthcare administration and retail product discovery. Rocket Mortgage reported that customers using its digital assistant converted four times faster, although the company has not published the full experimental design behind that comparison.
Decagon is following the same path. Its spring product release introduced proactive agents that remember customers and act at the right moment. Hunter Douglas has expanded its deployment toward conversational commerce, while 1-800-Flowers.com is applying Decagon across proactive engagement, sales, marketing and logistics.
Crescendo already combines service and shopping in one conversation. Lovepop reported that customers using its AI shopping assistant converted at roughly four times the normal chat-to-order rate and bought larger baskets. The example remains company-reported, but it shows how customer service can directly influence revenue.
PolyAI customers use voice agents to take reservations and bookings after normal working hours. One restaurant example generated $300,000 from after-hours bookings. Ada also positions its agents across the customer journey, though most of its published operating figures still concern service resolution.
Sierra leads because it has the widest set of high-value workflows already in production. Decagon is moving in the same direction quickly and could narrow this part of the gap.
If you want more recent data on this point, please see our latest agentic AI market report.
This chart, featured in our agentic AI market deck, shows the share of revenue generated by each customer segment in the agentic AI market
Which AI customer service startup has the strongest momentum lately?
Sierra still has the strongest momentum, with Decagon, Ada and PolyAI producing the clearest challenger momentum.
Sierra has combined rapid revenue growth with a large financing, international expansion and the launch of Ghostwriter, which aims to automate part of the work required to create and improve agents.
Decagon tripled its valuation to $4.5 billion in under six months, introduced proactive agents with persistent customer memory and continued adding large customers. Its missing absolute revenue remains the main gap.
Ada returned to more than 100% revenue growth, expanded voice ARR twelvefold and reached 146% net revenue retention. PolyAI doubled its disclosed enterprise customer count to more than 200 and launched tools that let enterprise developers build agents directly.
Crescendo also confirmed more than $100 million in annual recurring revenue, expanded into Europe and appointed co-founder Andy Lee as chief executive.
Sierra keeps first place because it is accelerating from the largest base. Decagon remains the most credible challenger.
Can Salesforce, Zendesk and NICE catch the AI customer service startups?
Large customer-service platforms can slow the startups considerably, but they have not yet matched Sierra’s focus or Decagon’s speed.
Salesforce, Zendesk, NICE, Genesys and ServiceNow already control customer data, workflows and large software contracts. Their sales teams can add AI agents to an existing renewal, which is much easier than introducing a new vendor through security, legal and procurement reviews.
Acquisitions are closing some of the product gap. NICE bought Cognigy, while Zendesk acquired Forethought. These deals give incumbents modern agent technology without waiting for internal development to catch up.
The incumbents can also bundle agents at low prices. A company already paying millions for Salesforce or Zendesk may choose an adequate built-in agent over a better standalone product if switching and integration costs are high.
Startups still move faster because their entire product and pricing model were designed around autonomous work. Sierra, Decagon and Parloa can connect across several incumbent systems and have no older seat-based business to protect.
The market will probably split between bundled incumbent agents and independent platforms used for more complex automation. Sierra’s lead depends on becoming an important customer-operations layer before capable agents become a standard feature inside every large service platform.
This chart, included in our agentic AI market deck, shows how autonomous AI agent platform technology has evolved over time
Which startups are actually ahead in AI customer service agents?
Sierra is the startup ahead overall in AI customer service agents, with Decagon as the closest challenger and Parloa holding a clear third position.
Sierra wins because it combines the largest comparable recurring-revenue business, the strongest large-enterprise portfolio, difficult production workflows, broad channel coverage and enough capital to keep expanding. As seen above, its annual recurring revenue exceeded $150 million, more than three times Parloa’s last disclosed figure.
Decagon ranks second despite withholding its absolute revenue. More than threefold growth, over 100 new enterprise customers and a million-call-per-month voice deployment at Chime provide enough evidence to place it above Parloa. Decagon could challenge Sierra directly if its new customers turn into large multiyear contracts and its eventual revenue disclosure approaches Sierra’s scale.
Parloa ranks third because it already exceeds $50 million in recurring revenue, retains and expands customers exceptionally well, and holds a strong position in enterprise voice. As noted earlier, its 150% net revenue retention means the existing customer base is growing quickly even before new contracts are added.
Ada takes fourth place. It has more historical interaction volume than most of the field, over 550 enterprise agents and strong recent growth across both existing and new customers. Its main weakness is the absence of an absolute revenue figure.
PolyAI ranks fifth overall and first among voice specialists. More than 200 enterprise customers, 45 languages and large live call volumes make it a proven global platform. Its narrower voice heritage keeps it below the broader omnichannel leaders.
Crescendo ranks sixth. A business exceeding $100 million in recurring revenue cannot be dismissed, but its mix of software, human operations and acquired service revenue makes it harder to judge as a pure AI-agent platform. Its partnership with Alorica and new chief executive could push it higher if enterprise platform adoption accelerates.
Maven AGI ranks seventh. Customer results from Clio, Papaya Pay, ClickUp and others are strong, and the company now serves more than 50 customers. It needs more aggregate revenue and deployment evidence before entering the next group.
Lorikeet ranks eighth. Its focus on complex fintech and healthtech support gives it a useful position, though funding, customer scale and production history remain far below the leaders.
For now, Sierra is ahead, Decagon has the best chance of catching it, and Parloa remains the only other startup with enough revenue and enterprise depth to form part of the leading group.
| Rank | Startup | Why it ranks here |
|---|---|---|
| 1 | Sierra | Largest comparable software business, strongest major-enterprise reach and broadest high-value workflows |
| 2 | Decagon | Fastest credible challenger with rapid enterprise growth and scaled chat and voice deployments |
| 3 | Parloa | More than $50M ARR, 150% NRR and a strong enterprise voice position |
| 4 | Ada | Huge interaction history, 550+ enterprise agents and renewed triple-digit growth |
| 5 | PolyAI | Leading voice specialist with 200+ enterprises across 45 languages |
| 6 | Crescendo | More than $100M ARR and 500+ deployments, discounted because of its hybrid software-and-services model |
| 7 | Maven AGI | Strong customer results and growing production proof, but still limited company-wide scale |
| 8 | Lorikeet | Promising specialist for complex regulated support that remains much earlier than the leaders |
If you want more recent data on this point, please see our latest agentic AI market report.
OUR METHODOLOGY
This analysis tests which independent startup is currently ahead in AI customer service agents. We compare eight companies across commercial scale, growth, enterprise adoption, product maturity, workflow complexity, voice capabilities, international reach, customer economics, defensibility and recent momentum.
We reviewed the freshest relevant evidence available and assessed each dimension separately. We gave the most weight to company-wide recurring revenue, customer expansion, production volumes, difficult workflows completed in underlying systems and results repeated across several deployments.
The companies do not report performance in a standardized way. Absolute recurring revenue was therefore separated from percentage growth, specialist leadership from overall leadership, and pure software revenue from models that also include managed services or human operations.
Resolution and automation rates were treated carefully because vendors use different definitions and apply them to very different tasks. We looked beyond the highest published percentage to the difficulty of the work, deployment scale, completed actions, repeat usage and evidence that customers were expanding their contracts.
The final ranking does not depend on one decisive metric. It reflects the accumulation of evidence across the full analysis, with companies ranking higher when commercial scale, demanding production use, customer quality and expansion reinforced one another.
Key sources include Sierra’s year-two review, Sierra’s financing and adoption update, Decagon’s Series D announcement, Parloa’s revenue update, Ada’s growth disclosure, PolyAI’s enterprise scale update, the Forrester study commissioned by PolyAI, Crescendo’s ARR and expansion announcement, Maven AGI’s Clio case study, and Zendesk’s Forethought acquisition announcement.
In our agentic AI market deck, we identify pain points entrepreneurs should prioritize
Related blog posts
- Who is winning the AI phone agent race?
- Who is setting the standard for healthcare AI agents?
