Who are the top conversational AI startups by revenue today?

The ranking of top startups in the conversational AI market by revenue

Below is a table ranking all the companies in this market by their current revenue scale. You can find our methodology at the end of this page.

If you want a deeper understanding of the market and its current dynamics, get our report covering the Conversational AI Market.

What are the top conversational AI startups by revenue right now?

The biggest conversational AI startups are already generating hundreds of millions of dollars a year, with ElevenLabs, Perplexity, Harvey and Glean forming the clearest group behind OpenAI’s much larger ChatGPT business.

There is one catch: private AI companies rarely report revenue in the same way. Some disclose ARR, others annualized revenue or run-rate, and a few figures are outside estimates. We kept those labels intact rather than pretending they are interchangeable.

OpenAI also needs special treatment. Its wider business includes subscriptions, enterprise products and APIs that go well beyond conversational AI, so using total company revenue would distort the ranking. The cleanest current figure we found is the $1B annualized revenue run-rate that OpenAI disclosed for ChatGPT Ads.

Among more focused startups, ElevenLabs has passed $500M ARR, the Financial Times reported Perplexity above $450M ARR, Harvey has crossed $400M ARR, and Glean has reached $300M ARR. Genspark, EliseAI and Sierra form another substantial group around $200M–$250M.

Several estimated figures belong in the discussion too. Sacra puts Manus around $450M in annualized revenue and OpenEvidence around $300M. We include them because the businesses appear too large to ignore, but they carry less weight than similarly sized figures disclosed directly by companies.

Rank	Startup	Best current figure	Metric	Evidence
1	OpenAI / ChatGPT Ads	$1B	Revenue run-rate	Company disclosed
2	ElevenLabs	>$500M	ARR	Company disclosed
3	Perplexity	>$450M	ARR	Financial Times reported
4	Manus	~$450M	Annualized revenue	Third-party estimate
5	Harvey	>$400M	ARR	Founder disclosed
6	Glean	$300M	ARR	Company disclosed
7	OpenEvidence	~$300M	Annualized revenue	Third-party estimate
8	Genspark	~$250M	ARR	Company-reported figure
9	EliseAI	>$200M	ARR	Company disclosed
10	Sierra	$200M	ARR	Company disclosed
11	Uniphore	>$200M	ARR	CEO disclosed
12	Legora	~$150M	ARR	Third-party estimate
13	Deepgram	>$100M	ARR	Reported company milestone
14	Cresta	>$100M	ARR	Company disclosed
15	CHAI	>$100M	Annual revenue	Company disclosed
16	Speak	$100M	ARR	Company disclosed
17	Decagon	~$100M	Annualized revenue	Third-party estimate
18	Sanas	$60M	ARR	Company disclosed
19	Parloa	>$50M	ARR	Company disclosed
20	Rogo	~$53M	ARR	Third-party estimate

Is OpenAI really the largest conversational AI startup by revenue?

OpenAI is clearly the biggest company in conversational AI, although its total revenue is too broad to compare directly with specialist conversational AI startups.

The cleaner number is ChatGPT Ads. OpenAI says that advertising reached a $1B annualized revenue run-rate in fewer than 200 days, with tens of thousands of advertisers already using the platform.

That figure covers only one ChatGPT monetization stream. It leaves out consumer subscriptions, business subscriptions, enterprise contracts and API usage. The full economic scale of ChatGPT is much larger, but we cannot separate every part cleanly enough for a fair startup ranking.

A $1B run-rate also does not mean that OpenAI has already booked $1B of annual advertising revenue. It describes the pace of the business at that point in time.

Even with those limitations, ChatGPT Ads alone currently sits well above the disclosed commercial scale of most standalone conversational AI companies.

Is ElevenLabs the biggest pure conversational AI startup?

ElevenLabs currently has the strongest case for being the largest independent pure-play conversational AI startup, with more than $500M in ARR.

The company ended 2025 at $350M ARR and said in May 2026 that it had already moved past $500M during the first four months of the year. Growing by more than $150M from such a large base in a few months is far more meaningful than posting triple-digit growth from a tiny starting point.

Voice agents are now a major part of that expansion. ElevenLabs specifically points to companies using its technology for customer support, sales, hiring and marketing operations.

That makes the company harder to describe as simply a text-to-speech provider these days. ElevenLabs still sells voice infrastructure, but more of that infrastructure is being used inside systems that conduct complete conversations and take actions for businesses.

Its evidence is also unusually clean for a private AI company: the ARR number comes directly from ElevenLabs and is very recent.

Is Perplexity really making more than $450M a year?

Perplexity has become a $450M-plus ARR business, putting the AI search and assistant company surprisingly close to ElevenLabs in commercial scale.

The Financial Times reported that Perplexity moved above $450M ARR after monthly revenue increased about 50% in March 2026. The company now combines consumer subscriptions, enterprise subscriptions and usage-based products.

Perplexity has also been pushing much further into agents. Its Computer product and related tools let users ask the system to carry out work rather than simply return search-style answers.

More than 100 million people use Perplexity monthly according to the same reporting, but the more interesting number for this ranking is how much revenue the company is extracting from that audience and its growing enterprise customer base.

Perplexity still does not publish audited financial statements, so we would place slightly less confidence in its number than in public-company revenue. The Financial Times reporting is strong enough, however, that we treat the $450M-plus ARR figure as credible current evidence rather than a loose market estimate.

How big is Harvey today?

Harvey has crossed $400M ARR, making the legal AI startup one of the largest specialist AI application companies in the world.

Founder Winston Weinberg disclosed that Harvey had passed $400M ARR while serving more than 3,000 customers. He also said that the company now works with 80% of the Am Law 100, 20% of the Fortune 500 and half of the Fortune 10.

Those customer numbers explain how legal AI can become so large without a mass-market audience. Harvey sells into organizations where lawyer time is expensive, contract values can be substantial and the product touches research, drafting, litigation, transactions and compliance.

The business has moved very quickly. Earlier in 2026, Harvey said it added more than $100M of net-new ARR in a single quarter.

Legal AI looks much less niche financially than the size of its user population would suggest. A product used by tens of thousands of professionals can generate the kind of revenue normally associated with software serving millions of people.

How did Glean reach $300M ARR so quickly?

Glean has reached $300M ARR only 15 months after crossing $100M, which puts the enterprise AI assistant firmly among the largest companies in this market.

Glean says its Fortune 500 customer count nearly doubled year over year, while more than 85% of customers now use the product across at least five departments.

That second number is particularly useful. Enterprise AI products often begin as isolated experiments inside one team. Glean appears to be spreading across organizations after the initial purchase, which gives the company much more room to grow inside each account.

Glean also shows that conversational AI revenue is not limited to customer-facing bots. Its assistant helps employees search company information, work with internal knowledge and increasingly use agents to carry out tasks across software systems.

The product began in enterprise search, but its current economics increasingly come from becoming a conversational layer over the company’s own information and workflows.

Is voice AI becoming one of the biggest conversational AI businesses?

Voice AI is currently one of the deepest parts of conversational AI, with several independent companies already reaching meaningful eight- or nine-figure revenue scale.

As seen above, ElevenLabs is the standout at more than $500M ARR. Further down the market, Deepgram has moved beyond the $100M level, Sanas says it has grown from zero to $60M ARR since going to market in 2023, and Parloa passed $50M ARR at the end of 2025.

The interesting part is the variety of business models. ElevenLabs sells voice technology and agents across many industries. Deepgram provides voice infrastructure. Sanas works on real-time speech. Parloa builds customer-service agents. Other startups focus specifically on restaurants, healthcare, recruiting, insurance or sales.

Voice works well commercially because businesses already spend huge amounts of money on spoken interactions. Contact centers, medical scheduling, sales calls, collections and restaurant ordering already exist at scale; companies do not need to invent a completely new workflow before buying AI.

Sanas gives us a good example of how fast this can move. The company recently said it had reached $60M ARR and was on track to exceed $120M. That is still a forward-looking target rather than current revenue, so we do not use the higher number in the ranking.

Which customer-service AI startups make the most money?

Sierra and EliseAI currently lead the customer-service and operational-agent group at around $200M ARR or more, while Cresta and Decagon have reached roughly the $100M level.

Sierra says it hit $100M ARR in seven quarters and $200M in nine. EliseAI has passed $200M ARR after five consecutive years of roughly 100% annual growth across housing and healthcare.

Cresta disclosed that it had moved above $100M ARR, while Sacra estimates Decagon around $100M in annualized revenue. Parloa is smaller for now, but its directly disclosed ARR has already passed $50M.

The cluster is more interesting than any individual company. We now have several unrelated startups selling AI into customer conversations and reaching tens or hundreds of millions of dollars in recurring revenue.

Customer service now has much stronger commercial proof than categories where nearly all the revenue sits with one exceptional company.

Company	Current commercial figure	Main use case	Evidence
EliseAI	>$200M ARR	Housing and healthcare operations	Company disclosed
Sierra	$200M ARR	Customer-facing AI agents	Company disclosed
Cresta	>$100M ARR	Contact-center AI	Company disclosed
Decagon	~$100M annualized revenue	Customer-support agents	Third-party estimate
Parloa	>$50M ARR	Enterprise customer-service agents	Company disclosed
Maven AGI	$7M 2024 revenue	Customer-service agents	Company disclosed
Replicant	Revenue undisclosed	Contact-center voice automation	Company scale metrics disclosed

Is Sierra really growing that fast?

Sierra is currently one of the fastest-growing enterprise conversational AI startups, doubling from $100M to $200M ARR in only two quarters.

As pointed out above, the company needed seven quarters to reach its first $100M and only two more to reach the second. That is acceleration at a scale where many software companies begin slowing down.

Sierra also says it now works with more than 40% of the Fortune 50, one-third of leading banks and five of the world’s ten largest healthcare companies.

Those are unusually large customers, which changes the math. Sierra does not need millions of paying users if a relatively small number of global companies run high-volume customer interactions through its agents.

The company is also pushing beyond one-off support questions toward longer tasks linked to customer outcomes and revenue. If that model keeps working, contract sizes can expand without Sierra having to chase a completely new customer for every extra dollar of ARR.

Which legal AI startup makes the most revenue?

Harvey is currently the revenue leader in legal AI, while Legora appears to be the closest specialist challenger.

As seen above, Harvey has passed $400M ARR. Legora’s current figure is less certain: third-party research puts the Swedish startup around $150M ARR, following very rapid growth from a much smaller base.

The gap is still large enough that Harvey sits comfortably ahead on the evidence we have today.

Legal AI now supports more than one company at nine-figure scale. That would have sounded aggressive only a few years ago for a software category serving lawyers.

Law firms can buy access to general models such as ChatGPT or Claude. Many are still paying separately for legal products because those products connect the models to firm documents, permissions, research sources and actual legal workflows.

That extra layer is proving valuable enough to build very large businesses.

Which healthcare conversational AI startups make the most money?

OpenEvidence and EliseAI currently show the most commercial scale among healthcare-focused conversational AI startups, although their business models are very different.

Sacra estimates OpenEvidence at roughly $300M in annualized revenue. The figure is not company-filed revenue, so we treat it more cautiously, but the direction is hard to dismiss: the estimate was around $150M at the end of 2025.

OpenEvidence is unusual because clinicians can use much of the product for free. Revenue comes heavily from commercial and advertising relationships around a very large physician audience rather than standard per-seat software alone.

EliseAI takes a more conventional enterprise route. Its agents automate conversations and operational work across housing and healthcare, and the company directly reports more than $200M ARR.

Below those leaders, the market is much younger. Companies such as Hippocratic AI, Assort Health and several healthcare voice-agent startups have strong deployment stories but much thinner public revenue evidence.

For now, healthcare conversational AI looks commercially real, although most of the visible revenue is concentrated in a small number of companies.

Are consumer conversational AI startups actually making serious money?

Consumer conversational AI can produce very large businesses, but revenue drops off quickly after a handful of winners.

As we saw previously, Perplexity has already moved above $450M ARR. CHAI says it now generates more than $100M a year, and Speak has crossed $100M ARR with its conversational language-learning product.

CHAI is especially interesting because the team says millions of people use the service every day while the company remains relatively small operationally. That is a very different cost and distribution model from enterprise software.

Speak monetizes another type of consumer conversation: language practice. The company publicly celebrated crossing $100M ARR after years of focusing on AI-driven spoken learning.

Character.AI shows the other side of the market. Its cultural reach and usage have been huge, but the strongest current revenue estimates we found are far below the leaders above.

A popular conversational product can have enormous engagement without turning that engagement into comparable revenue. Consumer scale helps, but pricing, retention and willingness to pay still decide whether that audience becomes a large business.

Where does Anthropic rank among conversational AI startups?

Anthropic is clearly one of the largest companies behind a conversational AI product, but Claude cannot be given a clean revenue rank from public information today.

Anthropic generates money from Claude subscriptions, enterprise deployments and model/API usage. Public revenue reporting generally combines those activities.

Using all of Anthropic’s revenue as “Claude revenue” would therefore credit the conversational assistant with spending that may come from developers calling Anthropic models inside completely different products.

The same perimeter problem appears with other foundation-model companies, which is why we isolate ChatGPT Ads for OpenAI rather than dropping OpenAI’s whole top line into the table.

Claude belongs among the most commercially important conversational AI products. The problem is measurement, not lack of scale.

Until Anthropic breaks out assistant revenue more clearly, placing Claude above or below companies such as ElevenLabs or Perplexity would create false precision.

Can we really compare ARR, revenue and run-rate in one conversational AI ranking?

ARR, recognized revenue and run-rate are useful together for judging conversational AI scale, but they should never be read as if they were the same accounting number.

ARR tells us the annual value of recurring business at a point in time. A startup can finish a year at $200M ARR while recognizing substantially less than $200M of revenue during that year.

A revenue run-rate can be even more sensitive to timing. It often takes the latest pace of business and expresses that pace over twelve months. That is useful for a company growing extremely quickly, but it does not tell us how much revenue has already been recognized.

Fiscal-year revenue is cleaner because it measures what the company actually recognized over a completed period. The drawback is that it can lag badly in a market where a startup may double within months.

That is why we rank with judgment rather than a mechanical sort. A recent, directly disclosed ARR figure can tell us more about a fast-growing private company today than an audited number from two years ago, while an audited revenue figure deserves more confidence than a similarly sized outside estimate.

Metric	What it tells us	Main limitation
Fiscal-year revenue	Revenue actually recognized during a completed year	Can become stale quickly
TTM revenue	Revenue recognized over the latest twelve months	Rarely disclosed by private startups
ARR	Current recurring contract base expressed annually	Not the same as recognized revenue
Revenue run-rate	Current revenue pace annualized	Sensitive to the chosen month or period
Annualized revenue estimate	Outside estimate of current yearly pace	Depends heavily on source methodology
Quarterly revenue	Revenue recognized during one quarter	Seasonal and incomplete
Users or customers	Adoption and commercial reach	Cannot be converted into revenue safely

How concentrated is conversational AI revenue at the top?

Conversational AI revenue is already quite concentrated, with a small group of startups accounting for several billion dollars of annualized commercial scale.

Taking the five strongest large disclosed or credibly reported figures in our ranking gives at least about $2.65B of combined annualized scale before adding Manus, OpenEvidence or any Claude-specific revenue.

This is only a directional aggregate. It combines different metrics, so we would never present the total as audited market revenue.

The concentration still tells us something useful. There is a large financial gap between the first group of companies and the long tail of conversational AI vendors still operating below $50M.

At the same time, the market is deeper than one blockbuster. Several companies now sit above $200M, and the list around $100M is getting crowded.

Which conversational AI startups are growing fastest right now?

Genspark, Sierra, Glean, ElevenLabs and EliseAI stand out because they have been adding very large amounts of revenue after already reaching meaningful scale.

Genspark’s reported trajectory is the most extreme. Public reporting around the company puts it near $250M ARR after the Super Agent business ramped from almost nothing in roughly a year. We would still treat that figure with more caution than a formal filing.

As seen above, Glean went from $100M to $300M ARR in 15 months. ElevenLabs moved from $350M at the end of 2025 to more than $500M during the first four months of 2026.

Sierra doubled from $100M to $200M in two quarters, while EliseAI says it has doubled ARR for five consecutive years.

These are stronger growth signals than a young startup moving from $1M to $5M. Every company here has already reached a level where adding another $50M or $100M means winning a substantial amount of real business.

Startup	Earlier level	Latest level	Approximate interval
Genspark	Near-zero early agent revenue	~$250M ARR	Roughly one year
Glean	$100M ARR	$300M ARR	15 months
ElevenLabs	$350M ARR	>$500M ARR	About four months
Sierra	$100M ARR	$200M ARR	Two quarters
EliseAI	Roughly half its current ARR	>$200M ARR	About one year

Does enterprise conversational AI make more money than consumer AI?

Enterprise conversational AI currently produces more large independent companies, while consumer AI produces a smaller number of enormous winners.

Harvey, Glean, EliseAI, Sierra and Cresta are all substantial enterprise businesses. Voice specialists such as Sanas and Parloa add another layer of companies selling directly into corporate budgets.

Consumer AI is much more uneven. Perplexity, CHAI and Speak have built large revenue streams, while ChatGPT sits in a category of its own because of its enormous reach and broader business model.

The difference comes down partly to pricing. An enterprise does not need millions of employees to justify a large contract if an AI product replaces expensive labor, handles customer volume or improves a high-value professional workflow.

Consumer startups can scale distribution much faster, but getting a large percentage of users to pay meaningful amounts is harder.

Right now, enterprise conversational AI looks like the more repeatable path to $50M–$200M businesses. Consumer AI can go much larger, but there are fewer proven examples.

Are AI agents replacing traditional chatbots as the big revenue opportunity?

AI agents are increasingly where the largest conversational AI startups are finding growth because customers want systems that can finish work, not simply answer a question.

Sierra handles customer interactions and outcomes. Harvey increasingly works across complete legal workflows. Glean combines enterprise knowledge with actions. Cresta and Decagon automate more of the support process rather than stopping after generating suggested text.

Even Perplexity has been pushing beyond search-style answers into products that can carry out tasks.

A chatbot that answers a question may save a few minutes. An agent that resolves the case, updates the relevant system and completes the next step can replace a much larger amount of work.

That gives vendors more room to charge based on outcomes, resolutions, usage or the value of the process being automated.

“Conversational AI” is therefore becoming a broader economic category than chatbots. Conversation remains the interface, while more of the revenue comes from the work executed behind it.

How many conversational AI startups have crossed $100M?

At least a dozen private conversational AI companies now have credible evidence around or above $100M in revenue, ARR or annualized revenue, with several more crossing the line when we include third-party estimates.

The strongest group includes ElevenLabs, Perplexity, Harvey, Glean, Genspark, EliseAI, Sierra, Uniphore, Deepgram, Cresta, CHAI and Speak.

Estimated figures would also bring Manus, OpenEvidence, Legora and Decagon into the nine-figure group.

The exact count depends on how strictly we define conversational AI and how much confidence we require from an estimated figure. That uncertainty should stay visible.

Reaching $100M is no longer an isolated achievement in conversational AI. There is now a real cohort of startups at that scale, and several have continued rapidly toward $200M, $300M or more.

Which conversational AI revenue numbers should we trust most?

The conversational AI figures we trust most are recent numbers disclosed directly by companies or executives and clearly identified as revenue, ARR or another specific financial metric.

ElevenLabs, Glean, EliseAI, Sierra, Cresta, CHAI, Sanas and Parloa all provide relatively clean examples. The companies tell us what the metric is and give enough context to understand the period it represents.

Perplexity also ranks highly in confidence because the Financial Times reported the current ARR figure directly, even though the company has not published audited accounts.

Estimates from firms such as Sacra are useful for companies where no direct number exists. They let us see that OpenEvidence or Manus may belong near the top of the market, but we would not treat an estimated $300M exactly like a company-disclosed $300M.

Old figures deserve another kind of caution. Kore.ai, for example, disclosed more than $100M ARR in early 2024. That tells us the company had already achieved serious commercial scale, but it is now too old to use confidently against businesses reporting fresh numbers.

The best ranking does not simply sort the biggest numbers. It also asks who said the number, what the number measures and how current it still is.

Are new AI startups already bigger than the old conversational AI companies?

Several new conversational AI startups now report annualized commercial scale comparable with or above established public conversational-software companies.

Cerence reported $251.8M of fiscal-year 2025 revenue. LivePerson reported $243.7M for 2025. SoundHound AI reported $168.9M, almost double the previous year.

Those figures are real recognized revenue and come from public financial reporting, so they are cleaner than private-startup ARR.

Yet several much younger companies now sit above those levels on current annualized metrics. That is a remarkable change from the previous conversational AI cycle, when the established vendors had the revenue and newer generative-AI companies mostly had usage and venture funding.

The comparison still needs care. A startup ending the period at $300M ARR has not necessarily recognized $300M during the year, while a listed company reporting $250M of revenue actually booked that amount.

Even after allowing for that difference, newer AI-native companies have clearly reached the same commercial order of magnitude.

What do the biggest conversational AI startups have in common?

The highest-revenue conversational AI startups usually connect conversation to something customers already value enough to pay for.

Harvey connects AI to legal work. Glean connects it to company knowledge and internal workflows. Sierra and Cresta connect it to customer operations. ElevenLabs connects voice to support, sales and communications. Speak connects conversation to language learning.

The strongest consumer exceptions can monetize a huge audience directly through subscriptions, usage or advertising.

This pattern explains why user count alone tells us so little. A specialized product with a few thousand large customers can earn more than a famous consumer application with millions of users.

These days, the conversation itself is rarely the most valuable part of the product. The value comes from the problem sitting behind that conversation: resolving a support case, completing legal work, finding internal information, booking an appointment, learning a language or completing a transaction.

So which conversational AI startups are really the biggest by revenue?

The clearest leaders today are ElevenLabs, Perplexity, Harvey and Glean, while OpenAI remains much larger if we include the directly attributable ChatGPT business.

The next tier is already substantial. Genspark, EliseAI and Sierra have reached roughly the $200M–$250M range, while a growing group including Cresta, CHAI and Speak sits around or above $100M.

Manus and OpenEvidence could belong very high in the ranking too, but their strongest current figures are third-party estimates, so we would not give them the same confidence as a directly disclosed ARR number.

Conversational AI has already produced a real class of nine-figure startups. This market is no longer being measured mainly through users, funding rounds or impressive demos.

Revenue is now large enough to show where companies are actually getting paid, and the strongest businesses currently tend to sit where conversational interfaces meet expensive workflows, enterprise operations or unusually strong consumer monetization.