Startup Deep Dive : Observe.AI — valued near $1 billion, then it cut jobs and reinvented itself twice

Before software got involved, a contact centre supervisor could listen to roughly 2% of the calls their agents took in a week. The other 98% simply vanished the moment the customer hung up. Observe.AI was built to close that gap — and by April 2022 it had raised enough money at a price that, per third-party estimates, valued the company at somewhere between $1 billion and $1.4 billion, a figure the company itself has never confirmed.

Last updated: September 2026

Less than a year later, in February 2023, the same company was cutting staff. The contradiction is the story: an “AI for call centres” business that grew fast enough to attract SoftBank, Zoom and Menlo Ventures as backers, then had to reinvent what it sold — twice — to stay relevant as the technology it was built on became a commodity.

Quick facts

What they do

Observe.AI sells software that listens to, transcribes and scores customer service conversations, then increasingly acts on them. Its customers are contact centre operators — banks, insurers, healthcare administrators, logistics firms, retailers and the business-process outsourcers who run call centres on behalf of other companies. The pitch has evolved in three stages: first, “quality assurance automation” that grades every agent call against a rubric instead of the 1-in-50 a human supervisor could realistically sample; then, generative-AI copilots that whisper real-time answers, compliance reminders and post-call summaries to human agents; and, from March 2025, “AI Agents” that can run entire customer conversations — over voice or chat — without a human on the line at all. The company’s own account is that it processes upward of five million interactions a day and serves more than 350 companies globally, spanning banking, healthcare, insurance, retail, travel and utilities.

The origin

The founding insight was simple and specific: contact centres generate a mountain of recorded conversation and use almost none of it. Swapnil Jain, an IIT Delhi graduate who had led user growth at Twitter and helped open its India office, teamed up with fellow IIT Delhi engineer Akash Singh and two-time entrepreneur Sharath Keshava Narayana — who had spent more than a decade building global sales teams at Unbxd, AWS and Akamai — to incorporate Observe.AI in May 2017. Their read on the roughly $400 billion global contact-centre market was that supervisors could listen to only a sliver of calls, so coaching, compliance and quality control were built on a rounding error of the actual conversations happening every day. The company went through Y Combinator’s Winter 2018 batch before raising its first institutional capital.

The struggle years

Observe.AI’s early years were not a straight climb. Its first outside cheque was a modest $1 million pre-seed in August 2017 from Emergent Ventures, followed by an undisclosed Y Combinator seed in March 2018 and an $8 million round that August led by Nexus Venture Partners — small sums for a company chasing a market as large as global customer service. It took until December 2019, more than two years in, to close a proper Series A of $26 million.

The sharper setback came later and had nothing to do with product-market fit. In February 2023, ten months after closing a $125 million Series C, Observe.AI cut jobs. Workforce-tracking sites and employee accounts on the professional forum Blind describe the reduction as part of a broader retrenchment across venture-backed software after the 2022 funding slowdown, with the company’s channel and go-to-market teams — expanded heavily during the pandemic-era contact-centre boom — among those affected; neither the company nor the trackers that reported it disclosed a precise headcount number. It was an uncomfortable moment for a business that had just been anointed one of the best-funded conversation-AI startups in the world.

The turning point

The turning point followed within months. On 20 June 2023, Observe.AI introduced what it billed as a purpose-built, 30-billion-parameter large language model trained specifically on contact-centre conversation data, alongside a new “Generative AI Suite” that could auto-summarise calls, surface real-time answers to agents and draft coaching notes automatically. Before that launch, the company’s business was almost entirely retrospective — grading calls that had already happened. After it, Observe.AI began selling a product that acted inside the call itself. The pivot compounded: in March 2025 the company introduced “VoiceAI Agents” under a unified platform positioning, moving from AI that assists a human agent to AI that can be the agent. By July 2026, Observe.AI and Amazon Web Services announced a partnership built around a DoorDash deployment scaling the platform’s automated quality coverage across 19,000 agents — a scale reference point that did not exist in the company’s marketing before the 2023 pivot. Estimated annual recurring revenue moved from roughly $30.1 million in 2023 to about $44.2 million in 2024, per third-party tracking, though the company has not confirmed those figures itself.

The money behind it

Observe.AI has raised $213–214 million across six rounds since 2017, according to multiple funding trackers including Crunchbase-sourced data and Inc42’s funding database. The shape of the capital tells its own story. Nexus Venture Partners backed the earliest institutional rounds and stayed on the board through Jishnu Bhattacharjee. Menlo Ventures, through partner Steve Sloane, led the $54 million Series B in September 2020 and became the company’s anchor growth investor. The $125 million Series C in April 2022 was led by SoftBank Vision Fund 2 — SoftBank’s second, more disciplined AI-and-software vehicle — and, notably, added Zoom Video Communications as a strategic investor, a signal that a company synonymous with video conferencing saw contact-centre AI as adjacent enough to bet on directly. Scale Venture Partners, via board member Andy Vitus, rounded out the syndicate.

The valuation question is where the record gets murky, and it is worth naming that plainly rather than picking a number. Observe.AI has never published a post-money valuation for its 2022 Series C. Two independent valuation trackers that model private markets put the figure differently: TechStackIPO estimates it at $1.0 billion as of the April 2022 round, while Caplight’s estimate for the same round is $1.4 billion. Both label their number an estimate, not a disclosed figure, and both sit well above PitchBook’s own tracked mark of $304 million for the company’s 2020 Series B. Read together, the range says Observe.AI was probably worth somewhere between $1 billion and $1.4 billion at its funding peak — comfortably into “unicorn” territory by most reasonable readings — but the company’s own silence on the number is itself notable for a business that has been vocal about almost everything else it has shipped.

How it makes money

Observe.AI runs a business-to-business subscription model, not a per-call marketplace. Enterprise customers pay recurring software fees typically scaled to the volume of interactions processed and the modules switched on — quality automation, real-time agent assist, or the newer autonomous AI Agents each carry their own pricing tier, layered on top of a base analytics platform. Data aggregator Getlatka has pegged the company’s average contract value at roughly $294,000, consistent with a product sold to enterprise contact centres running thousands of seats rather than small businesses. The part people tend to get wrong is treating this as a speech-to-text company: transcription is a commodity input, licensed or built cheaply, and Observe.AI’s actual margin sits in the layer above it — the scoring rubrics, compliance workflows and, since 2023, the proprietary language model tuned on contact-centre-specific conversation patterns that a generic off-the-shelf model handles less precisely. Costs run heaviest in cloud compute for real-time transcription and inference at call volume, plus the enterprise sales and customer-success headcount needed to land and expand large accounts — the same go-to-market cost centre that took the brunt of the 2023 layoffs.

The numbers

Observe.AI does not publish audited financial statements; as a private company it is under no obligation to. The figures below are third-party annual recurring revenue estimates from the data aggregator Getlatka, not company-confirmed or audited numbers, and Observe.AI has not disclosed profit or loss for any year.

Two things stand out. First, the growth rate: Getlatka’s estimates imply ARR nearly tripled from 2021 to 2024, a trajectory that would explain why investors kept writing cheques through a difficult 2022–23 funding market. Second, the gap between that growth and the layoffs of February 2023 — a company growing revenue can still cut costs if it grew headcount faster than revenue, which is exactly what the channel-team expansion described in the struggle-years section suggests happened.

Where the money comes from

Observe.AI’s revenue is not broken out by geography or segment in any public filing, but its own customer disclosures point to a business concentrated in a handful of verticals: banking and fintech, healthcare and insurance, retail, transportation and logistics, hospitality, and utilities — sectors that share large, regulated, high-volume contact centres where a compliance-monitoring feature is as valuable as the productivity gain. Publicly named customers span that spread: Pearson and Concentrix in education and BPO, SoFi and DailyPay in fintech, Signify Health in healthcare, DoorDash in logistics-adjacent consumer services, and Public Storage, Trupanion, Super-Sod and Freeman across retail and services. Operationally, the company is split across a US commercial headquarters in Redwood City, California, and a Bengaluru engineering base — typical of the India-founder, US-headquartered structure common among enterprise SaaS companies built by Indian founders for a primarily American enterprise customer base. The surprise, given the founders’ Indian origin and engineering base, is how little of the disclosed customer roster is Indian: the named case studies are almost entirely US enterprises, which suggests the actual go-to-market centre of gravity has always been American contact centres, with India serving as a cost-efficient build centre rather than a market.

The risks

Three risks sit underneath the growth story. The first is bundling: large contact-centre-as-a-service platforms — NICE CXone, Verint and Five9 among them — can fold conversation-intelligence features into their existing platform contracts at minimal incremental price, which is a structurally difficult position for a standalone vendor selling the same capability as an add-on. The second is model commoditisation: Observe.AI’s 2023 differentiation rested on a proprietary 30-billion-parameter contact-centre language model, but as general-purpose foundation models from OpenAI, Anthropic and Google have grown more capable and cheaper to fine-tune, the technical moat a purpose-built model once provided narrows every year, forcing continual reinvestment to stay ahead of what a customer could plausibly build with an off-the-shelf API. The third is data sensitivity: the product’s entire value proposition depends on ingesting recorded calls that routinely contain payment details, health information and other regulated personal data, which raises the compliance stakes of any breach or model-training misstep well above those of an ordinary SaaS tool, particularly as the company pushes further into autonomous agents that act on that data rather than merely analysing it.

The takeaway

Observe.AI’s history argues that in enterprise AI, the moat is rarely the model — it is the willingness to keep discarding the product you just built. The company was a QA-automation vendor, then a generative-AI copilot vendor, then an autonomous-agent vendor, inside roughly three years, and each shift arrived just as the previous category started looking replicable by a well-resourced platform competitor or a generic foundation model. The lesson for any company selling “AI for X” is that the technology underneath will commoditise faster than the market expects; the businesses that survive are the ones that treat their current product as a temporary position to be traded up from, not a moat to be defended indefinitely.

Frequently asked questions

What does Observe.AI actually sell?

Software for contact centres that transcribes and scores customer service calls for quality and compliance, assists human agents in real time with AI-generated prompts and summaries, and, since March 2025, autonomous AI agents that can handle full voice or chat conversations without a human agent.

Who founded Observe.AI and when?

Swapnil Jain, Akash Singh and Sharath Keshava Narayana incorporated the company in May 2017, going through Y Combinator’s Winter 2018 batch before raising institutional venture capital.

How much has Observe.AI raised, and what is it worth?

The company has raised $213–214 million across six rounds, most recently a $125 million Series C in April 2022 led by SoftBank Vision Fund 2 with Zoom, Menlo Ventures and Scale Venture Partners participating. Its valuation from that round was never officially disclosed; third-party trackers estimate it at between $1.0 billion (TechStackIPO) and $1.4 billion (Caplight).

Did Observe.AI lay off employees?

Yes. In February 2023 the company reduced staff, reported by workforce-tracking sites and corroborated by employee accounts on the professional forum Blind, though neither the company nor those sources disclosed an exact headcount or percentage affected.

Is Observe.AI profitable?

Not disclosed. As a private, venture-backed company, Observe.AI does not publish audited financial statements. Third-party estimates put its 2024 annual recurring revenue at approximately $44.2 million, up from about $30.1 million in 2023, but profit or loss figures are not public.

What is Observe.AI’s revenue?

Observe.AI is an AI-powered conversation intelligence platform for contact centres, analysing voice and text interactions to automate quality assurance, coaching and compliance across enterprises. It processes over four billion minutes of conversations annually. The company’s revenue details are covered in its annual filings and public disclosures, reflecting growth in line with the broader saas / ai opportunity in India.

What is Observe.AI’s shareholding structure?

Observe.AI’s cap table includes founders, institutional investors, strategic partners, and an employee stock option pool. The founders retain meaningful control while investors hold preference shares acquired through successive funding rounds.

Related Deep Dives

• Tekion — Startup Deep Dive
• TaxBuddy — Startup Deep Dive
• DataWeave — Startup Deep Dive
• Browse all Startup Deep Dives

Shareholding and Cap Table

Observe.AI has raised capital from a mix of institutional investors, angel investors, and strategic partners. The cap table reflects the company’s fundraising journey and strategic priorities.

• Governance: Observe.AI’s board includes independent directors and investor nominees, ensuring balanced oversight of management decisions.
• Dilution history: Each funding round diluted existing shareholders proportionally. The founders have managed dilution carefully to retain strategic control.
• Secondary sales: Select early investors and employees have exited positions through secondary transactions, providing partial liquidity.

Key Milestones

Source: Publicly available data, published reports, industry research, and other credible sources.
Figures are as of September 2026. Currency converted at $1 ≈ ₹96.0 as of 18 September 2026 (Trading Economics).
Found an error? Write to us and we’ll correct it in the open, dated, on the piece.