The scene, todayThe ticket that never reached a human.
It is 9:47 a.m. somewhere inside a Nestle office, and an employee is asking for a payslip correction. There is no email thread. No queue. No three-day wait. A small chat window answers, pulls the record, files the ticket against the right ledger code, updates Workday, and closes itself. The employee says "thanks" out of habit. The thing on the other end is not a person. It is Leena AI - and across roughly 500 enterprises in more than 100 countries, this is what an ordinary Tuesday now looks like.
Leena AI is, in plain language, an autonomous AI agent for the back office. It handles the small, dreary, expensive interruptions of corporate life - the HR forms, the password resets, the expense queries, the policy questions nobody can ever find the answer to. It does not pretend to be a human. It just resolves the thing and moves on.
"We guarantee a 70% self-service ratio. In writing."Leena AI sales contract clause, paraphrased
Caption: an unusual commitment for a category whose entire history is built on hedging.
The problem they sawThe back office was a graveyard for time.
Software has spent two decades fixing the customer-facing parts of a company. Stripe fixed checkout. Salesforce fixed sales. Zendesk fixed support. The inside of the company - the employee-facing part - somehow got skipped. People still email "HR" like it is 2009. People still file IT tickets and wait. People still ask the same five questions about the same five policies, every single day.
The founders called this, with some affection, the "Siri for employees" problem. Everyone has a voice assistant in their pocket. Almost no one has one at work. The cost of that gap, when you actually measure it - hours per employee, per month, multiplied by every employee in a Fortune 500 - is not a rounding error. It is a budget line item that nobody has the courage to name.
If software ate the world, agents are eating the org chart. Leena AI is hungry.YesPress, on the state of agentic AI
Caption: a category nobody markets, because every company secretly hopes you haven't noticed how bad it is.
The founders' betThree roommates, one stubborn thesis.
Adit Jain, Mayank Goyal and Anand Prajapati met long before they had a company. They met as friends - the kind of friendship that, several years and one Y Combinator batch later, would survive a pivot, a pandemic, and the awkward middle period when "AI startup" stopped being a brag and started being a cliche. Leena AI was admitted to Y Combinator's Winter 2018 cohort. The original product was, charitably, an HR chatbot. The thesis was more ambitious than the product: that the entire internal employee experience would eventually run on conversational AI, and that whoever built the rails would own a category nobody had named yet.
The bet, in retrospect, was timing. They were a chatbot company in a chatbot world, which would have been a slow death sentence, except they kept building - past chat into workflow, past workflow into knowledge, past knowledge into autonomous agents. By the time the GPT moment arrived in 2022, Leena AI had seven years of enterprise plumbing already in place. They were not retrofitting LLMs onto a help desk. They had already wired the help desk.
"Leena AI is building the Siri for employees."VentureBeat, 2021 - on the $30M Series B
Caption: the funny thing about being early is that nobody believes you until you are obvious.
The productWorkLM and the brain behind the brain.
The thing Leena AI built, technically, is two things stacked. The first is WorkLM - a proprietary large language model fine-tuned on more than 10 million multi-application enterprise prompts. It is not GPT in a sweater. It is a model that has been steeped in the dialects of SAP, ServiceNow, Workday, Salesforce, Oracle and Microsoft 365 until it knows the grammar of an enterprise ticket the way a London cabbie knows the streets.
The second is the agent. Leena AI's autonomous agent breaks down a request into sub-tasks, dispatches each one to a specialist application agent, and coordinates the whole performance through a central super-agent the company calls, with very little drama, the "Leena AI Brain." If that sounds suspiciously close to how a real team works - one person triages, one person executes, one person follows up - that is the point.
Caption: numbers from public filings, customer pages and partner marketplaces. Approximate, but defensibly so.
A short, slightly opinionated timeline.
2018
Founded. Joins Y Combinator (Winter 2018). Raises a $2M seed. Calls itself an HR chatbot, mostly because that is what investors will fund.
2020
Closes an $8M Series A led by Greycroft. Quietly pivots from chatbot to workflow platform while nobody is watching.
2021
Raises a $30M Series B led by Bessemer Venture Partners. Total funding crosses $40M.
2023
Announces WorkLM, its proprietary enterprise LLM. Stops calling itself a chatbot.
2024
Autonomous Agent goes general availability. 100+ languages. Listed on AWS, Azure, SAP and UKG marketplaces.
2025
Voice-enabled agentic AI ships in July. AI Colleague Studio launches in October, letting customers build their own agents.
Where Leena AI's agent does its quiet work.
// share of resolved enterprise requests by function (illustrative, per public case studies)
Source: aggregated from Leena AI public customer case studies and G2 reviews. Numbers vary by deployment.
The proofNames you recognize, doing things they will not brag about.
The customer list reads like a slightly weary global stock index. Nestle. Coca-Cola. Mercedes-Benz. Sony. Etihad Airways. AbInBev. Puma. These are companies that do not, as a rule, hand their internal support stack to a startup. They moved because the math worked. A contracted 70% self-service ratio is not a marketing line - it is an SLA, and SLAs are how grown-up companies actually buy software.
The partnership shelf is similarly serious. Leena AI ships through the SAP Store, Microsoft Azure Marketplace, AWS Marketplace and UKG Marketplace, with deep integrations into ServiceNow, Workday and Salesforce. The investors - Bessemer, Greycroft, B Capital, Y Combinator, Elevation - are not the kind who fund moonshots without a path to revenue. Leena AI is, by most reasonable measures, a real software company, which in 2026 is increasingly the more interesting story.
"Enterprise AI without the demo magic: it actually closes tickets in production."Recurring sentiment, G2 reviews
Caption: the highest compliment any enterprise tool can receive is "it works on a Wednesday."
The missionEffortless work, fewer interruptions, more hours back.
If you ask Adit Jain what Leena AI is for, the answer is consistent and a little understated. It is for making work feel less like work. Not by removing humans, which is the easy applause line, but by removing the friction that fills the gap between humans and the systems they have to navigate. The HR portal that needs a manual. The IT request that needs three approvals. The expense receipt that needs a screenshot of a screenshot.
That mission is unfashionable, in a way. It is not building general intelligence. It is not curing anything. It is, instead, the deeply practical project of reclaiming the small, stolen hours of the working week. Which, multiplied by the global workforce, is a very large number indeed.
What people actually do with Leena AI.
Reset passwordsFile expense claimsUpdate payroll details
Onboard new hiresResolve IT ticketsAnswer policy questions
Run leave workflowsPull HR documentsSurface knowledge
Manage SLAsTrack sentimentBuild custom AI colleagues
Why it matters tomorrowThe agent that learns the company.
The bet under the bet is this: in five years, every enterprise will have an internal AI colleague that knows the company better than most of its employees. It will know the policies. It will know the systems. It will know which approval chain is genuinely necessary and which one is just political theater. Whoever builds that colleague first - and gets it right - will sit somewhere very close to the heart of how work gets done.
Leena AI is not alone in chasing that future. ServiceNow has Now Assist. Glean and Moveworks are in adjacent rooms. The category will be loud before it is settled. What Leena AI has, that not everyone else does, is seven years of unglamorous enterprise plumbing - integrations, schemas, edge cases, language packs, contractual self-service guarantees - that you cannot fake your way through with a clever demo.
A chatbot answers. An agent finishes the job. That is the difference Leena AI is selling, and a growing list of CIOs is buying.YesPress, closing argument
Back to the scene9:47 a.m., somewhere inside a Nestle office.
The employee closes the chat. The payslip is corrected. The ledger is updated. The clock has moved forward by less than a minute. Somewhere in a data center, a model named WorkLM has logged the interaction and gotten slightly better at the next one. Somewhere in San Francisco, a small team of engineers is shipping the next sub-agent. And the queue, that long, mournful queue of internal tickets that used to define a Monday morning, is shorter than it was an hour ago.
Leena AI did not invent the help desk. It just made it answer faster than the question.
