Adit Jain

The through-line

Adit Jain's recurring argument is that value in enterprise software comes from depth, not breadth, and he learned it the hard way. His first product, Chatteron, was a no-code platform on which anyone could build a chatbot for any purpose, and it reached 30,000 businesses while converting barely twenty of them into paying customers [2][4]. His diagnosis has not changed in the years since: "we were trying to be everything for everyone" and so "we essentially were nothing for nobody" [2]. Everything Leena AI has become follows from the correction, which was to look at the small group of customers who were paying and using the product heavily, discover they were companies with a thousand-plus employees building internal HR and IT bots, and then spend six and a half months shadowing those people to understand what they were actually trying to fix [3][4].

The ambition has scaled up while the method stayed constant. The early framing was a Siri or Jarvis for the enterprise, narrowed at first to HR because a three-person team could not take on the whole enterprise application landscape at once [3][4]. The current framing is AI colleagues that remove manual work from the back office entirely, across HR, finance, procurement and IT, so "humans can do so much more with their time" [1]. Jain's belief now is that "all large Enterprise workflows are going to be disrupted" by large language models, and that the timeline is compressed: the future he is being asked about is "not 3 years away. It's it's 6 months, 9 months away" [1][4].

On going deep rather than broad

The lesson from Chatteron is the one Jain returns to most often and the one he offers other founders. Users could build customer support bots, marketing bots and sales bots on the platform, "but at the end of the day you were able to do nothing because you weren't going deep enough" [4]. A related failure was that conversational design was still an unfamiliar craft, so businesses signing up could not build good experiences themselves and never scaled past a few hundred messages a month [4]. His prescription is blunt: "have a very clear cut value proposition and hit that really hard," and he models it with his own pitch to a Fortune 500 CIO, which is to reduce IT helpdesk costs by 70 percent by automating incoming tickets, "that's it nothing else" [4].

Depth also defines what he thinks a serious product looks like now. Automating an entire job role requires real surface area, because an IT operations or HR ops person may be expert at fifty things but their work touches Salesforce, ServiceNow, Workday, SAP, SharePoint and more [1]. Leena AI's answer is out-of-the-box integrations with all of those, so customers "can literally configure them in minutes and hours and days instead of quarters and years" [1]. That, he argues, is why CIOs buy: a fast time to value and an ROI-positive product that "goes really deep in their enterprise," with customers able to get running in a day and save millions within a week [1].

On the interface being the least interesting part

Jain resists the framing of Leena AI as a chatbot or a conversational AI product. Conversational AI, in his view, "is an interface or itself it's an information delivery medium," and so are the app, WhatsApp, Slack, Microsoft Teams, Workplace by Facebook, Outlook and the native desktop builds [3]. What organisations actually need is end-to-end service delivery, and so the company built a full service management suite of knowledge management, case management and ticketing underneath: "the real magic is essentially happening in the back end service delivery" [3]. The interface layer is simply where an employee types or records a question and gets an answer back as text or voice, with long answers pushed to the screen rather than read aloud [3].

The underlying problem the interface hides is fragmented knowledge. In large companies information sits across one system or five or fifteen, plus emails, Dropbox and Box, plus tribal knowledge held by a single person or team [3]. Jain describes the contrast that made this vivid to him: in college you could call anybody and get things done fast, whereas inside a company you often have no access to the system and are just waiting for someone to reply [4]. Globally, across more than 500,000 employees using the product, the top three question categories he sees are money, meaning salary, taxes, deductions and reimbursements; leave and time off; and insurance and benefits, followed by requests for employment and visa letters [3].

On what automation does to jobs

Jain does not soften the displacement argument, and he does not treat it as tragedy either. Asked directly whether the technology replaces people, he answers that "technology always replaces people at the end of the day in the long term," compares it to simple agricultural technology, and frames the payoff as freed time for more valuable work or more leisure [3]. His concrete example is an airline HR support team that went from around 102 people in January 2020 to 68 after deployment, achieved by automating simple, repetitive, non-value-add queries [3]. His observation is that cost saving is not usually what customers optimise for: HR teams are chronically short-staffed, and he doubts anyone will ever meet a CHRO who says they have excess people, so the gain is redeploying that manpower onto strategic work [3].

He applies the same logic outside his own category. He is preoccupied with whether the SDR to account executive handoff still makes sense given that it causes friction and AEs can now run full cycle with AI, and treats questions like this from early-stage founders as useful pressure on his own thinking [1]. The general claim is that "every workflow in an Enterprise today can be disrupted using large language models," from processing invoices to purchase orders, and that the bottleneck is organisational rather than technical: enterprise workflows are slow "not of the lack of Technology but the lack of you know wanting to change" [4].

On how to find a workflow worth automating

His method for founders entering this space is deliberately unglamorous and mirrors what he did after Chatteron. Rather than name specific opportunities, which he declines to do because there is "a sea of opportunities," he tells people to reach out cold, introduce themselves honestly, ask for fifteen to thirty minutes, and learn how someone's work actually happens and where their time goes [4]. The test is repetition: "if you hear the same thing from 10 people you know exactly what to automate," and then the instruction is to go hit it hard [4].

On selling to large enterprises with no credibility

Jain has led sales at Leena AI since day one, and did the same at Chatteron, so his advice here is first-hand [4]. The starting condition is acceptance of ignorance and rejection: engineers arrive with zero idea how to do sales, build relationships or establish credibility, and enterprise deals involve hundreds of thousands of dollars rather than a consumer's small purchase, so "you have to be okay with failure because you're gonna get the phone shut on you" [4]. He generalises the same temperament to entrepreneurship overall: "you have to be okay with rejections" [2].

His tactical move is to invert the meeting. Use an institutional tag like IIT to get in the door, then explicitly do not sell: say you are building something in the space, that you have not figured everything out, and ask the experienced person to spend thirty minutes explaining their problems [4]. Only after they have told you does he suggest naming what you are building, admitting it is very early, and asking to partner and start a pilot over the next three, six or nine months, positioning yourself as hungry and eager to learn [4].

On raising money and what investors are buying

The hardest part of a first raise, in his account, is not the pitch but the disorientation: founders straight out of college have no idea where to start and burn three to six months working out what investors care about [2]. He thinks that gap has narrowed considerably, and now that he sees pitches from the other side he notices second and third-year students communicating far more concisely than his own cohort managed seven years earlier [2].

His core point about what is being bought is consistent across stages. Even at Series B, "any BC's biggest bed is going to be you and uh probably not the business," so the job is to come across as clear, authentic and trustworthy [2]. As an angel investor he says he backs the team and the space, because at the earliest stage "the idea the product Etc is so fluid" that founders will spend time in the market, learn and pivot anyway [2]. Investors already assume most things will go wrong, so the two things that matter are selling yourself properly and articulating clearly why you are excited about that specific space, having genuinely researched it [2]. The most common self-inflicted wound he sees is founders who have done great work and never mention it anywhere, when "the biggest product that you're selling in an angel investor pitch is yourself" [2].

On data privacy in conversational AI

Jain's position is that conversational AI does not warrant a separate regulatory regime, and that the absence of one is fair [2]. His reasoning is that the system is not generating or extracting new information: when an employee asks why their salary is lower, Leena AI needs to identify them and read their salary slip, attendance and leave data, all of which already exists inside enterprise systems [2]. The same holds for a telco's voice bot explaining a roaming charge, so the same standard enterprise software protections apply, and the ones that matter to Leena AI are GDPR, ISO 27k, SOC 2, CCPA and FedRAMP [2]. The most sensitive data he handles is employee information rather than customer information, which the product generally does not touch, with exceptions for sales and operations use cases such as a Coca-Cola salesperson querying real-time distributor volumes [2].

On language and accuracy

Accuracy in conversational AI varies with how well-resourced a language is. Widely spoken languages such as English, Spanish, French, German and Hindi have plenty of available resources, while less spoken or more complicated languages do not, and accuracy varies wildly as a result [2]. His practical read is that roughly forty to fifty languages are now well supported in NLP, which covers what the product needs across major global and major Indian languages [2].

On culture as a living organism

Jain describes three pillars. The first is curiosity, which he treats as an operational requirement rather than a nicety, because a high speed of innovation depends on people who "don't take no for another" and keep marching forward [2]. The second is refusing to read criticism as negative; his analogy is that if someone warned you a car was coming you would thank them and brake rather than get defensive, and he says he embraces criticism precisely because it exposes blind spots [2]. The third is patience and trust, or what the company calls always giving the benefit of the doubt: if a colleague has not come back to you, assume they are busy and remind them rather than assume they are not working, on the reasoning that if ninety-five percent of people are good, the remaining five will be weeded out by the system anyway [2].

He is clear that the basics do not change but their implementation must. Having gone from three people to 450 in four years, he argues each stage requires going back to first principles about what the values look like on the ground [2]. Culture, in his phrasing, "is also like a living organism organism and it keeps evolving," and he actively harvests input for it by talking to CEOs and CHROs about how they think about culture in their own organisations [2]. He treats mentorship the same way: what you need to learn cannot come from one kind of person, so the people you rely on evolve with the stage, and the dichotomy for a first-time founder is that you do not yet know anyone who has done it, so you keep talking to people until you find whoever can get you to the next step [2].

On growth, timing and the size of the prize

The trajectory Jain describes runs from a $2 million seed round after Y Combinator in late 2018, through 3x growth in 2019, to a period he characterises as unexpectedly good [4]. COVID, he says, had an extremely positive effect on the business and the industry, because companies that had previously assumed employees could just walk over and ask someone were forced into adopting available technology [3][4]. More recent figures he cites are 4x growth over twelve months, more than $20 million in revenue, a target of another 4x, and 500-plus enterprise customers including Coca-Cola, Puma, Sony, Vodafone, Nestlé and Estée Lauder [1][4].

His stated horizon is long and specific. He has been at this for eight years, "before it was cool," expects to keep going for at least ten more, and believes "there's a 100 billion plus company in the making in this segment," with the conviction that Leena AI has the best product on the market and therefore "a right to win" [1].

Takeaways

• The founder failure mode he names most often is breadth: Chatteron reached 30,000 businesses and about twenty payers because "we were trying to be everything for everyone" and ended up "nothing for nobody" [2][4].
• The fix was ethnographic, not analytical: he spent six and a half months shadowing the HR and IT teams who were paying, and built Leena AI around the problem he found, which was broken access to knowledge inside large enterprises [3][4].
• Out-of-the-box integrations with Salesforce, SAP, Workday, ServiceNow and Oracle are the moat, because they compress deployment from quarters and years to minutes, hours and days [1].
• Treat conversational AI as an interface only; the defensible product is the back-end service management suite of knowledge management, case management and ticketing [3].
• To find something worth automating, interview ten people about where their time goes, and act only when you hear the same problem repeatedly [4].
• Selling into a Fortune 500 with no track record works better as a learning request than a pitch: ask for thirty minutes on their problems, admit you have not figured it out, then propose a pilot [4].
• Investors at every stage are betting on the founder rather than the business, so clarity, authenticity and demonstrable knowledge of the space matter more than the idea, which will change anyway [2].
• Conversational AI needs no special privacy regime because it surfaces data that already exists inside enterprise systems, governed by GDPR, ISO 27k, SOC 2, CCPA and FedRAMP [2].