A venture capitalist once asked Deepak Kumar never to contact him again. Roughly seven years later, Kumar says, the investor replied to that very email with an invitation for coffee. Adaptiva was attracting attention. The earlier message remained conveniently attached. Even venture capital has a version-history problem.
Kumar tells the episode with an engineer’s appreciation for an inconsistency. After the early rejections, he financed the business himself and sought people willing to help it grow. “I realized I needed to find allies, not VCs.” The encounter makes a lively opening to a quieter story: a founder spending years on the work that happens after a product demonstration ends.
A better kind of work
Kumar grew up in India. He has described an early fascination with computer science and the pace of American technological innovation. Moving to the United States offered the prospect of making things he could first picture in his head. That attraction to imagining and building would follow him through communications technology, corporate computing and eventually his own company.
At Nortel, he managed a group developing advanced IP telephony products. His career at Microsoft included the Windows NT Networking team and the Systems Management Server 2003 team, where he was lead program manager. His later biographies identify him among the principal designers of SCCM, the management software familiar to generations of enterprise administrators.
Those roles placed him inside a particular kind of ambition. Making software operate on one computer is a task with boundaries. Making software arrive, install and behave across a large organization introduces offices, networks, schedules and people who have quite enough to do already. The product’s usefulness depends on all of them.
There was another strand to his Microsoft work. Kumar contributed to a Thinkweek paper for Bill Gates associated with Project Greenwich, part of the ancestry of Microsoft’s later enterprise communications products. He describes that experience as a lesson in how a technical idea can grow into a change in the way people work together. The thread connects naturally to his earlier interest in networking.
The network already sitting on the desk
Adaptiva, founded in 2004, grew around the practical problem of delivering software across distributed computer fleets. Corporate devices need updates. Updates need to travel. Sending identical material repeatedly over limited connections can turn a routine deployment into a contest between the IT department and everyone else trying to use the network.
Kumar offered a concrete example in 2014: an SCCM client of about 180 megabytes. Across hundreds of thousands of machines, the repeated copies become a considerable delivery job. OneSite’s peer-to-peer approach could bring that content to a location once and let nearby computers share it. The useful insight was sitting on the desks: the recipients could also help with distribution.
He described bandwidth prediction, pooled storage, local memory and peer-assisted deployment as parts of the design. Content could arrive before the moment administrators needed to install it. That separation matters. Moving the package and deciding when to apply it are different jobs, with different opportunities to cause inconvenience.
By 2015, Kumar was also explaining why the update process itself needed simplification. Customers sometimes skipped several versions and updated only after months had passed. OneSite’s Update Manager was intended to handle those accumulated changes through a customized patch. An update tool that needs an elaborate updating ritual has, after all, acquired a rather personal sense of irony.
His later writing returned to the resources already present on corporate devices. In 2022, he argued for using spare computing and storage capacity at the edge to distribute updates among peers. Remote work made the question more pressing: a management system had to reach machines wherever work was taking place. The network diagram could no longer assume everyone was conveniently gathered inside it.
The questions he puts to customers have become another way of examining that burden. Introducing a 2023 patch-management study with Ponemon Institute, Kumar organized the problem around visibility, deployment and process management. The report found that 54% of respondents considered detecting vulnerabilities among the hardest parts of patching, while 50% identified assessing application risk and exposure. For a designer, those answers are a reminder that the installation command is only one part of the job. Before issuing it, someone needs to understand what is there and what matters.
The engineer takes the CEO’s chair
In early 2020, Adaptiva documented Kumar’s transition from chief technology officer to chief executive. Davinder Singh became CTO and Shailesh Kumar became vice president of engineering. The founder’s job now included leading the company around the architecture he had spent years developing. His title changed; the product problem remained recognizably his.
Adaptiva reported more than doubled revenue in the first quarter of that year, compared with the year-earlier period, and nearly 425,000 added seats. These were company-reported business results, rather than a measure of Kumar’s personal fortune. They do help explain the change in scale: a technical proposition had accumulated customers, obligations and an organization to run.
His language about the work still makes room for aesthetics. “To me, it’s an art form,” he has said of software. He describes visualizing its underlying design. A corporate deployment system may be an unlikely gallery piece, but elegance has a practical audience here: the person who has to make the parts cooperate.
There is a useful check on that artistic instinct in his stated leadership approach. Kumar says he encourages teams to challenge assumptions while connecting their ideas to customer outcomes. The design has to accomplish something outside the designer’s imagination. A satisfying internal structure earns its keep when somebody can use it without inheriting all of its complexity.
He also has a life beyond the console: his public biography identifies him as an outdoorsman and hiker. Alongside the networking work, patents and technical publications, it is a small reminder that a career spent thinking about connected machines can leave room for going somewhere without them.
In a 2021 essay, he pictured systems that could maintain themselves with little intervention. The attraction of that idea lies partly in where it leaves the IT staff. Every recurring maintenance step occupies a place on somebody’s calendar. Kumar’s architectural choices invite a plain test: when the system changes, does the person’s working day improve? That is a less photogenic ambition than a new device or a consumer app. It is also a question an enterprise customer can bring back to the next renewal meeting, with fairly specific answers.
Permission to stop the machine
The newer challenge is persuading people to let software perform more of the process. A patch can fix a vulnerability and still interrupt someone’s work. The administrator responsible for an estate of computers has reasons to be cautious. Calling a product autonomous does little, by itself, to settle the question of who answers for the result.
Kumar has described trust as a substantial obstacle for Adaptiva. His answer includes controls that let administrators pause or roll back updates. The ability to intervene gives an otherwise abstract assurance a visible shape. A system can do its routine work while leaving the person responsible with authority over what happens next.
OneSite’s deployment documentation shows how that thinking becomes settings. Administrators can distinguish patches by attributes, set installation behavior and restrict approvals to particular business units. Built-in strategies offer choices about initial approval, phased approval and rollout timing. A fleet does not have to behave as though every application and every department has the same needs.
An editorial outline of Kumar’s approach to autonomy.
That balance also appears in the subjects he chooses to discuss publicly. In April 2025, he spoke on Apple @ Work about expanding Adaptiva’s patch management into Mac and Linux. The problem was extending beyond the Windows world that had shaped much of his earlier career. Different operating systems add variety to the environment; the people managing them still need control over the work.
The rules come before the rush
By June 2026, Kumar was pressing a distinction between automating individual tasks and completing the whole chain of work. A process can contain automated stages and still wait on people between them. His concern includes the coordination between identifying vulnerabilities and carrying out repairs. A fast component offers limited comfort when the next handoff is sitting in a queue.
In August, he pushed the discussion toward capacity. An organization can debate an update deadline without knowing whether it could actually respond to an urgent problem tomorrow. His preferred division of labor puts strategy with people and repetitive execution with software. He also argues that a system should recognize situations beyond its authority and bring a person into the decision.
“The decision to involve a human should itself be part of autonomous behavior.”
Deepak Kumar · August 2026
His October 14 writing makes the preparation explicit. Organizations need to decide priorities, differentiate strategies, consider end users and establish the validation required before software acts. A browser update and a sensitive business system may warrant different treatment. Moving quickly requires those differences to be worked out beforehand, when there is time to think.
That brings the story back to the administrator’s desk. Kumar’s career has passed through telephony, Microsoft management software, peer delivery and autonomous patching. Across those changes sits a persistent design question: how much repeated work must a person carry because of the way a system was built? His answer keeps returning to architecture, clear rules and the freedom to intervene. The coffee invitation is an enjoyable footnote. The work left off someone else’s calendar is the more durable result he is trying to build.