Kevin Gao and the art of doing less

Christmas Day is an unusual date to start charging for something people have been getting free. For Kevin Gao, it became a dividing line. Comm100 introduced fees on December 25, 2011, after roughly two and a half years of free service. Existing users kept the features they already had. The business around those features had to change.

Then came a second decision. In mid-2012, Gao concentrated resources on live chat, cutting back from a six-product effort. His original ambition had been an integrated communications suite. For the time being, one product would carry it. “We decided to rely on ourselves,” he said of the earlier funding decision. Investors had not offered terms he considered acceptable.

There is a pleasing inconvenience in this beginning. Founders are encouraged to have a large vision, and customers are encouraged to expect a low price. A company must eventually put those two preferences in the same room with its expenses. The conversation can become awkward. Gao’s answer involved a smaller immediate assignment.

The useful distinction is between the size of an ambition and the number of things a team can finish. They need not grow together. Narrowing the work can give an idea somewhere to stand. In Gao’s case, the later return to a broader platform makes that early retreat worth a closer look.

A buyer with a builder’s complaints

Before Comm100, Gao founded Dynamsoft in 2004 and worked as its CEO and system architect through 2009. His education joined computer science at Zhejiang University with an MBA at the University of British Columbia. He had experienced customer communications as a manager dealing with disconnected tools. Comm100 followed in 2009.

That background helps explain the kind of advice he later gave software buyers. In a July 2015 guide to choosing live chat, Gao began with a question about what the organization actually needed. A small online shop and a large enterprise could arrive at different answers. A list of features was only the beginning of the purchase.

He urged readers to distinguish requirements from conveniences, compare equivalent capabilities, and try the software. He also questioned reviews that awarded credit for a feature without examining how it worked. The advice is almost comically sensible. Software shopping has an impressive capacity to make sensible questions disappear under a comparison table.

There is a particular kind of technical salesmanship here: asking a customer to be a more demanding evaluator. It gives the demonstration a harder job. An automatic invitation, for example, has to arrive at a useful moment; merely possessing the button does little for the person receiving it. Gao’s writing keeps returning to that distance between a capability and a helpful experience.

Scripts, with permission to improvise

Gao’s January 2015 writing on canned messages addressed a small indignity familiar to anyone who has used customer support. A reply can arrive quickly and still feel useless. Reusable answers save typing, especially when an agent handles several conversations. They also risk making a customer feel like the latest recipient of a very busy rubber stamp.

His advice was to keep those messages personal and refresh them as products and policies changed. He compared capable representatives with actors who adjust a script when the role requires it. That is an unexpectedly theatrical image for support software, but an apt one. A script helps someone prepare. It cannot anticipate every person who walks onstage.

In another 2015 piece, he treated consistency as a matter of procedures, training, and observation. Managers could use conversation monitoring to identify problems and help agents improve. The attention to everyday practice matters. A product may supply the tools, but a team still needs shared expectations about how to use them.

His email-marketing advice carried a similar concern for the recipient. In a 2013 article, he recommended changing the approach as customers moved through different stages of a relationship. He warned against excessive contact, encouraged relevant information, and asked marketers to consider whether they would read their own material. That last test is inexpensive enough to survive almost any budget meeting.

Together these pieces suggest an approach to product leadership grounded in the receiving end of communication. Speed, consistency, and personalization all matter, but each can become irritating when pursued mechanically. For Gao, the practical question running through the work is how a system helps a person respond appropriately to another person.

“Our dream never changes.”Kevin Gao · on narrowing the product strategy, 2013

The bot needs someone to correct it

By 2018, Gao was explaining how to improve customer-facing AI. The work began with likely customer questions, a body of answers, and training. A confidence score could flag an uncertain response for review. Human agents could also mark an answer as unhelpful, prompting an expert to examine it. The machine’s output became material for further work.

That explanation contains a useful antidote to the effortless talking machine. Somebody must decide what a satisfactory answer looks like. Somebody must notice when the software has missed. A support organization accumulates that knowledge through awkward questions, unusual cases, and corrections. Turning it into something a bot can use is an operational task with technical consequences.

Gao also discussed the difficulty of adding languages. Related language families can offer useful similarities; a language outside those families may require different processes. The point resists the idea that a conversation system can simply be copied into another market. Communication carries structure as well as vocabulary. A translated interface is only part of the journey.

His company was also building its management capacity. In March 2018, Comm100 announced a North American leadership team spanning marketing, enterprise sales, solutions, and customer success. Robin Jones, Jeff Epstein, and Tony Sandhu were among the appointees. Gao described investment in people alongside additions to the platform. Growth required more than another development sprint.

The connection between those two jobs is easy to overlook. Training a bot and building a leadership team both force a founder to distribute judgment. The decisions can no longer all live with the person who started the company. For a founder, each adds a question of responsibility to the question of capability.

Count the conversations, then ask what happened

Gao’s introduction to Comm100’s 2022 live-chat benchmark report put another discipline beside the product work: comparison. The report examined more than 36 million chats handled during 2021. Its categories included workload per agent, waiting time, conversation duration, satisfaction, and mobile use. Results were broken down by team size and industry.

Those distinctions change the meaning of a number. A short conversation can reflect an easy question or an unfinished answer. A busy agent can be productive or overloaded. Comparing a team with a relevant peer group is more useful than admiring a large total. The report’s purpose was to help teams see where their service needed attention.

For a founder whose business sells conversation software, examining the conversations is a sensible extension of the job. It also introduces a constraint on the sales pitch. Once the waiting time and the outcome are visible, a customer can ask whether the system is helping. The software has to answer for what happens after installation.

Six jobs, a second time around

In April 2025, Comm100 introduced AI Agent, AI Copilot, and AI Insights. The division of work was explicit: one addressed customers, another assisted human representatives, and the third helped managers examine conversations. Gao was still CEO. His company was once again offering a connected set of tools around the support desk.

The details were more prosaic, and therefore more revealing, than the product names. Copilot could help prepare replies and summaries. Insights could track resolution and flag issues. AI Agent was designed to carry context through an interaction and work with enterprise systems. These were different jobs within the same conversation, each with a different user.

In September, Comm100 added AI Knowledge, AI Quality Assurance, and AI Training. The announcement addressed the maintenance work behind the customer-facing exchange: finding gaps in documentation, checking interactions against guidelines, and helping new agents practice. A customer’s unanswered question could become a clue that the organization’s information needed repair.

The training product used chat-style scenarios and adaptive questions. Knowledge tools could recommend revisions or draft articles from conversations. These were product capabilities the company announced, rather than evidence that every deployment produces the same result. Their significance in Gao’s story is the choice of problems: learning, checking, and keeping information current.

By October 12, 2026, Comm100’s release notes described audio, video, and screen sharing within Anytime Chat tickets, alongside a service-account option for connecting Google Drive. Visitors could accept or decline a session request while continuing the conversation. It was another attempt to join activities that might otherwise send people elsewhere.

The old ambition is recognizable in these small pieces of workflow. A conversation should be able to change shape without losing its place. Someone should be able to step in with enough context to help. The tools may now include AI, but the inconvenience Gao has spent years addressing remains familiar to anyone who has repeated a question to a second department.

His current statement of purpose puts understanding people and their need for connection at the center of support innovation. The practical challenge is to make that intention survive the machinery. Gao’s career offers one concrete way of approaching it: choose a manageable part of the conversation, make it work, and earn the right to connect the next part.