Zoom pursues the path to happiness through AI

Other capabilities

Other key capabilities allow the system to understand what users are seeing in the core Workplace app, remember prior interactions and provide suggestions relevant to their work. In addition to identifying and completing next steps and action items, with the launch of Zoom Tasks in late 2024, AI Companion 2.0 will be able to help users get even more done by surfacing, tracking and completing actions across Zoom Workplace.

By employing what it called a federated AI approach – making use of a number of major AI models including OpenAI, Anthropic and Llama – Zoom said it can optimise AI quality and performance, furthermore enabling AI Companion to produce outputs with 36% fewer errors in transcriptions and 15% fewer errors in post-meeting summaries compared with Microsoft Copilot, cited by a number of execs at Zoomtopia as the company’s chief rival. The company also announced a Custom AI Companion add-on, available in the first half of 2025 for $12 per user per month, allowing organisations to tailor AI workflows to their specific needs.

While going full bore on AI in the product sets is something new for the company that really made its mark during the early years of the pandemic as the leading video conferencing and collaboration platform, Zoom has been on a marked corporate pivot to maximise the opportunities in what it calls the true modern workplace.

That is, a world of not just home working and office working, but one that now supports the changing nature of the employee experience, that has not just seen the shift from traditional office spaces to home but also to include experiential working hubs and engagement for anywhere there is connectivity. That isn’t just fixed environment or mobile; it now very much includes new compute space such as vehicles. Employers and employees have experienced the benefits of a more flexible approach to work and want to use advanced, invariably AI-driven technologies to maintain them.

The bottom line is that more and more communication will take place across more channels, with collaboration across oceans, and enabling this will be AI. Those businesses that best adapt to these changes and optimise the benefits of technology and flexible working, while maximising the times that teams spend face to face, will inevitably be those that succeed.

The upshot is that Zoom is not just a video conferencing company anymore. In this past year or so, it has been emphasising that it now fundamentally provides a platform to solve business communications such as contact centre and IQ for sales, and a platform to learn from customers to respond better to their needs.

GenAI changing everything

Speaking at Zoomtopia about how he feels the new technologies can disrupt the market, Huang – who boasts 30 years working at Microsoft – stressed that GenAI was changing everything.

“Microsoft is the leader in productivity,” he said. “Then with the internet coming, Google dominated the Web. They brought the Google Docs and [its] workspace really enhanced collaboration. But they didn’t break the traditional three pieces [of office technology]. They have a PowerPoint equivalent called Slides. They have an Excel equivalent called Sheets. So, they still have more or less the same set of productivity tools Microsoft defined. Generative AI is completely changing what was defined by Microsoft.

“Why we can be successful? There are three things, three phases,” said Huang. “GenAI can help you to focus on those important things you care about, that is action-oriented tasks. Zoom AI Companion will identify next steps as to who should be doing what. That will not stop just with the form of meeting … information will flow into every corner of Zoom Workplace that will stay in the whole lifecycle. So, if you have a meeting, you get an identified action item. If a week later, you get a chat message, AI Companion will identify whether that is related and the action item can flow into a Zoom [session] you’re working on. So, that’s [why] AI Companion 2.0 is going to really differentiate as a very different [product].

“Microsoft has made the statement that – I think it’s good – that CoPilot is your UI [user interface] for AI,” he said. “It is challenging the definition of what the AI UI is, and we should not abandon what is optimised for the graphics interface. ‘What you see is what you get’ is actually very powerful. What we are trying to do is bring those two together seamlessly. That information flow from graphics interface into conversations. Let’s say you are chatting with someone. We can bring the calendar scheduling automatically, filling … what is the best time slot for subjects.

“AI Companion will actually have the context to have that conversation,” said Huang. “Innovation will flow [into a] conversation’s graphical user interface. That is what I call a UI. An AI-first user interface, a UI combined with task completion, information flow. The way to do this is [to] have a federated AI backend. We are training our own small language model that is running on the device to understand broader context and complexity. The search engine will challenge both Microsoft Beam and Google. We are putting the rest of the whole world together [from disparate AI models]. That’s our package.”

AI-first work platform

For Hashim, the product launches encapsulated just how the company had been on a journey to become a work platform, and for the company, to be an AI-first work platform at that.

Explaining what she meant about the latter, she said: “When we talk about an AI-first work platform, with generative AI there’s really a possibility for it to help people free up time. And because if you free up time from the routine, the mundane, then [people] can actually do more creative work, more engaging and more insightful work, connect more with teammates and so on. We started on that [journey] with our AI companion, which we launched a year ago, and we have seen really good response to our meeting summary, meeting questions. [You can think of] AI Companion 2.0 as a single brain which works across all of Zoom Workplace and is contextual.

“Custom AI Companion will help our customers connect to their own data sources,” said Hashim. “It could be their custom vocabulary. It could be their internal knowledge databases, including, say, employee handbooks, policies, the third-party apps you might work with like ServiceNow, Jira, Salesforce. You’ll be able to connect to all of those. It will also be able to get a lot more personalisation in terms of avatars for clips or personal coaches. We also have products which go deep for industries like education and healthcare.”

All of these announcements were very much geared towards AI. In terms of AI models, Hashim confirmed the company was working with OpenAI, Anthropic, Meta and Perplexity to allow people to develop their own large language models, and that it was also working on developing its own small language models.

Yet the success of the new Zoom products, like any AI-based technology, is very much dependent on the regulatory environment of the countries in which the products will be rolled out – especially Europe. Hashim assured that this would not be an issue, and aiming to assuage worries, she noted that the company already works “very closely” with regulations and regulators, in particular in terms of the way it has built its AI companion. Hashim specifically stressed that Zoom does not use any customer data for training zoom models or third-party models.

“We give customers very fine-tuned, granular controls so they can turn every feature on and off,” she said. “So, for example, they can choose whether they want to turn on some features, whether they connect to anything like Google or Microsoft. We also give them a lot of control over data. For example, how many days do they want to retain it? What do they want to delete immediately? We have what we call a Zoom-only model of zero-data retention, because we work with so many enterprise customers.

“So, we’ll continue to work with the regulation and responses,” said Hashim. “But I would say the bar for us, we set it to be very high. We work with so many large enterprise customers, and we are very responsive to their asks. 57% of Fortune 500 companies have enabled Zoom AI Companion at this point, and they wouldn’t do it unless they [undertook] a very thorough analysis of the privacy, compliance and security before they would turn it on.”

This will be central to the mission of fulfilling Zoom’s stated aim to “revolutionise and transform” the way people work. Assessing how much of a transformation and revolution the new products will deliver compared with the first version of AI Companion, Hashim said she sees it as a step change.

In addition to the functional enhancements, the launches herald the arrival of a new business model. A Custom AI Companion add-on will become available in the first half of 2025 for $12 per user per month, allowing organisations to tailor AI workflows to their specific needs. This could mean the ability to connect preferred apps and data sources, with AI Companion acting as, suggested Zoom, a “super agent” across multiple platforms. Hashim illustrated this with an example of AI Companion pulling employee data from the Workday product creating a service ticket in the Zoom environment.

Such capability will pit Zoom directly head-to-head with Google Meetings and Microsoft Teams in the premium space, both of which having considerable financial and technological heft to compete effectively with Zoom. Hashim was confident as to how the company would take on its rivals – especially when it came to deployment of AI-enhanced capabilities.

“I would say Zoom Meetings is head and shoulders over everyone else that we know,” she said. “When we are talking to our customers, first of all, Google and Microsoft are charging a lot for the AI. They’re charging $30 per user per month. It is a lot of money. And in terms of their [AI], if it is Microsoft, it’s Open AI; if it’s Google, it’s Gemini. So, we take this smart, federated approach. We can manage cost well through a federated model.

“And then the reality is that most customers use a lot of applications, because Microsoft and Google, even if they’re using these products, we connect to them. We include an AI Companion 2.0 connection to Microsoft Outlook or Gmail, Calendar or Docs, it’s included at no additional cost. An average enterprise customer uses more than 200 apps [and] our AI Companion add-on is going to provide the connectivity to these apps in a very seamless way. We also very much expect to have to earn customers’ love through great execution and quality, which we will continue to do.”

Key trends

Among the key trends in the unified collaboration and unified communications (UCC) spaces has been the rapid transition, or perhaps more accurately, evolution, of the services to address the contact centre, and then to look at the whole issue of customer experience as a whole.

Zoom now recognises that it needs to have a viable and competitive contact centre service, and at Zoomtopia, Hashim explained that work in the area had actually been carried out over the past few years, and that the key was to understand the different needs of the different users of a contact centre.

“We are actually seeing really good growth interaction on the contact centre,” she said. “It’s really [like] what you saw in our phone product. From a contact centre point of view, the way we have approached this is to go very high on usability. We are thinking about the experience of the end user, but also the agent and supervisor. We think of all of them as being distinct users, and each and every one of them deserving a great experience. If you are a customer, you want to be able to connect flexibly to any kinds of channels, whether it is chat or email or video or phone, or even WhatsApp or social media, we support all of those kinds of things.

“We have a virtual agent, like a customer bot, which we are expanding to be able to detect multiple intents so that it can be more and more of a natural conversation,” said Hashim. “From an agent point of view, we include Zoom AI Companion in all our contact centre licenses. As an agent, you can get help with things like sentiment analysis and summarisations.

“So, as thoughts move from one agent to another, you can get the summary,” she said. “It’s a much better experience for customers [and] also for agents, because the customer is not irritated with them because they are repeating [themselves]. If you are a supervisor, then we [can] help you analyse conversations to guide your agents. But we are also adding capabilities. We include AI companion. We have an advanced product called AI Expertise. We are adding things like dynamic agent guides, automatic supervisor flagging. We think about how to make the agent’s life and the supervisor’s lives easier.”

Customer base

Looking at where these tools are being deployed and by whom, Hashim could not reveal specific customers, but noted that the company had a “pretty broad” base of about more than 1,300 customers from a contact centre point of view – working in industries such as insurance, banking and healthcare – and that this number was growing “quite rapidly”.

However, like any business, there will be challenges to overcome to attain this growth. When asked what key technological challenges Zoom will face over the next year, Hashim pinpointed AI quality, something that she said will need to be addressed by everyone – an issue exacerbated by the fact that enterprise data is so broad and so complex.

“In order to continue to use that data to create really meaningful answers for the user, it’s an ongoing challenge,” she said. “We all have to be respectful and mindful of that challenge, and that’s where we try to do a lot of work with our federated approach, but we also go really deep.

“AI Companion 2.0 is a complex product – with chat and emails and documents and meetings – with so much happening,” said Hashim. “You have to be able to connect to all of these in seamless ways, pull out the data, analyse it, be able to add even more data, synthesise it. You have to go in really deep to fix the quality and make sure that the quality is good.

“Our AI team has … just started creating more small language models and testing the quality, and we will be able to add that to our competitive mix as well, in order to get better results and make them more cost effective,” she said. “We are giving a lot of value to our users by including knowledge of cost.”

User experience

In addition to specific technological challenges, Hashim, like Huang, also stressed that getting the user experience right was important. “Especially when you’re talking about a cross platform, having an experience that is natural and intuitive, and not kind of cluttered, but gives the right controls and choices – we obsess a lot about [that],” she said. “Getting that right takes a lot of focus, it takes education … and takes intensity of purpose.”

Huang conceded that the offering does not currently work as fast as he wants it to because of latency, yet stated that AI Companion is moving beyond what rival products can do through what he calls an “engineering miracle” of integration – and through this, changing Zoom itself, and ultimately delivering happiness.

“Our focus and differentiation [are] task completion to action,” he said. “We all have the same time constraint. No matter what you do, time is not on your side. So, when we can focus on action or rate of a task completion, you gain time. In order to achieve that, we want to bring the best AI backend. We want to bring the best AI user interface.

“I believe this is fundamentally going to change the industry,” said Huang. “We believe innovation requires us to focus on the most important things to all of us, the main constraint [being] time. We can really help you save time. We are not just a meeting company anymore, nor a video communications company.

“We are an AI-first work platform company for human connection and delivering happiness,” he said. “It’s a fundamental value, no matter where you are.”