Amidst the AI revolution, e27 presents a new article series showcasing how organisations embrace AI in their operations.
Amir Movafaghi is CEO at Mixpanel. Prior to Mixpanel, he served as CFO at Spiceworks Inc., an IT network and marketplace connecting companies to technology solutions across industries.
Previously, he held various leadership roles at Twitter, where he helped it scale from 150 to 4,000+ employees and led it through its IPO.
In this edition, Movafaghi shares how his company has embraced AI.
Edited excerpts:
How do you perceive the AI revolution and its potential impact on your industry and workforce?
The potential of AI has been spoken about for some time, but it’s only since generative AI models became available to the masses that people and businesses have started to notice. That’s because generative AI is just the next interface of computing, unlocking huge productivity gains across various sectors and industries.
In the world of SaaS, the rules are changing. It has long been the case that productivity has required technical formulae or exhausting interfaces. Generative AI is unframing all of that. Have some code that you need to generate, translate, or verify? You can now click a button to get AI to write and organise it for you. New efficiencies like these, and the wow factor they bring, are things we’ve never seen before in software.
In our world of analytics, it means making everything more accessible. If anyone can now query data in plain English by asking the AI a question, it means everyone in an organisation can participate, not just a select few more technical-minded colleagues. Making it easier for anyone to gain insights from data will increase collaboration at companies, helping teams to have higher-quality conversations to solve problems more quickly and with better outcomes.
In what ways has your company embraced AI technologies to improve operational efficiency or enhance business processes?
Earlier this month, we introduced our first step into generative AI. It’s called Spark, and our focus has been to help speed up workflows and simplify how people ask questions about their data.
It works quite simply: if you have a question about your data in Mixpanel, you can just ask it in plain English. You might want to know, “Which market was responsible for website traffic yesterday?” or “How did a particular cohort of users respond to a message or push notification?” Spark will build the right report to get you the right answer, complete with the corresponding chart.
This works for any user of any type across an organisation. For example, a financial services app that has just launched a new ‘tap to pay’ feature and a nontechnical user wants to find out the performance of the feature amongst different user cohorts. With Spark, you can now get a quick answer by asking, “Which group of users have used ‘tap to pay’ the most in the last week?” And the AI would understand the question and build the query in the platform to generate a report.
Similarly, a marketer could ask a question regarding trends related to advertising periods and compare it with money spent on advertisements to understand if a campaign had been driving users to a website or had affected the use of an app. A salesperson could use AI to see revenue changes over time or understand if users were making it through the cart or abandoning it early.
Also Read: Saison Capital, Mixpanel team up to launch a product manager peer-support community
This is all very important for us at Mixpanel because we’re working hard to allow companies to understand the impact of their actions on the user experience throughout that user’s entire journey with the company. It’s becoming possible to use Mixpanel to measure how users engage with ads quickly, the actions they take in the product, and how they respond to messages.
Making this linkage for the complete understanding of the full user journey means companies can understand how their actions, like building a new feature, impact bottom-line revenues. Our vision sees every function in a company having this same view so teams can easily understand and focus on what’s working — generative AI accelerates this transition.
But this is just the start of the journey. Large Language Models (LLMs) will continue to evolve and impact analytics for years to come, and we’re excited about the potential of the technology for our users and how anyone can build better products.
Can you share specific examples of how AI has been integrated into your workforce to streamline operations or drive innovation?
At Mixpanel, we initially focused on integrating OpenAI’s enterprise model into the Mixpanel analytics tool. It helps users ask questions about their data more easily and quickly by asking the AI to build a query in Mixpanel. Mixpanel has always been easy to use and has never required complex coding, but AI takes this UI and ease of use to the next level.
What challenges or concerns did you encounter when implementing AI technologies within your organisation, and how did you address them?
Mixpanel is trusted by many of the world’s most exciting companies to look after their data. We take this responsibility extremely seriously, so we knew we needed an enterprise LLM and an initial use case for AI where we didn’t need to expose any customer data.
That’s why we’ve focused on natural language chat for analytics query building. It pushes our vision of ‘analytics for everyone’ forward by making Mixpanel even easier to use, but we don’t share any customer data.
We also needed to ensure the AI’s work could be easily verified. To achieve this, we allow users to review the query the AI has built, so they can be sure the chart it generated answers the right question. Generative AI is still developing, and it is crucial to ensure humans can review its work.
How do you ensure transparency and uphold ethical considerations in using AI technologies within your organisation to mitigate privacy concerns?
After testing a variety of LLMs, we opted to integrate OpenAI LP’s GPT-3.5 Turbo Large Language Model, a technology similar to ChatGPT, which is capable of humanlike speech and understanding, to allow its users to “chat” by simply asking a question and the AI does the work for them.
A lot has been said about the risks associated with the technology, which was an integral consideration in our decision. OpenAI LP’s GPT-3.5 Turbo is an enterprise model, so our users will not need to contribute their data to the LLM, and it will only be used to increase the speed and reduce the effort of building queries. In essence, Mixpanel analyses the underlying data, not the LLM. The LLM makes it easier to ask questions with Mixpanel.
We’ve also made transparency central to Spark. As a guiding principle, any generative AI feature we deploy in Mixpanel will be able to “show its work,” which means you’ll always be able to check for yourself exactly how analysis or other content is being generated. For example, when Spark builds a report, it’ll be viewable and editable like any other report, meaning you can go into its query builder view and see details like what events are being used.
Also Read: AI tools enhance efficiency but can never replace human creativity: Gia Ngo of Give.Asia
How do you ensure that AI technologies complement your workforce’s existing skills and expertise rather than replacing or displacing human workers?
While there are reasonable fears that technology will ultimately replace humans, I think it is often overstated and misplaced. Yes, some organisations have focused on automating specific roles once occupied by a human, but I think many of these decisions will only lead to short-term productivity gains. Those who just deploy the technology to displace or replace workers will neglect the real value and transformation that this technology can bring.
For me, the real value for businesses lies in how humans and AI will enhance each other’s strengths — the speed and scalability that AI brings, coupled with the communication, teamwork, creativity, and social skills of humans. The value is in how we as humans can collaborate with technology – how we can enhance what these machines are capable of and how those machines can augment what we do best.
Our use case is a good example. The AI does the manual element of query building, but the creative quality of the human knows the right question to ask of the company’s data.
How do you envision the future collaboration between humans and AI? What role do you see AI playing in augmenting human capabilities?
We’re going through a time when most people and organisations are consuming ‘off-the-shelf’ models and getting to grips with what these models are capable of. However, the biggest value will come when businesses and users begin customising and fine-tuning these models to address unique and specific needs.
While we’ve focused mostly on helping speed up existing workflows, the possibilities for what more customised use cases of AI can bring for human capabilities are endless, from scalability to improving decision-making to personalisation. This is already beginning to take shape, but we have a long way to go.
For example, in the future, companies might be able to use data insights from Mixpanel about different cohorts of customers to personalise the messages, images or content they display to users. Mixpanel can provide user insight, and generative AI could work with that to curate the right experience for that user. It’s an exciting future.
What advice would you give to other company founders looking to leverage AI in their workforce?
Exploring AI is not just a nice-to-have — it’s a must. Generative AI, in particular, opens up a new world of possibilities, and the technical and economic requirements are not prohibitive. The downside of not doing anything is quickly becoming that you will just fall behind competitors. However, it is important to balance this need with assessing requirements around data privacy, IP protection, security, and governance to ensure risk is well managed.
The other thing I would say is that generative AI is almost purpose-built for this community, particularly for new founders and entrepreneurs. Not just because of the limited barrier to adoption but more so about how it allows you to rapidly build, test new prototypes, test new concepts, and continually iterate at speed, which is something, particularly in software, that we’ve never had. Teams really need to be considering how generative AI can augment their own capabilities.
—
Fundraising or preparing your startup for fundraising? Build your investor network, search from 400+ SEA investors on e27, and get connected or get insights regarding fundraising. Try e27 Pro for free today.
The post The value for biz lies in how humans, AI will enhance each other’s strengths: Mixpanel CEO appeared first on e27.