The heady, accelerated growth across the Web3 space in 2021 not only brought wider public adoption, pumping portfolios and a plethora of new projects; its natural corollary was, of course, user numbers. Despite only hitting the 5 million monthly active users (MAUs) mark in April 2021 – growing by a factor of five in comparison with the preceding six months – MetaMask’s MAU count surpassed 21 million in November and is now over 30 million

Naturally, this massive influx of users necessitated a corresponding surge in MetaMask’s support capacity: at roughly the same time as MAUs went past 21 million, v1 of the LivePerson bot went live. Engaging the support of a market leader in conversational AI has joined the dots between several existing support mechanisms that were already offered, including a two-tiered ticketing system, a knowledge base, and a community forum. As user numbers rose, the bot’s AI has helped efficiently redirect users to existing content to help resolve their queries faster and without the need for email-based correspondence. Where the bot recognizes that it can’t link the user to the correct resource, it escalates to an agent. 

In adding the conversation bot, we are also providing users with a means to access support the way they prefer; after all, the most effective type of support will always be the one people are happy to use. Self-service resolution, where users are signposted towards solutions without contact with support staff, or with AI assistance, is now most people’s preferred method. A Harvard Business Review study suggests 81% attempt to self-resolve before looking into direct contact. 

Implementing and developing LivePerson is part of our long-term strategy for equipping our support services to meet user expectations, and reflects our commitment to investing in and prioritizing customer experience. In contrast to some Web2 software companies, we do not rely on customer inertia – the tendency to stick with the familiar – to keep our user base engaged. Instead, optimizing the customer experience is our core focus, as evidenced by the significant resources and expertise devoted to implementing the bot in collaboration with LivePerson.  

Implementing v2: Responsive, Efficient and More Accurate

Much like Web3 in general, user needs are in flux. Accordingly, the level of support users need is a moving target. This is why MetaMask’s AI bot was always intended to be improved upon throughout successive iterations. The first of these, v2, went live in mid-February and was the result of a significant collaborative redesign of bot behavior, comprising months of work. 

Consensys and LivePerson teams analyzed lessons learned from the v1 bot and intensively restructured bot flows – i.e. the sequences a user clicks through when they access the support conversation – to better align with users’ language choices, reflect the most common issues, and support users’ ability to intuitively self-resolve.  

We’re proud to report that these efforts have paid off. When compared with a weekly snapshot of the v1 bot taken in January, the v2 bot as of 6 March evidenced that: 

  • Customer satisfaction – a score out of five gathered after the conversation ends – has more than doubled

  • The bot’s ability to help users resolve their issues without escalating to an agent increased by over 70%

The benefits don’t stop there. With the bot pulling more weight, agents can react more responsively when the bot assigns users to them. Using the timelines above as reference points (January to March), the new bot supported a drop in agent escalation wait times of 93%, with users now able to begin talking to agents within minutes – quite the improvement. 

What’s on the Horizon?

We’re confident that MetaMask Support is now in a considerably better position to provide prompt resolutions to users’ problems, and we hope that you – as users, the intended beneficiaries – have felt this change too. 

Rather than resting on our laurels, we’re going to continue working on future improvements to the bot, as well as broadening LivePerson AI integration to encompass other aspects of MetaMask Support. In addition to freeing up agents to focus their attention on assigned tickets and conversations, MetaMask Support will have even greater capacity for whatever levels of demand future bull runs can conjure. 

In addition, our roadmap includes new features which will make the Support experience more interactive, efficient and Web3-oriented:

  • AI integration to more social media platforms. We already have AI on Twitter that ensures our MetaMask Support handle automatically replies to tweets likely to be targeted by scammers, reminding them of key wallet security principles. We’re looking at adding AI-driven, anti-scam features to other platforms, such as Discord

  • In the future, the bot will be able to create tickets based on information already gathered in the conversation, rather than requiring manual entry by the user or an agent. This will apply to escalating issues that require a higher level of technical support, such as wallet security, and benefits both users and Support agents.

The headline improvement to the bot flows has also been accompanied by plenty of behind-the-scenes tweaks that optimize function. Natural-language understanding, for example – a term describing the bot’s ability to interpret user responses – is continually being adjusted in light of performance data. The advanced AI behind NLU will add flexibility and resilience to the bot, better supporting users to self-serve.

Ultimately, we intend to roll out LivePerson integration for other Consensys products and services that experience high demand for support, such as NFT platforms. In the same way it has for MetaMask, we envisage that this solution will provide additional agile support for users, streamline user education, and ensure that accessing support doesn’t leave you needing emotional support. 

To access MetaMask Support and the conversation bot, head to Remember to never give your Secret Recovery Phrase to anyone, regardless of how legitimate they appear.