The Power of Personalization

Personalization is tailoring the tools, services, and goods workers use to their unique personalities. At the lighter end, this can be achieved by mapping characteristics that describe a person (e.g., age, seniority, geography, affinity). Think about how a product’s front-end may be different if the sign-up process elicits these traits (e.g., a different landing page for people from different countries, different music or film recommendations). 

While standard personalization has historically relied on broad characteristics or traits, hyper-personalization goes further by tailoring a user’s experience on categorically unique traits, elicited by their behaviors, interactions, and individual characteristics. Think of a content portal where the content delivery is calibrated to a person’s usage patterns. While we take that for granted as a part of the consumer technology experience, hyper-personalization holds similar promise for the enterprise. 

To level set, imagine the first day of work for an investment banking analyst. The young graduate steps into the office for orientation, where they receive a packet of information, including their login and password for their new work terminal. Is any of the software they boot up personalized to them? Is their Workspace data terminal aware of their persona? Chances are it isn’t, in large part because these tools were built to address a huge spectrum of workflows. This anonymizes the user experience to fit the product. 

The current instantiations of LLMs aren’t personalized, nor are most enterprise software solutions. Pull dynamics of information retrieval don’t cater to the individual user’s persona, while push dynamics such as Unprompted AI are predicated upon it. What are the other challenges towards implementing hyper-personalization at the enterprise? Let’s examine the consumer space for some clues. 

Returning to the example of a social media app, take a step back and think of how many gestures are compatible with an app. It ends up being a handful (swipe, double tap, hold, pinch…what else?). Now compare that to the number of functions or use cases in a spreadsheet program. Easily thousands. You can begin to see why collecting user interaction data from an enterprise software solution is far different than a social media app.  

Here lies the rub: consumer technology personalizes in a peculiar fashion. It does so by surveilling one’s activity, using big data algorithms to look for “generally right” patterns, and applying that. It’s a low risk and generally effective consumer interface. If a recommended video misses the mark, it is simply swiped away. While annoying, a consumer doesn’t abandon using the product over a bad recommendation. 

It is our belief that true hyper-personalization requires a knowledge graph on users, and most approaches towards knowledge-graph creation comes with some form of attention harvesting. Very few software solutions or companies that are in the business of selling standalone products develop these. Meanwhile, if a recommended article or reference is wrong in an enterprise level, it can lead to stranded adoption. It can’t be “roughly right”, it must almost always be right! 

This is a different solution space than what ModuleQ wants to deliver to the enterprise: greater salience of information and thus greater productivity. ModuleQ would rather you interact less with the information that we serve because it is the right information, not needing to hold your attention. Our guiding light is to deliver the right information at the right time to the right person, which is why our approach to personalization is the inverse of most consumer tech. 

Privacy is more central to the way we organize our workspace than the way consumers interact with their phones. Data scale is another. Even a large company taps out in the thousands or even tens of thousands of employees. The ability even then to pool them to make valid inference on usage is debatable (how does a junior employee’s use of Excel map with an MD’s? Sales vs Trading?). Social apps are sub-scale when they have hundreds of millions of users. As such, it is a drastically different solution space from a product design standpoint. 

Calibration can also be a problem in trying to deliver great personalization for the enterprise worker. Understanding the validity of signals across their specific workflows will be challenging. Knowing which contacts and tasks are important to them is also a challenge, especially given the unique ways different individuals and companies utilize them.  

Take for example calendar information. At ModuleQ, we tap into the calendar resources of our investment banking clients to enrich aspects of their daily workflows, such as meeting prep. This delivers a degree of personalization around the type of information we send to our users.  

But calendars are used differently by different people. Some supervisors have meetings on their calendars simply for visibility, while others only list meetings which they’ll be attending. This means the validity of the contextual information around a calendar invite can differ and likely needs to be enriched by other contextual clues such as email, user habits, and reinforcement learning.

There are a few ground level concerns for enterprises looking to adopt personalization to improve workplace productivity. As we mentioned earlier, data integrity and security are key for enterprise deployment. While most consumers are willing to breeze through the security and privacy concerns regarding their data when clicking a Terms of Service checkbox, most enterprises (especially in regulated industries) pay close attention to their data and security. 

To that end, having solutions that can sit on-premise greatly improve the ability to apply personalization. AI tools such as PDF remain securely within the security perimeter of the organization. Embedding technology on-prem is thus a key driver of enrichment. ModuleQ has partnered with trusted brands in the financial services space, notably LSEG and Microsoft. This allows us to deliver on data privacy and integrity while still having access to the rich back-end, compute, and market data that can push personalization from a pipe dream to a daily reality. 

Becoming Iteratively Better 

What gets us really excited about personalization is the ability for improvement to be far faster than a linear trajectory. As more data gets integrated, as more tools connect the dots across that data, and as greater reinforcement fine-tunes that personalization, the productivity boost delivered to the user will only accelerate. And so, without some degree of user feedback, personalization can get its tires stuck on many a pothole.  

To fine-tune our process without hampering our users, ModuleQ provides a simple thumbs up/down associated with each of our information delivery cards, ensuring that a user’s preference can help tune and personalize their experience. As more usage patterns emerge through these reinforcement signals, the product’s ability to deliver greater relevance improves, creating a positive reinforcement loop. 

The Promise of Hyper-Personalization 

A future of hyper personalization at the enterprise is in sight, and it is being driven by advancements in AI technology, improving tools and knowledge graphs of users. It’s important that the industry continues to press on personalization, rather than retreating to the old modality of prompting to access AI. Enterprises must embrace the challenges to reap the rewards of personalization. It must be done in an ethical manner which empowers the worker rather than a surveillance apparatus. 

We recognize that it’s a thorny problem. It requires the synthesis of subject matter expertise in the domain with cutting edge technology. It needs a different design ethos that is more “bottoms up” rather than “top down.” It needs to handle security and privacy. If done right, it may be the key to unlocking the next wave of productivity growth at the enterprise level.  

At ModuleQ, we will continue to trailblaze in the field of hyper-personalization for knowledge workers. We will continue to adopt a safe and responsible approach; one that does not use surveillance techniques but takes a persona first approach to improving the work experience of knowledge workers. We’re excited to deliver a hyper-personalized work experience, unlocking the next wave of productivity.