Leveraging LLMs to improve Klue's new user onboarding

At Klue, we recently held a week-long hackathon to prototype product developments 🤖

10 teams, composed of engineers, designers, and product managers participated in the hackathon, resulting in an inspired variety of prototypes and working demos that leverage large language models (LLMs) to improve and expand various parts of the Klue web app.

While these initiatives are all still in the prototype stage and might not all make it into production, they will certainly inform future product development and how Klue’s software can continue to pave the way as a leader in the competitive enablement (CE) category. This article describes one of Klue’s initiatives that intends to take advantage of an LLM, in order to build on our existing machine learning capabilities.

Defining the Problem

The foundation for most hackathon projects at Klue is defining a problem space, i.e. identifying the pain point that a new product development initiative can solve.

Team Love at 1st Sight (hereafter abbreviated as Team 😍), composed of Product Manager Jessica Zhang, User Interface Design Vira Romanko, and Software Developer Ian Sweetland, decided to tackle the problem of improving Klue’s First Time User Experience (FTUE).

New user onboarding is key to a positive first impression of an app, and early retention metrics. It’s crucial for a Software-as-a-Service (SaaS) company to provide users with a good FTUE in order to help those customers leverage the full value of their platform. UserPilot and UserReport both provide an overview of the importance of the FTUE for software companies.

Klue provides in-app tutorials, extensive help docs, and 1:1 customer success support to help customers get value out of our platform as quickly as possible. Still though, there’s always room for improvement in any company, and Team 😍 aimed to take Klue’s FTUE to new heights with a completely redesigned onboarding flow, designed to incorporate an LLM, providing the capability to create a customized onboarding plan for each individual user.

Here, we’ll describe their proposed solution in detail and the prototype functionality that Team 😍 designed and demoed during Klue’s hackathon.

Improving the FTUE

The hackathon team designed an onboarding wizard that would prompt users to provide the relevant information that influences what they need from a CE SaaS platform.

Then, that information would be passed to an ML backend model that uses ChatGPT, in order to prepare a series of customized steps those users can take to get value out of Klue quickly.

They devised a structure composed of five steps, each intended to provide an LLM with a relevant piece of information that would inform the creation of a custom onboarding plan.

Five things Team 😍 asked users about, to provide an LLM with a detailed prompt, are:

• The user’s role like Sales Rep, AE, Exec, PM, or something else
• Their goals like tracking CI news or understanding the market
• How familiar they are with their top competitors at present
• The market segment they focus on (i.e. by Product, Region, or Persona)
• The types of updates that interest them, like product launches or price changes

The team hypothesized that, with these key pieces of information, they could compose a detailed prompt to ChatGPT that would help the LLM identify the most important areas of the user’s CE strategy for them to focus on early in their use of the Klue software.

This could include recommendations about the competitors for which they already have the most valuable CI or where key info is still missing, which content to review in order to identify useful talk tracks, and which Battlecards would be most relevant to their team early on.

Team 😍 showed off a prototype of what this type of personal set up plan could look like:

Love at first sight indeed! 🥰

Into the Future

The prototype received an enthusiastic response, and folks at Klue are optimistic that this LLM-inspired initiative would have significant positive effects on Klue’s FTUE.

We expect that this effort would help new users understand what to expect from our software, and the specific ways it provides individuals and their teams with value, while also being capable of enabling multiple roles in organizations that compete in diverse industries.

Broader efforts to incorporate LLM capabilities into Klue will also play a significant role in Klue’s goal to be the ultimate platform for compete content curation. They will help users narrow down incoming data so they can prioritize what’s most important to them and their team. They will also enable users to drill down into specifics to glean more insights, enabled by competitive intelligence, that our CE software can provide.

In the long term, Klue can use LLMs to help improve our identification of relevant content, and put the right information in front of the right eyes, at all the right times. Klue already provides value to a large number of high-profile organizations in filtering the endless stream of data online into a manageable flow that is digestible by individuals and teams. Leveraging emerging technologies like LLMs will help us enable new users to recognize this value early in their experience with Klue. That benefits everyone 🚀

Learn more about Klue and building an effective compete program on our website.