In this article, I’m using several AI-related terms interchangeably. The text relates to Seam - a chat-based SaaS unlocking the power of data insights and operations for business users. Seam is built on top of GPT-4, a Large Language Model. Large language models fall under generative AI (models capable of generating different types of content based on their training). Generative AI falls under a much broader category of Machine Learning, which refers to a subset of artificial intelligence algorithms capable of analyzing patterns in data and performing tasks without explicit instructions. All terms are encompassed by an umbrella of “Artificial intelligence,” referring to any computer or machine aiming to simulate a human-like thought process. While this language makes sense in the context of Seam, it’s worth noting that the terms are not interchangeable in general, and each refers to a specific, distinct software category.

Introduction

Living through a hyper-growth era of a tech company is a thrill like no other. Everything is moving fast, and there are countless opportunities for individual members and teams to make a significant impact. As a marketing or sales associate leading the charge in this growth, there are plenty of rewards to reap, but also a fair share of responsibility to handle. There’s just one problem. Modern operations teams need data to make good business decisions. Still, as startups move to scale-ups and focus on rapid growth, there’s simply too much data to keep up. Revenue, Marketing, and Sales Operations teams must juggle dozens of sources to surface relevant information about prospects and products. It’s not hard to imagine the uncertainty business users face when interfacing with large, scattered datasets. In practice, team members who can’t write SQL depend on data analysts and engineers to perform even basic operations. This inevitably slows the momentum and creates a non-trivial opportunity cost.

Seam’s mission to give business users the power of data

Data access shouldn’t be limited to technical users. Unfortunately, however, this is precisely what happens as companies reach a certain level of growth. Seam solves this problem by giving power back to the business user. The LLM chatbot interface allows them to surface, transform, and sync any data from their company stack back to the tools they already use (without writing a single line of SQL). 

What is the Modern Data Stack?

As companies have begun handling immense volumes of data in recent years, a new generation of cloud-based tools known as the Modern Data Stack (MDS) has emerged. These tools assist companies in gathering, storing, and analyzing their data more efficiently. With the cost-effectiveness and scalability of cloud computing compared to traditional on-premises solutions, the adoption of MDS has surged, becoming the standard for most tech organizations. There are many components to the modern data stack - each serving a different purpose:

• Data sources.This is where all company-wide data originates. Data sources typically include products (product usage and user behavior), websites (data about visitors), and customer relationship management tools (customer data). An average company can have hundreds of data sources, including marketing or financial platforms, internal databases, or social media accounts. (Examples: Salesforce, Hubspot, Google Analytics)
• Data integration tools. The role of data integration is to pull data from all sources across the organization and push them into a centralized data warehouse. These systems are called ETL (short for extract, transform, and load). (Examples: Airbyte, Segment, Fivetran).
• Data warehouse.The warehouse serves as the single source of truth for all company-wide data. Data is typically structured for the purpose of further analysis. (Examples: Snowflake, Redshift, BigQuery)
• Data transformation.At this stage, data is being cleaned, organized, manipulated, and finally processed. This stage of the stack makes it possible to extract valuable insights in the next steps (for example, by removing any errors, inconsistencies, or duplicates from the dataset). (Examples: dbt, Airflow)
• Data analytics.These tools analyze company data across multiple sources to identify patterns and trends. Data can then be visualized as graphs or dashboards so it’s easier to understand. Finally, data analytics tools can use that information to help businesses forecast future trends and, in turn, allow organizations to make better decisions. (Examples: Mixpanel, Amplitude, Looker)
• Reverse ETL.After data is analyzed, users need to return new insights to their original data sources (for example, a CRM). Reverse ETL tools are similar to Data integration tools but work in the opposite direction (instead of pulling data from sources, they push it into them). (Examples: Hightouch, Census)

Live demo

Interested in trying out Seam for yourself? The live demo is waiting.

Try live demo

Business users in the modern data stack

As I’ve touched on briefly in the opening paragraph of this article - business users need access to data. They often work with data analysts to extract critical insights that can influence their sales, marketing, and revenue operations decisions. 

How Seam removes complexity in place of pure insights

Seam's simplicity is its most significant advantage. Considering all the types of tools mentioned in the previous section, it becomes evident that interacting with data in its current form can pose considerable challenges for business users. The complexity is further compounded when we factor in additional tools for data governance and machine learning, which I’ve not mentioned for simplicity. As I’ve passionately reiterated across all Seam-related content, data is not and should not be reserved for SQL-savvy users only. At the same time, given the complex nature of this infrastructure, it's understandable why SQL proficiency often becomes a barrier to entry into the world of data.

Seam removes this barrier by offering users a simple chat interface. Any data can be retrieved in seconds with a simple natural language prompt. The same is true for actions that would typically require a data transformation tool. Operations like cleaning records or removing duplicates from a table can be performed with a simple request. Beyond retrieval and transformation, users can also create data models by simply asking AI to save a relevant query. They can then sync this data to their tools, replacing the need for a reverse ETL like Hightouch. 

Seam’s user personas

While Seam focuses on the business user as the primary persona, it’s more than capable of accommodating the needs of a more data-savvy user. 

What is a user persona?

Before diving deeper, let’s briefly define the term “persona” and why it is so crucial in building a successful product. 

It’s no secret that end users rule the software landscape. Tools that don’t cater to their needs fail to achieve product-market fit and have no chance for long-term adoption. Learning as much as possible about different types of users, their work, objectives, and pains is essential in creating software that will provide them value. Since SaaS products often serve diverse use cases, clearly defining multiple personas ensures that the needs of all users are represented and align with the company's business objectives. Investing time in building user personas is the first step in building user-centric products that address real needs and fill a gap in the market.

A user persona is a set of distinct characteristics that, based on real data and insights, define the needs of a subgroup of your software’s target users.

It’s worth mentioning that properly built user personas should always rely on data. In the early stages of developing new products, it’s common to build proto-personas based on anecdotal experience, assumptions, and ideas about how the target audience should look. This approach, however, isn’t reliable and doesn’t accurately represent the types of users who will use the product. 

Best practices for creating user personas

Searching the phrase “user persona” in Google Images returns almost exclusively charts containing photos, names, and demographic information like age, location, or job title. This approach, however, demonstrates an outdated way of thinking about personas. While the idea behind including additional details was to make the persona seem more “real” and easier to empathize with, it often introduced bias into the process while providing very little value. Assigning physical attributes like gender, age, or ethnicity might influence how you think about your users in the entire product development cycle. This can skew your perspective or lead to inaccurate conclusions and, eventually, bad product decisions. 

User personas should instead be focused on the person's needs and wants. Not who they are but what they want to achieve and what problems your product solves for them. It’s also important to note that personas are context-specific, so sadly, there’s no “one size fits all approach.” That being said, key questions to ask when building a persona are:

-list-normal-

• What are the user needs?
• What are their goals?
• What are their motivations?
  

-list-normal-

While ascribing physical attributes to your personas might be more harmful than beneficial, it’s still a good idea to differentiate between them. Think of names and visual representations that, while distinctive, aren’t specific to any geographic location, gender, age, or occupation. 

Finally, it’s important to remember that since personas are built based on data, they are meant to evolve as you gather more information. Verifying the understanding of goals and motivations with a broader sample of users is likely to uncover new insights. This cycle of verification and improvement is the best way to ensure your personas accurately reflect the reality of the end users. 

The business user

Business users across operations teams struggle with accessing crucial data. They’re often discouraged by the technical complexity of extracting even simple insights and don’t want to rely on data analysts to perform essential everyday tasks. At the same time, they need access and control over customer and product data to inform their decisions around customer success, sales, marketing, revenue operations, and more.

Business users want a simple way to answer their daily data-related questions. They want a companion that will help them understand the world of data and leverage its power without needing to learn SQL. Having a tool like Seam makes them more efficient and, therefore, more enthusiastic about their jobs. Uncovering new use cases with Seam is an exciting journey that they’re motivated to embark on after seeing the tool's initial value. 

The technical user

Technical users, on the other hand, are familiar with the company’s data infrastructure and navigate it confidently on a daily basis. They are generally open to all tools that add value to their workflow and make the team’s job easier, but they can be skeptical about letting non-technical users into their proven systems.

"Since technical users are the only ones who can access the data, they are constantly bogged down by service requests from business users. This means they have less time to do the strategic work that they want to work on because they are stuck servicing requests. The entire system is a struggle"

Outcomes - how we’ve made sure both personas can leverage Seam for their needs

One of the challenges of designing for multiple personas and use cases is maintaining a balance between the goals they want to accomplish within the product. In Seam’s case, the main interface is the chat; however, more advanced functionalities also need to be easily accessible in the appropriate context. 

When designing the interface, we mapped out two distinct user paths intersecting at critical action points. Our goal was to create a single interface that could be used by both personas. We’ve achieved this by allowing users to interact with Seam in multiple ways. Users of varying levels of data savviness can accomplish the same goal by following one of the parallel paths that best fit their use case. 

Writing a model with AI vs SQL

A great example of this approach can be seen in the way users create a new data model. They can choose between  “Ask AI,” which opens a prompt window, and “Custom query,” which allows power users to write SQL directly in Seam.

All sources at your fingertips

Both technical and business users can interact with the chat interface and process queries using natural language. This view, however, includes additional features, neatly stored in a side panel accessible in the top right corner of the page. A drill-down view containing the logic behind the provided answers is revealed upon clicking the' Details' button. While business users might appreciate the ease of use and intelligent suggestions the chat interface provides, power users can dive deeper to track the model’s reasoning - including data sources, filters, definitions, functions, and table joins.

One of our goals when working on this functionality was to show only context-relevant data. Since a conversation users are having with the model can contain hundreds of responses, we’ve tested different ways of adapting the content of the details panel to the response currently viewed by the user. The interaction we’ve chosen in the end attaches a separate panel to the top of each response. Since each panel is encapsulated in a box, it’s easy to distinguish between data sources from different queries.

We’ve rejected the fixed panel version that simply replaced the data information as the user scrolled through their responses. This view didn’t differentiate between queries adequately and felt less embedded within the conversation. Additionally, since the responses generated by the model vary in length, a side panel spanning the viewport's height wouldn’t provide enough real estate to display the entire logic for more complex queries, forcing the user to scroll within the component, and creating a confusing experience when browsing multiple responses in the conversation.

Access for business users - peace of mind for technical users

Seam’s ease of use lowers the barrier of entry to data analysis, which previously prevented operations teams from leveraging their company data to reach their goals. By utilizing a familiar chat interface and enhancing it with unobtrusive suggestions, we’ve designed software that is inviting and allows even hesitant users to get initial value with a single click. This enables operations teams to uncover business intelligence traditionally reserved for data analytics teams. The ability to run queries on top of companies existing tools ensures the integrity of their sources and systems, allowing technical teams to free up the time spent assisting business users without worrying about unintended changes, or disruptions in existing processes. 

Giving non-technical users the power over data results in a significant advantage from the business perspective. Companies like Redpanda (a Series C streaming data platform from San Francisco) and Apploi (a Series B workforce management platform from New York) implementing Seam are seeing massive benefits in the first months. For example, because Seam replaces multiple tools across several use cases, Apploi has saved six figures (Link to case study on Seam’s website) on tooling alone. Both companies have saved hundreds of hours annually by automating manual, non-scalable processes with Seam. Redpanda’s team was able to uncover new insights and recover over 700 hours annually. (Link to case study on Seam’s website).