Workato AI Initiatives

Over the past year and half, I worked on various initiatives for Workato focused on incorporating AI onto the Workato platform.

Workato GO is an AI-based enterprise search that combines the power of search, AI chat, and agents to not only find documents and process content with AI, but take action.

Role & Team

I led research for Workato Go across multiple phases of development with various product stakeholders, partnering closely with PM, interaction design, and engineering to establish directional findings.

My manager selected me for the role as a top company initiative, competing for enterprise productivity on the level of Glean and Google.

Below I outline various projects that led to the launch of Workato GO, including strategic exploration, directional guidance, iterative testing, and QA initiatives.

Research Phases

The research took on various phases as Workato established AI presence in the platform:

  • Understanding task-based prompting styles across a diverse, technical population

  • Development of Copilot as an on-platform assistant for recipe construction

  • Emergence of the standalone product offering, Workato GO, to bridge search, AI, and app automations outside of the Workato environment

1 - Strategic Context

As LLMs became popular in 2023, leadership wanted to explore the addition of AI elements into the Workato platform. To inform user input, I partnered with my manager to run parallel efforts - a survey (me) and informal internal interviews (him). The survey provided the team with LLM prompt types for technical and non-technical audiences in the US, and informed the type of task. This work, combined with the interviews, impacted our approach to Copilot and Workato GO.

2 - Copilot Foundations

Post-launch copilot saw low adoption, and I joined the product team to understand usage patterns, styles, and to provide usability support of new Copilot features. I first aided our product designer by running a quantitative analysis of trigger types during recipe creation to restructure the optional start page for Copilot emergence. In parallel, I ran a study with Workato developers (new and experienced with Copilot) to understand various design elements like the new Sketches feature, prompting methods, and error recovery. I worked with a separate designer to develop concepts for future Copilot interactivity methods by uncovering unmet needs/wants.

This project impacted the product roadmap for prioritization of reliability/stability over new features, and moved a concept around optimization to the fore, which is now in production (after a small evaluative study).

3 - AI QA Method

As a fast-follow to my Copilot Foundational work, I independently developed a novel methodology that has impacted both our research and QA processes. The paper was published on the Workato blog, and details how the novel methodology works. Essentially, I used the Copilot study to determine variations in input style. I then ran a Snowflake analysis (SQL) to pull top recipe tasks, and had AI simulate user input and Copilot output in order to reach an idealized recipe state (JSON). This method of running a study with user inputs prior to design, as well as to QA AI features prior to launch helped inform both design and engineering efforts.

4 - Sales Personas

As a separate initiative unrelated to Copilot or prompting, I ran several studies both internal to Workato as well as with Workato users, and external to Workato with AEs to understand how AI may impact their current job functions. To do this, I ran interviews to go deep into a day in the life, even screenshotting their browser tabs, and shadowing them throughout different processes. This work led to the development of an AE customer journey map, and concept development for a widget based on sales cycle for a standalone app.

5 - Workato GO Foundations

Workato Go developed quickly, with engineering and design working lock-step to impact one another. Before a functional POC developed, I ran a qualitative study using designs made in Cursor to understand how users with familiarity with other enterprise search products may expect to use our product, and what could differentiate us from others. The output was a slide deck that was presented up to the VPs and C-level eng and product leaders to help understand core tenants of the design framework as well as model alignments and differentiators.

I worked with my designer to distill the findings into actionable design concept modifications.

6 - GO Early Access Program (EAP)

Workato GO evolved rapidly after that initial study, and in order to keep the team informed, I developed an EAP comprised of three core deliverables to update on a weekly cadence: rapid interviews with early adopters, ticket issue summaries on a customized dashboard built in Workato’s Workflow Apps platform with analytics from Jira, and design sprints (RITE of new prototypes). I concurrently worked with engineering to develop our analytics needs post-product launch to understand adoption, churn, and behavioral analytics.

The results led to both engineering and design bug fixes, UI enhancements, and product direction validation.