Designing a Data Source Discovery App — What’s New in v2.26

When I wrote last August's product review, I described the kDS Data Source Discovery App as ready for enterprise beta. Eight months later, version 2.26.8 reflects sustained development in the areas that matter most to daily users: workflow clarity, respondent lifecycle management, project structure, and better visibility into the organizational hierarchy. The admin dashboard is the operational hub of the app, so its evolution is a good proxy for the product itself.

Contacts & Interviews: A More Complete Workflow

The Contacts & Interviews panel has grown from two actions to four, making the end-to-end admin sequence visible as a coherent pipeline: confirm role → generate interview → send interview email → collect responses.

Manage Role Confirmation Email

Before a respondent sees an interview, there is now a formal role confirmation step. Since the kDS interview engine tailors questions to each respondent’s role and industry, assigning the wrong role means the AI-generated questions will miss the mark. The confirmation step catches that misalignment before an interview is generated, protecting the quality of everything downstream.

Manage Interview Email

The interview email workflow has been promoted from Quick Actions to a first-class entry in Contacts & Interviews, where it belongs. It governs the outbound email carrying each respondent’s secure, token-authenticated interview link. Previously, workflow steps were distributed across three panels. Now they’re in one place.

Optimizing Interview Generation

While testing, we discovered that the interview question generation process was timing out mid-run because the web server had a hard limit on how long a single request could take. The fix was to move the question generation into a background thread so the server responds immediately while the work continues behind the scenes. Users never see this change directly — but they notice its absence when a generation silently fails and leaves a respondent with an incomplete question set.

Summaries and Analytics: The AI Pipeline, Now with a Dashboard

The three-stage AI pipeline transforms raw interview answers into actionable intelligence: topic-level summaries with data flow mappings, an executive-ready report surfacing key themes and strategic recommendations, and structured data extracted into relational tables. The pipeline turns unstructured SME conversations into a documented picture of how data moves through an organization.

At the individual interview level, admins can now explicitly choose between a First Summary (initial analysis from interview responses), a Follow-up Summary (builds on initial findings using follow-up responses), and an Executive Summary (cross-cutting synthesis for decision makers). Processing takes 2–5 minutes; the app redirects automatically when complete.

The summarization pipeline has also been optimized since last year — prompt structure, response parsing, and relational schema routing have all been refined based on real engagement data. That work deserves its own post, and I plan to cover it in a future installment.

On the admin side, the key change is the separation of generation from consumption. Version 2.26 adds a distinct Analysis Dashboard with two dimensions: Role-Based Analysis, which surfaces cross-role patterns and technology usage across similar job functions, and Business Unit Analysis, which identifies cross-functional dependencies at the organizational unit level. Triggering analysis and reviewing results are now two clearly labeled steps.

System Management: Projects and Respondents

Project Data

Project Data management formalizes something always implicit but never surfaced in the admin UI. A discovery engagement has a scope, participating organizations, a timeline, and discrete phases. The Project Data function defines that project envelope, giving better context for organizing interviews and analysis outputs by engagement — and laying the foundation for multi-project reporting down the road.

Respondent Data

Respondent Data management consolidates the respondent lifecycle — invite, role confirmation, interview completion, reassignment — into a single screen. Previously this required navigating through several parts of the app. The panel also shed redundant entries from v2.25: industry classification now happens automatically at org registration time, and client context lives under Project Data.

Quick Actions: The Functional Hierarchy Diagram

The new View Functional Hierarchy button opens an in-page modal with a live Mermaid diagram of the full organizational hierarchy — parent organization → subsidiaries → business units → roles → respondents. Previously, the only way to see this structure was to navigate the Configure Data screens one level at a time. The modal fetches a live rendering from /hierarchy without leaving the dashboard, with graceful error states for engagements still being configured.

Where Things Stand

The through-line in v2.26 is clearer workflow sequencing and better separation of concerns. The underlying capability — interview engine, analysis pipeline, security architecture — was already there. This release cycle adds better administrative ergonomics around it. Beta feedback has consistently pushed toward making the app feel like a guided workflow rather than a collection of tools, and that remains the direction. Next up: the respondent-facing experience and shareable analysis output formats for stakeholders who need results without logging in.

If your organization is working through undocumented data assets, reach out at talk2us@keeshinds.com. We’re actively onboarding select beta partners.