# Astreya Pictor Platform

By Munawar Khel / Astreya / Product UI/UX Designer / 2015–18 / 6 min read

Published at https://munawarkhel.com/work/astreya/aes-pictor/

Shifted operations from reactive to proactive: real-time visibility, automated AI-driven detection, and instant alerts let teams catch regressions and bottlenecks as they emerge, shrinking anomaly-to-action time from days to minutes.

- Company: Astreya
- My role: Product UI/UX Designer
- Timeline: November 2015 – June 2018
- Platform: Web (Enterprise SaaS, desktop) with a responsive marketing site
- Category: Enterprise Product Design
- Status: New design, delivered
- Tools: Figma, Figma Prototyping
- Team: Cross-functional (Product/Design, Engineering, Data Science, and Data Engineering)

## Context

Astreya Pictor is Astreya's proprietary AI-powered process intelligence platform designed for IT and operations leaders. It ingests process data from external systems, automatically maps real-time workflow processes, highlights operational inefficiencies, and uncovers bottlenecks. Using built-in intelligence, it helps teams visualize exactly where work stalls across different operational environments, benchmarking behavior to deliver live, actionable insights and prescriptive alerts via Slack, text, and email. Its guiding promise is simple: you cannot improve what you cannot see.

The platform exists because Astreya's own delivery generates enormous volumes of process data. As a global IT managed-services provider resolving roughly 1.5 million tickets and managing 500,000+ assets a year for the world's largest technology companies, Astreya runs its most important work across ITSM, ticketing, and asset-tracking systems that produce static, backward-looking dashboards. Astreya Pictor was built to replace that retrospective view with live, interactive workflows, layering automated detection and AI-driven alerting on top.

The product is organized around four core capabilities: Data Integration (ingest process data from anywhere seamlessly into Pictor), Process Analysis (analyze data to uncover hidden patterns, visualize flows, and identify inefficiencies), Intelligent Detection (AI models that automatically discover what is slowing down the workflow), and Insights and Alerts (real-time operational notifications routed directly to the right stakeholders). My work covered the end-to-end design of the Astreya Pictor platform and its marketing site.

## The problem

Large organizations run their most important work, service delivery, asset performance, sales pipelines, and claims processing, as multi-step processes that span many teams and systems, yet they have almost no reliable visibility into how those processes actually behave. Running this on ETL scripts, BI dashboards, spreadsheets, and disconnected alerting tools created the same failures everywhere.

Teams knew how a process was designed but not how it ran in practice, where cases looped, skipped steps, stalled between handoffs, or split into unexpected variants. Anomaly detection was manual and reactive: KPI degradation like falling conversion, slowing deal velocity, or spiking resolution times was discovered only after the damage was done. Tooling was fragmented, with separate systems for ingestion, visualization, alerting, and task management and nothing connecting pipeline to discovery to statistics to detection to action, so every new model or data connection meant custom code and IT tickets. There was no standardized way to define a healthy ideal path and measure deviation against it. And insights reached the wrong people, in the wrong format, at the wrong time, so generic, context-free alerts bred fatigue and inaction even when a problem had been detected.

The design challenge was to turn this fragmented, engineer-dependent, reactive workflow into one coherent product that non-technical analysts could operate end to end, without losing the depth that data scientists and operations leaders needed.

## What I did

I designed Astreya Pictor as one integrated platform where data flows from ingestion through analysis, detection, and action without ever leaving the app. Everything shares a single navigation model, Company to Connections to Datasets to Analyses to Detections to Apps, and one custom design system, so each capability reads as part of the same product rather than a separate tool.

The platform is composed of a responsive marketing site plus five main modules, each framed by the specific failure it removes. The Integration Framework turns per-integration custom code into a no-code, self-service flow. The IQ Datasets layer gives analysts a governed, self-service on-ramp for shaping raw data into well-formed process data. IQ Process Analysis translates event logs into live, interactive workflow maps with node and edge statistics, variant comparison, and ideal-path overlays. The Heatmaps layer adds a directed-graph canvas with live per-stage metrics and three overlay modes for comparing actual against expected flow. App Packaging orchestrates everything downstream, from detection rules to conditional subscriptions to contextual notifications to automated prescriptions. And the Control Panel consolidates health, errors, audit trails, and recent activity into one scannable operational view.

Each layer addressed a specific failure mode surfaced during the work. Together they turn a set of disconnected analytics tools into one closed feedback loop: ingest, discover, benchmark, detect, act, and validate.

## Outcomes

- Shifted operations from reactive to proactive: real-time visibility, automated AI-driven detection, and instant alerts let teams catch regressions and bottlenecks as they emerge, shrinking anomaly-to-action time from days to minutes.
- Removed a chronic bottleneck by giving analysts independence, through self-service dataset configuration, no-code model integration via JSONPath mapping, and configurable detections and notifications that no longer depend on data engineers per iteration.
- Replaced weeks of manual investigation with minutes of interactive exploration using workflow mapping, variant comparison, and heatmap and ideal-path overlays that surface rework loops and quantify the impact of each deviation.
- Delivered the right insight to the right person at the right moment through conditional subscriptions, role-based prescriptions, and contextual notifications routed via Slack, text, and email, replacing generic, fatigue-inducing alerts with targeted, actionable ones.
- Established operational confidence at scale: the Control Panel's unified health, error, and audit views reduce time-to-detection and create clear accountability across a large ecosystem of interdependent systems, which matters most in compliance-sensitive environments.
- Gave the platform a consistent, scalable foundation through a single navigation model and a purpose-built design system that carries a coherent experience from marketing site to operational dashboard.

## Decisions

### One navigation model and one design system across every capability

What: I unified the entire platform under a single navigation spine (Company to Connections to Datasets to Analyses to Detections to Apps) and one purpose-built design system, rather than designing each capability as its own tool.

Alternatives: Treating ingestion, analysis, detection, and alerting as separate products with their own patterns, which is closer to the fragmented tooling the platform set out to replace.

Why: The core value proposition was a closed feedback loop from ingest to act. A shared spine and component library let data flow through the whole loop without context switching, and gave the product a coherent experience and a durable base to extend with new data sources, AI detections, and apps.

### No-code, self-service integration and dataset preparation

What: I designed guided API configuration with JSONPath response mapping and a dataset wizard that maps the core transactional fields process analysis depends on (Case ID, Timestamp, Process Step), with column-level statistics and inline augmentation.

Alternatives: Keeping model integration and data shaping as engineering tasks, or exposing raw configuration only technical users could complete.

Why: Analyst dependence on data engineers for every iteration was a chronic bottleneck. Making integration and preparation self-service dropped connecting a new model from days of development to minutes and let the rest of the platform always run on well-formed process datasets.

### Progressive-disclosure workflows with switchable overlays

What: For workflow analysis, I designed a directed-graph canvas with an adjustable deviation control and three overlay modes (Ideal Path, Common Path, Heatmap), plus a tabbed workspace of Graph, Statistics, Data, Variants, and Charts.

Alternatives: A single static process map, or dense statistics tables without a spatial view of where workflows concentrate and stall.

Why: Operations leaders needed to see where work clusters and how far reality drifts from the intended path. Progressive disclosure kept the common case readable while letting a user reveal uncommon paths on demand, collapsing weeks of manual investigation into minutes of interactive operational exploration.

### Detection packaged into publishable apps with role-aware, contextual alerts

What: I designed App Packaging so detection logic (regression, threshold, or AI pattern-based) is packaged into publishable apps, with conditional subscription rules, rich notifications built through variable interpolation and live preview, and prescriptions assignable to a person or executed by an automated bot via outbound webhooks.

Alternatives: Generic, broadcast-style alerts, and keeping routing, subscriber management, and follow-up actions in separate tools.

Why: Context-free alerts to the wrong audience were the reason detected workflow problems still went unaddressed. Routing the right insight to the right person in the right format, and wiring a follow-up action to it, replaced five or more disconnected tools with one workflow and cut mean time-to-action from days to minutes.

Tags: Product Design, Information Architecture, Interaction Design, Visual Design, Design Systems, Prototyping, Enterprise SaaS, Marketing Website, Data Visualisation