More than a Chat Bot: An AI Support Ecosystem for Industrial Plant Engineers

Manuel Ortiz
Nov 11, 2025
5 min read

Updated: Nov 28, 2025

2024-2025

A zero-to-one design strategy for an AI ecosystem, built to empower industrial engineers and scale operational expertise.

AI-Enhanced Research - Systemic Design - Strategic Design - UX Design

Eye-level view of a designer's workspace with sketches and a laptop showing wireframes — *A simplified high-level diagram for the AI ecosystem*

Summary

My Role:

Lead Designer & Strategist, responsible for the end-to-end design of a complex, zero-to-one AI product—from initial ambiguity to a validated service blueprint.

The Challenge:

Plant engineers lacked immediate, trustworthy access to critical operational data, leading to costly, multi-hour downtime. My challenge was to design a generative AI ecosystem that would complement existing support channels, increasing user efficiency and reducing support costs.

The Outcome:

Delivered an initial MVP that reduced human-handled support tickets by 10% on first release.

Defined the complete product strategy and service blueprint for a multi-faceted AI ecosystem, aligning competing stakeholder and user needs.

*Plant Engineer doing a software activation on a machine. AI generated image.*

NDA Disclaimer: This Project is protected by an NDA. To respect client confidentiality, this case study focuses on the strategic process and uses abstracted, anonymized visuals.

My Role and Strategic Contributions

My contributions focused on strategy, leadership, and de-risking this highly ambiguous project:

Facilitating Alignment: organized a three-day alignment workshop with 10 cross-functional stakeholders (engineering, business, experience) to define the core problem and build consensus.
Service Blueprinting: I mapped the "To-Be" journey, ensuring the transition from self-service (AI) to human support (Phone) felt like a single, intentional conversation, not a handover.
Product Strategy: Formed the end-to-end product strategy, translating the vague “let’s use AI” brief into a concrete, phased product roadmap.
AI-Led Research: Pioneered a research synthesis process by leveraging a custom-trained AI agent to analyze over 20,000 data points (surveys, interviews, usage stats) and extract actionable, evidence-based insights.
Prototyping: Drove the validation process through rapid prototypes with Maze and live user demos to de-risk assumptions and build stakeholder alignment.

Collaborators

This project was a deeply cross-functional effort that combined design, engineering, and business strategy.

Product Manager: Defined key success metrics and ensured alignment with executive stakeholders throughout the project.
Business Analyst: Facilitated information gathering and maintained business alignment and clarity at every stage.
Lead Engineer: Partnered closely on feasibility mapping, technical validation, and data access for AI prototyping.
Design Tech Director: Led the design team effort and facilitated key stakeholder conversations to align on vision and execution.
Data Scientist: Developed the initial proof-of-concept AI models and supported interpretation of system behavior and model accuracy.
Visual Designer: Translated complex AI workflows into intuitive, human-centered interface patterns derived from wireframes and user flows.

Each collaborator brought unique expertise that enabled us to move quickly from concept to validated MVP while maintaining a strong user focus throughout.

Context

The Business Problem:

The client was facing a crisis of value. Surveys showed that while customers valued the expertise of support technicians, they found the price "steep" and the digital experience lacking. The subscription felt like "car insurance"—essential for emergencies, but a waste of money day-to-day.

The User Problem:

The "Doom Loop." Our data showed users getting trapped in circular navigation, bouncing between irrelevant FAQs and a search bar that returned "simplistic or vague" results.

The Scale:

We weren't guessing. We had to synthesize a massive, fragmented dataset: 18,235 survey responses, 1,500+ loyalty comments, and 181 deep-dive interviews across varying timeframes and regions.

The Design Journey

Chapter 1 — Finding Clarity

The Challenge

We started with an overwhelming amount of information. Our first step was to understand the real problem, not just the assumed one.

The Action

I consolidated all available interviews, surveys, complaints, and usage metrics and used them to train a custom AI agent. This allowed me to synthesize thousands of data points in days, not weeks, and identify clear patterns.

The Key Insight

We discovered a psychological split. Users viewed the service as "safeguard" (like insurance) for downtime. When they did use it, they were often expert engineers who had "already done the basic troubleshooting." They didn't need a basic chatbot; they needed expert answers fast. The current search tool was failing them ("40% of the time I don't find what I'm looking for").

Chapter 2 — The Messy Middle

The Challenge

A key challenge emerged: stakeholders were heavily invested in a chatbot solution, but our research showed users were frustrated with all existing support options. They didn’t want another channel — they wanted their existing tool, the knowledge base, to actually work.

The Action

I used the data to push back. Our research explicitly stated: "AI should be a back-end enhancement to search... not a separate tool." I ran a rapid prototype test comparing the Stakeholder's "Avatar" vs. my "Unified Search."

The Key Insight / Pivot

The test results were definitive — the AI-enhanced search had an overwhelmingly positive response. This data allowed me to pivot the stakeholder conversation from “either/or” to a hybrid strategy:

Focus on the AI search as the primary solution.
Use the chat interface only for a niche, high-friction task it was perfect for: software license activation.

Chapter 3 — Getting to the Solution

The Challenge

We had validated our core concept, but the research uncovered many other problems. We had to prioritize and create a long-term plan.

The Action

Recognizing we couldn’t solve everything at once, I created a strategic roadmap to phase the rollout. My primary role became supervising the information architecture, ensuring that all new components — the search, the chat, and a future smart guide — felt like one cohesive, intelligent ecosystem, not a collection of separate tools.

The Key Insight / Validation

The final validated search concept achieved a 90% acceptance rate in final user tests and, upon initial release, correlated with a 10% reduction in human-handled support tickets. This hard, data-backed proof gave the client confidence to fund the full, multi-phased roadmap.

The Solution and Design Rationale

Feature 1: The AI-Enhanced Search

This interface directly solved the core trust and findability problem. By summarizing complex technical documents into a single, AI-generated answer (with citations), it empowered engineers to find their own solution in seconds, not hours.

Feature 2: The Niche-Case Chatbot

Instead of a generic “How can I help?” bot, we designed a hyper-specific chatbot to solve one high-friction task: software license activation. This turned a major user pain point into a simple, automated success story.

This also solves the "Doom Loop." If the AI answer isn't enough, this feature captures the user's context (machine type, error code) before connecting them to a human, ensuring they reach the right expert immediately.

Impact and Reflections

The Impact

10% Reduction in Support Tickets: The initial MVP release of the enhanced search function immediately reduced the volume of human-handled support tickets by 10%, with projections to reach 30% as the full ecosystem rolls out.
Secured Full Project Funding: The data-backed, phased approach secured stakeholder buy-in and funding for the complete, multi-year AI support roadmap.
Established a New AI Strategy: This project set an internal precedent for how the client validates and implements AI functionalities — shifting from “AI for its own sake” to validated, user-centered AI solutions.

Personal Reflection

This project was a masterclass in senior-level negotiation. It proved that the best way to manage stakeholder expectations is through user-backed data. By prototyping both their idea and our research-led alternative, I wasn’t just pushing back — I was facilitating an evidence-based decision. That, ultimately, is the most collaborative and persuasive way to lead.

One of the most important lessons I learned early in my career is that design must start with understanding the user. Without a clear grasp of who the users are and what they need, even the most visually appealing product will fail.