Building AI Agents That Work With Your Real Data

‍A conversation with Zinier's Andrew Wolf on why most FSM AI will break down with a custom solution

During a recent conversation, I asked Andrew Wolf, Zinier's Co-founder and CPO, how Studio Z changes the game for AI deployment compared to traditional FSM platforms. His response got to the heart of why most field service AI projects struggle.

Deploy AI agents in any solution for any use case using Studio Z - our low-code/no-code builder. Studio Z's AI agent builder works with custom data where other FSM platforms fail.

No-Code AI Agents for Custom Field Service Data

"We want to be able to deploy AI agents in any solution for any use case using a no-code environment," Andrew explained. "That no-code environment for us is the AI agent builder in Studio Z."

I asked him what separates this from other FSM products? "Other FSM solutions simply aren't capable of deploying LLM-based agents on top of custom data and custom solutions. They may have some agentic capabilities working in really specific, narrow use cases, but once you start introducing custom data, then it doesn't really work."

The benefit of Zinier's approach centers on building "a repository of configurable agents that can be used to solve different types of use cases for different customers. Every client solution is different, every client has different pain points and business needs. As a result, the AI use cases to solve those pain points can be dramatically different."

Custom data breaks standard AI - while other platforms work only in narrow use cases, configurable agents solve different pain points for every client's unique business needs.

Why Standard FSM Platforms Fail with Custom Fields

The problem becomes clear when you move beyond standard fields. "An AI agent in a point solution would be configured to work with the fields within that task, but no further than that," Andrew noted. "That's where it would start getting into difficulty anytime there's custom fields."

For Zinier customers, "the level of customization is much higher. We've got custom modules, custom fields across all of our personas."

Traditional FSM platforms "just have some predefined chatbot or something, and it is limited to the scope of their out-of-the-box product. Here it can be augmented and work with your actual solution."

This extends beyond data queries to documentation. While other platforms might give you "a chatbot where you can ask questions on how to use the product on their out-of-box documentation, with documentation agent you want to be able to use it to ask questions about your custom operating procedures and manuals."

AI agents in point solutions work only with standard fields - custom modules and fields require repository of configurable agents, not predefined chatbots limited to out-of-box products.

Zinier's AI and Automation Discovery Workshops

Andrew described our systematic approach: "We run an AI and automation discovery workshop where we work with the customer to identify areas where they can drive automation and leverage AI. I put them together because automation could mean opportunities to leverage recommendations or other types of automation that are more rule-based, and where there are opportunities to leverage AI agents."

This distinction matters. During a recent workshop, a client proposed three "AI use cases" that revealed different solutions:

Use Case 1: "If I have a high priority task that comes in, I want to automatically assign it to a technician based on skills and proximity." Andrew's assessment: "That can be solved with auto-scheduling or with a specific recommendation that is completely rule-based."

Use Case 2: Automatically populate timesheets based on task activity and location. "This is automation, but it's actually how the timesheet product was built. It has built-in rules that automatically tag certain activity based on the status of the task."

Use Case 3: "Technicians to be able to have access to their operations and task manuals that are relevant to the task at hand and then be able to ask questions about it." This represents genuine AI territory.

"For things that are based on business rules, workflow automation and rule-based automation is better than using Gen AI firstly because Gen AI is still relatively expensive compared to other automation, it takes longer, and it's not 100% deterministic at this point."

Three-step process identifies real AI opportunities: Rule-based automation for scheduling, workflow automation for timesheets, genuine AI for operations manual queries and conversations.

Documentation and Data Queries

The documentation use case illustrates AI's sweet spot. You can provide manual access through "a workflow action that takes the user to a list of manuals specific for that type of task." But to augment that, "you could train the documentation agent on those manuals and allow technicians to have conversations about different best practices and how-to questions."

This becomes particularly valuable for newer employees. "These types of use cases are particularly beneficial to younger or newer employees as they're getting onboarded. Having the ability to ask questions when you haven't built up the expertise of a senior technician who's been doing this for a decade."

The business impact: "It can prevent a second truck roll from having to happen. If you don't have access to information, you don't know how to do something."

Data intelligence benefits

Every interaction provides learning opportunities. "If you notice that people are asking about a certain thing, you get much more data than if they just go read a manual. We might get a log that they read a manual, but we don't know what piece of information they're searching for."

Train documentation agents on your custom manuals - younger employees ask questions, prevent second truck rolls, gain data intelligence on information searches beyond simple manual access.

Natural Language Queries for Mobile Field Teams

Andrew sees particular value in simple data queries. "Maybe I just need the task ID of John's last task. If it gives it to me right away, then I don't have to go to task list view, filter, find John, do all this other stuff."

The mobile advantage becomes clear: "I'm a heavy user of the mobile version of ChatGPT at the moment using voice command, it's so easy. It's actually easier for me to push ChatGPT and ask a question than open up Google, type the question, hit enter, check web pages, and figure out where the answer is."

This applies especially in field service contexts where hands-free operation matters.

Voice-activated mobile queries beat manual filtering - get task IDs, data insights hands-free while in the field, easier than opening apps and typing searches.

Configurable AI vs. Predefined Solutions

The key difference comes down to flexibility. Traditional platforms offer "predefined" capabilities limited to standard use cases. Zinier's approach lets you "leverage the full power of AI and agents on a custom solution to meet those pain points."

The process mirrors our overall solution approach: "Similar to building out a custom solution for a customer, we're taking that same approach with AI and automation where we identify use cases and then use Studio Z and the AI agent builder to build out those custom use cases in their solution."

This is only possible "if you have agents that are configurable and configurable in a way that allows you to deploy it on custom data and on a custom solution."

Custom solutions need custom AI - leverage full power of configurable agents on your specialized workflows, not predefined capabilities limited to standard use cases.

The Implementation Process

Success requires matching problems to appropriate solutions. "It's our job to interpret the intention behind customer comments and figure out the actual pain point they're trying to solve, then solve it in the best way possible whether it's with Gen AI, recommendations, or customization to the product using the platform-based workflows."

The discovery workshop approach ensures you don't become "enamored with the technology" but instead "focus on the problem to be solved."

For field service organizations dealing with custom data structures and specialized workflows, this architectural difference determines whether AI becomes a practical tool or an expensive experiment that works only with standard data.

Match problems to right solutions - discovery workshops interpret customer pain points, solve with Gen AI, recommendations, or platform workflows based on actual needs, not technology fascination.

Ready to explore how configurable AI agents could work with your custom FSM data? Schedule a demo.