Let’s say a service technician is learning a new hydraulic system. During classroom training, an instructor takes a unit apart, explains how it works, and walks through key procedures. The technician leaves with a binder of printed materials and a link to access training videos.

A month later, they’re replacing a valve on that system at a customer site. They flip through the binder and find the relevant procedure, but the written instructions are lengthy. Remembering the training videos, they pull out their phone and search for the right segment in a longer video. Eventually, they piece together the information they need to get the job done.

A few weeks after that, they encounter a problem that wasn’t covered in the training. Trying to track down the information they need will take too long, so they call a remote expert to walk them through it in real time.

Those are three different moments of need: learning something new, applying it on the job, and working through an unfamiliar problem. It’s the same person working on the same product, but each situation calls for different guidance — and the worker is left to navigate between different resources that live in different places.

AI offers a chance to give that worker one assistant across their changing needs, without repeating the fragmentation of today’s frontline tools.

Today’s frontline tools are fragmented

Tools for frontline work often address one part of the job. Across an organization, that leads to a patchwork of solutions that are fragmented in multiple ways:

• By moment of need. Training, work instructions, and troubleshooting are often handled by separate tools, even when they support the same task.

• By device. An experience built for a headset may not carry over to a laptop or phone, and vice versa.

• By function. Engineering, manufacturing, training, sales, and service teams each have their own mix of tools and resources, even when they're working with the same product.

• By content type. Many tools can’t work with 3D models at all, and there’s often a steep learning curve for the ones that do. Some tools are video-centric, while others make it hard to bring in video.

The worker has to navigate those divisions to get the help they need. Something useful in a training session may be difficult to find in the field. A troubleshooting guide may identify the component to replace, then send the technician somewhere else for the replacement procedure.

This makes it hard for workers to know what resources to trust. In a Forrester study commissioned by Staffbase, more than a third (35%) of U.S. frontline workers and managers reported encountering conflicting, outdated, or unclear information several times a week or more.

Adding AI to each individual tool doesn't solve that problem, especially if it reproduces the same fragmentation.

The same product knowledge supports every moment

Underneath those separate tools is a common foundation: the products, the parts, and the procedures for working on them. That knowledge starts in engineering. Training modules, service manuals, work instructions, and troubleshooting guides — all of those resources are built from the same underlying knowledge and data, just packaged differently for a different moment.

An AI assistant for physical work needs to understand that engineering foundation, including CAD models and spatial context. It can then draw on the same knowledge across training, guided work, and troubleshooting.

In a recent article, we looked at how AI can turn a technical procedure into conversational guidance. The videos, documents, and models that help explain the work can attach to the steps they refer to, so the right resource comes up when it's needed.

That's what separates an assistant built for physical work from a general-purpose one. A general assistant can search documents to find the right paragraph in a manual. But for physical work, finding an answer is only part of the help someone needs. An assistant also needs to guide them step by step, explain verbally and visually, know exactly which part an instruction refers to, and let them ask questions any time (even while their hands stay busy).

When the relevant assets and knowledge are organized into a structured procedure, one assistant can support every moment of need:

• A worker learning something new can put on a headset and walk through it at life-size scale

• On the factory floor, an operator can get step-by-step guidance by interacting conversationally with the assistant on their workstation PC

• When a technician takes a service call in the field, they can prop up their phone beside the equipment and talk through it with the assistant, hands-free, using earbuds or AI glasses

Why frontline workers need a single AI assistant for physical work

If the knowledge is the same in every moment, the assistant should be too. When each worker has one assistant that stays with them from training through troubleshooting, it can build on what they've already done: the questions they asked, the steps they worked through, and where they got stuck. When the technician returns to a task months later, the assistant can draw on that history.

That's better for the company, too. Experts review, approve, and maintain a shared procedure, and every moment of need draws on that same version. As a procedure gets used, the questions people ask and the steps where they get stuck provide insights on where to improve. Once an improvement is ready, it can be pushed live instantly and used everywhere.

That continuity is harder to maintain when each tool brings its own assistant.

That's the easy path for AI in frontline work: one assistant for training, another for work instructions, another for support, each working from its own copy of the knowledge. IT leaders are already worried about where that leads. In an OutSystems survey of 1,900 IT leaders, 94% said they're concerned that AI sprawl is increasing complexity, technical debt, and security risk. For the worker, five assistants result in the same problem as five tools.

From fragmented tools to one assistant

Frontline workers move between learning, doing, and troubleshooting as they work on the same products. One assistant can make the company's product and procedure knowledge useful across every moment of need, adapting guidance to the device that fits the task.

That's the idea behind Spark, Campfire's AI assistant for physical work. Spark is currently in Preview. Explore the Spark Preview to see how your own procedures could become conversational guidance.