How a 77-Year-Old Company Got 71% of Staff Using AI Daily

According to VentureBeat, at 77-year-old promotional products company Gold Bond Inc., CIO Matt Price drove AI adoption by embedding it into high-friction workflows like messy ERP intake and document processing, not by rolling out a generic chatbot. He built a group of “super-users” to find Gold Bond-specific examples and train others, wiring models like Gemini and OpenAI into daily tasks. The payoff was stark: daily AI usage rose from 20% to 71% of employees, and 43% reported saving up to two hours a day. The company, a major supplier in the $20.5 billion promo industry, uses a multi-model approach with Gemini, ChatGPT, and Claude for different tasks, all backed by sandbox testing and human review. Early wins included automating purchase orders from chaotic formats and creating AI-assisted “virtual mockups” for product visuals.

The Secret? Workflow-First, Not Tool-First

Here’s the thing most companies get wrong: they start with the shiny AI tool and then go looking for a problem. Price did the exact opposite. He looked at the work people genuinely hated—manually keying order details from a jumble of emails, faxes, and web forms into their ERP—and asked how AI could simply erase that friction. The solution was a pipeline: Google Cloud ingests the documents, LLMs from Google and OpenAI extract and structure the data, and a completed purchase order pops out the other side. It’s not glamorous. It’s plumbing. But that’s where the real efficiency gains live. This is the core lesson for any enterprise, especially in industrial and manufacturing sectors where processes are king. You don’t need a chatbot to answer questions; you need to eliminate the questions altogether by fixing the broken process.

A Pragmatic, Multi-Model Kitchen

Gold Bond’s tech stack isn’t about loyalty; it’s about pragmatism. They use Gemini inside Workspace because it’s there, ChatGPT for backend automation, Claude for QA and reasoning checks, and even smaller models for edge experiments. It’s like having a kitchen with different knives for different jobs. This agnosticism is smart—it prevents vendor lock-in and lets them pick the best model for a specific task. For instance, using Recraft for generating product mockups is a perfect fit for a visual industry. And using AI to generate complex Google Sheets formulas? That’s a genius low-code way to empower non-technical staff. They even use AI to compress project planning, which Price says leads to “a lot fewer meetings, which is great.” Who can argue with that?

Change Management: Trust, But Verify

Adoption didn’t happen by magic at a legacy company. Price faced natural apprehension. His strategy? Leverage a small “cool group” of eight early adopters to test tools and land use cases, who then become evangelists to train the wider team. But alongside promotion came serious guardrails. They use Promevo (their Google partner) for implementation and change management, added data loss prevention controls, and centralized access with tools like LibreChat. The mantra is “human-in-the-loop.” Any public-facing content gets human approval, and outputs are always verified. Price even asks models for their sources—”Give me all the work cited.” He’s setting a “temperature of trust,” which is a brilliant way to frame it. You have to trust the process enough to use it, but you verify every output. This balanced approach is critical for scaling AI responsibly in any operational environment, from a promo products warehouse to a factory floor where reliable industrial panel PCs are the bedrock of daily data interaction.

The Big-Picture Lesson

So what’s the takeaway for other enterprises drowning in AI hype? Price’s advice is to start simple and basic. Don’t try to build an autonomous agent that runs the company. Find the tedious, repetitive, error-prone workflow that everyone groans about and use AI to streamline it. Measure the impact with real metrics, like their Kaizen events that compare old and new processes. And be brutally realistic about the tech’s limits. “Agentic solutions can only go so far,” Price cautions. The goal isn’t to replace people; it’s to make their jobs less annoying and more productive. That’s how you get from 20% to 71% daily usage. It’s not about intelligence; it’s about integration. And that’s a workflow problem IT is uniquely equipped to solve.