Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
In a 77-year-old promotional items company Gold Bonds Inc.CIO Matt Price knew that adoption of generative AI wouldn’t come from deploying a chatbot. Employees needed to integrate AI into the work they already hated doing: complicated ERP management, document processing, and call tracking.
Instead of providing references, Price assembled a small group of “super-users” to surface Gold Bond-specific examples and educate the rest of the organization. They then integrated Gemini and other models into high-friction workflows, supported by sandbox testing, guardrails, and human review for everything public-facing.
The reward came in the form of behavioral change, not hype: daily AI usage increased from 20% to 71%, and 43% of employees reported saving up to two hours per day. “I wanted to take everyone on this journey,” Price told VentureBeat. “After we redefined some expectations, people started paying attention. Our adoption took off.”
ERP streamlining, product visualizations
Gold Bond, Inc. – not to be confused with the skin care company – is one of the largest suppliers in the industry. A $20.5 billion promotional products industryproducing personalized and corporate gifts for 8,500 active customers.
Orders, quotes and sample requests arrive via website, email, fax, etc., in every format imaginable. “So it gets very complicated,” Price said.
AI proved to be a natural solution. Previously, employees manually entered order details into the ERP. Now, Google Cloud ingests incoming documents and standardizes them, while Gemini and OpenAI extract and structure fields before pushing a completed purchase order into the system, Price said.
From there, Gold Bond grew into a pragmatic multi-model approach: Gemini inside Workspace, ChatGPT for backend automation, Claude for QA/reasoning checks, and smaller models for edge experiences.
"We’re pretty hesitant about using AI technology,” Price said. Gold Bond is largely set up like a Google store, with implementation and change management led by First Google partner Promevo.
Early wins included summarizing phone calls, drafting emails, and reviewing contracts. A more advanced use case is AI-assisted “virtual mockups” of branded products; teams use Reinvent to iterate on sample visuals before sending previews to clients, Price said.
Employees also use AI to generate Google Sheets formulas (including Excel-style XLOOKUP logic), while NotebookLM helps create an internal knowledge base for procedures and training.
Other ways Gold Bond uses AI internally:
-
Presentations: A job that used to take four hours now takes about 30 minutes, Price said.
-
Code auditing: Developers run NetSuite scripts, then use two models to review them before moving on to testing.
-
Research: Track trends and tactics of importers in response to tariffs.
AI also compresses planning at an early stage. “We go back and forth with the AI and come up with a high-level blueprint that we can then build for execution,” Price explained. “We get to concepts a lot quicker. We have a lot fewer meetings, which is great.”
To quantify the impact, Price’s team hosts Kaizen events – short workshops that document basic workflows and compare them with AI and automation-assisted versions.
To validate multi-LLM workflows, Gold Bond tests changes in a sandbox environment and runs quality assurance scenarios before deployment. “Our technical team, along with subject matter experts, approves before shipping changes or integrating them into production,” Price said.
Change management is essential
Adoption was not automatic: in a legacy company, change management was the job. “It’s just a little bit of apprehension, it’s something different,” Price said.
Most users start with Gemini because it’s integrated with Workspace, then move to ChatGPT, Claude, or Mistral when they need different features or a second opinion.
Price relies on a “nice little group” of about eight early adopters to test cutting-edge tools; once they find a use case, they train the rest of the team.
“You can’t just look at something like new software," noted John Pettit, CTO of Promevo. "We really need to change people’s thoughts and behaviors on this subject.
But even though Price’s team encourages widespread use, blind trust is not an option, he stressed.
Gold Bond added policies, DLP controls, and identity layers to reduce the use of shadow AI. It also uses LibreChat to centralize access to approved tools, enforce paid/approved use, and block certain models when needed.
Human intervention is mandatory: content intended for the public goes through approval and the results must be verified. “You have to set the right confidence temperature, but check,” he said. Even with strong prompts, the results still require verification. “You take the data, you can’t just take it and blatantly use it. »
For example, it will ask for sources and reasoning — “Give me all the works cited, where you get this data from” — and treats that verification step as part of the workflow, he said.
Price also warned against excess. “Agent solutions can only go so far: humans still need to be involved,” he said. “Some people have bigger visions than technology is capable of.”
His advice for other businesses: don’t overwhelm yourself with hype. Start simple. Start basic. “Provide detailed prompts, test them, play with them.”
