John Davie, Founder & CEO of Buyers Edge Platform and CEO of CollectivIQ, leverages data and AI to drive smarter enterprise decision-making.
Since starting my career nearly 30 years ago as the founder of food procurement company Buyers Edge Platform, the technology challenges that have crossed my desk have typically been infrastructure-related, largely tied to whether our systems could handle anticipated growth. However, very recently, and almost in an instant, a new and perhaps even more critical challenge emerged: Our 1,200 employees were using different, sometimes overlapping AI models, each with their own workflows, pricing and dangerous levels of inaccuracy.
When AI became a serious factor, I expected the usual rollout consisting of pilot programs, internal champions and a lot of convincing. None of that happened. Instead, people just started using it. As a technology advocate, I was all for it. Within a few months, teams were drafting content with it, running research through it and handing off repetitive tasks to AI models.
Everyone was excited about a different aspect of AI, and most employees went all-in at what seemed like lightning speed. One person swore by one model, while someone two desks over insisted a competing platform was better for the same task. A few liked how Gemini wrote, while some developers trusted Claude for coding.
From where they sat, each person had found what worked for their specific role or set of tasks. From my viewpoint, thousands of quiet decisions were being made with no one tracking the outputs, accuracy and how they showed up in critical business decisions.
For a while, that approach seemed to be working. People were productive, and nobody was complaining. As the usage climbed, however, I started asking questions nobody could answer. How much were we really spending across all of these tools? Which ones were earning their keep? How carefully were people checking the accuracy of outputs they were leaning on to make actual decisions? And the one that nagged me the most: As employees were gaining valuable insights, where was the visibility for long-term value and growth?
The prompts people refined, the workarounds they discovered and the small process improvements that actually moved the needle lived and died inside a third-party tool. We had no way to capture it, pass it around or build on it.
Around the same time, I started seeing the same dynamics surface across the broader business community. A 2024 study by Microsoft and LinkedIn found that 78% of surveyed employees who used AI relied on personal accounts and self-procured tools rather than enterprise-approved licenses, creating isolated workflows and outputs that remain siloed from the broader organization.
Leaders were also growing increasingly concerned that critical business decisions were being made on inaccurate or unverified information. The consequential errors resulting from AI hallucinations cost organizations $67.4 billion globally in 2024, according to a report published by Four Dots (citing AllAboutAI research). We were generating something valuable every day and letting it evaporate. The challenge had shifted from getting people to adopt AI to figuring out how to govern, measure and scale that adoption in a way that minimized risk and created lasting value.
The financial side of the equation was also becoming more difficult to ignore. AI vendors were charging steep markups on top of model costs, often bundled into per-seat licenses priced at $20 to $40 a month per user, regardless of how much an employee actually used the tools.
Paying that high of a rate across our company, including for employees who rarely touched the tools, was hard to justify. What we needed was cost-effective pricing tied to actual usage, with the visibility and controls to keep spending in check. Solutions with that kind of flexibility and oversight weren’t available.
All of these considerations are what eventually pushed us to build a solution of our own, CollectivIQ. We wanted a way to reduce the fragmentation we were seeing across the organization, help employees validate often inaccurate information more efficiently and ensure the knowledge being developed every day didn’t disappear inside individual tools and accounts. The goal was to create a system that could bring together multiple perspectives, increase confidence in outputs and help our organization learn from itself over time.
Building our own AI platform solved those problems for us, but the lessons behind it apply far more broadly. If I could go back in time and apply what I know now to early AI deployment at Buyers Edge Platform, I’d establish three priorities from the very beginning:
1. Understand How Employees Are Already Using AI
Foster an open dialogue where employees can share what models they rely on for specific tasks and where they’re seeing limitations. Leaders benefit from visibility into which tools are being used, how they’re being used and where they influence important business decisions.
2. Treat Every AI Interaction As An Opportunity To Build Institutional Knowledge
The most valuable innovations come from the refinements my employees make along the way, from prompt building to streamlining workflows. Leaders should create opportunities and forums for employees to document and share successful use cases and to build on each other’s discoveries across the organization.
3. Demand Governance From The Start
Employees need the freedom to experiment and find the best solutions for the business, but leaders need the visibility, cost controls and accountability that make AI adoption sustainable over the long term.
AI is moving faster than its predecessors, but the sequence is the same. Like many transformative technologies before it, AI’s success will depend less on the sophistication of the tools and more on the people, systems and processes underpinning them.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
