Recently, Andrej Karpathy, a co-founder of OpenAI and the former Director of AI at Tesla, posted something that should keep every executive up at night. He described giving an AI coding agent a paragraph of instructions to build a video analysis system. The agent worked for thirty minutes, hit problems, solved them on its own, wrote the code, tested it, and delivered a working system. What would have been a multi-day project for one of the world’s greatest software developers took AI half an hour. He didn’t touch anything.
That’s the capability curve. It’s going vertical. And it’s not waiting for your org chart to catch up.
The problem isn’t speed. The problem is depth. Your AI agent can see the entire internet. But it can’t see into your company.
The Four-Layer Test
There’s a simple diagnostic for this called The Four-Layer Test. Each layer represents how deeply AI is connected to your actual data, and most organizations are far lower than they think.
Layer 1: AI plus the internet. ChatGPT, Gemini, Claude, straight out of the box. The AI knows everything the public web knows and nothing about your organization. Your competitors have this same capability. There is zero differentiation here.
Layer 2: AI plus your documents. You’ve connected ChatGPT or are using CoPilot and they can see the files you upload, your meeting notes, reports, and spreadsheets. It’s better than Layer 1, but still limited to whatever one person remembers to feed it. If your AI strategy depends on employees manually uploading the right files or your SharePoint being your only AI source of truth, you’re here.
Layer 3: AI plus your Software-as-a-Service (SaaS) tools. Your CRM has embedded AI. Your support platform has embedded AI. Your marketing tool has embedded AI. Each one works well in isolation. None of them talk to each other. This is where most companies think they’re winning. They’re not – quite.
Layer 4: AI plus your corporate systems. Consistent definitions across departments. Governed access to unified data. AI agents that can take real action across business processes, not just generate text about them. This is where AI stops being a tool your people use and starts becoming infrastructure your organization runs on.
The Fragmented Data Trap
Layer 3 is the most dangerous position in this framework because it feels like progress.
At Layer 1, you know you’re getting generic answers. You calibrate. You apply judgment. You treat the output as a starting point.
At Layer 3, you have five different AI systems, each trained on one slice of your business, each giving you confident, specific answers. Your Salesforce AI says the customer is happy. Your Zendesk AI says they’ve filed three complaints this month. Your finance tool says they’re ninety days past due. Three systems. Three truths. One customer.
Fragmented truth is more dangerous than no truth, because either nobody questions it or nobody trusts it.
MuleSoft’s 2025 Connectivity Benchmark surveyed over a thousand IT leaders and found that the average enterprise runs 897 applications. Only 29 percent are integrated. That number hasn’t improved in three years. Companies keep bolting AI onto disconnected systems. They went from one AI that knows nothing about them to five AIs that each know one very specific thing.
And here’s what nobody talks about: the vendor incentives are designed to keep you here. Every SaaS company with an AI feature wants your data inside their walls. They’re not building toward your unified intelligence. They’re building toward their own stickiness. The more data you feed each tool, the harder it is to leave, and the less likely you are to connect it all together.
Layer 3 isn’t a technical problem. It’s a business model trap.
What Digging Deeper Looks Like
The specific platform matters less than the pattern. Organizations that reach Layer 4 share three things: a single source of truth that spans departments, governance that controls who and what can access it, and the ability for AI to act on that data, not just read it. The technology varies. The architecture doesn’t.
Guardian Life is a Fortune 500 insurer. Their tax department was running a multi-billion dollar operation on spreadsheets. Not a decade ago. Recently. Finance teams worked across disparate legacy systems, including decades-old PeopleSoft installations. They couldn’t see their own business. According to Oracle, they consolidated onto a unified cloud platform and deployed core financials in nine months. The tax department went from manual data collection to automated reporting. Financial Planning & Analysis (FP&A) could suddenly plan with accuracy. The win wasn’t a smarter model. It was removing the walls between data that already existed.
Dave & Buster’s had the same disease at a different scale. Data lived in siloed SQL databases across more than 200 venues. Sales, inventory, customer feedback, and operational metrics all sat in separate systems. They unified everything onto a single data platform. Databricks reported the results: 22% increase in promotional redemptions, 15% reduction in food waste, 30% savings on annual data platform costs.
Two industries. Two scales. Same lesson. The intelligence was already there, but it was trapped in silos.
The Gap is Compounding
IDC reported that among four thousand business leaders, companies with mature AI strategies, the ones that invested in integration and governance alongside their models, achieved 10.3 times return on their AI investment. The average company saw 3.7 times. That’s not a marginal difference. That’s a different world.
Every month your company spends at Layer 1, the organizations at Layer 4 pull further ahead. And there is no finish line here. AI is not an Enterprise Resource Planning (ERP) migration with a go-live date. The models change. The capabilities compound. The organizations that treat this as a project to complete are the ones that will wonder, in two years, why their competitors seem to be operating in a different reality.
Run the test. Pick one question that requires data from sales, support, and finance. Something like: which at-risk accounts should we save this quarter? If answering it requires humans to export spreadsheets and stitch slides together, you’re not at Layer 4. Find your layer. Start digging.
The tool doesn’t matter. The depth does.
Getting Started with AI for Connected Systems
If implementing an AI strategy across your company systems sounds intimidating, Doyon Technology Group is here to support. We have expert consultants on staff that can advise on the best tools and data governance strategy to help make your transition a success.
Connect with us today to request an intro call with our team and see what’s possible for your organization.
––––––
About the Author
Greg Starling serves as the Head of Innovation & Growth for Doyon Technology Group. He has been a thought leader for the past twenty years, focusing on technology trends, and has contributed to published articles in Forbes, Wired, Inc., Mashable, and Entrepreneur magazines. He holds multiple patents and has been twice named as Innovator of the Year by the Journal Record. Greg also runs one of the largest AI information communities worldwide.
Doyon Technology Group (DTG), a subsidiary of Doyon, Limited, was established in 2023 in Anchorage, Alaska to manage the Doyon portfolio of technology companies: Arctic Information Technology (Arctic IT®), Arctic IT Government Solutions, and designDATA. DTG companies offer a variety of technology services including managed services, cybersecurity, and professional software implementations and support for cloud business applications.