A few weeks ago, I had a conversation with an operations leader that’s been stuck in my head ever since.
She’s smart, experienced, and running supply chain for a Fortune 500 company. She’s also frustrated.
Her team is mandated to use 15 different corporate systems. Every single day, they monitor multiple shared inboxes, manually fill spreadsheets, copy-paste data between systems, and coordinate via endless email chains just to get work done.
Fifteen systems. Zero orchestration.
She called it “duct tape and spaghetti,” which is a perfect description of enterprise operations in 2026. And here’s what keeps me up at night: this isn’t an outlier. This is the norm.
Every enterprise is living this reality. Billions of dollars spent on “digital transformation,” and operators are still drowning in manual work, toggling between systems, and praying nothing breaks in the handoffs.
Now layer in the AI agent explosion, and the confusion multiplies.
Last week, I sat with a customer’s SVP of Strategy and Transformation. He opened with the question I’m hearing everywhere: “Why do we need your AI capabilities when we can build this in-house? And how do we even choose between all the vendors pitching us?”
Here’s how I framed it for him:
You’re being pitched by three categories of vendors:
The reality is that all of us, including FourKites, use the same foundational models – OpenAI, Claude, Gemini, Llama. The technology is increasingly commoditized.
So what’s actually different?
The answer is data. Specifically, the data that AI agents are trained on and have access to in real-time.
Every other vendor – and I mean every single one – works exclusively with data inside your four walls. Your ERP data, your TMS data, your warehouse data, your planning data.
FourKites has proprietary data outside your four walls.
We see real-time performance across 500,000+ trading partners in 176 countries – carriers, suppliers, manufacturers, 3PLs. We process 3 million daily events across the entire supply chain ecosystem. We know which suppliers consistently miss ASNs, which manufacturing facilities are experiencing quality issues affecting ready-to-ship dates, which distribution centers have capacity constraints, which carriers are reliable under which conditions – patterns that don’t exist in any internal system.
More than visibility data, our intelligent network provides the external reality your internal systems can’t see.
And here’s where it gets interesting: Many of our customers share the same trading partners, creating network overlap that enables use cases no one else can replicate. When we see the same supplier defaulting on commitments across multiple customers, when we detect manufacturing issues rippling through the network before they hit your dock, when we can benchmark performance and optimize based on network-level intelligence – that’s not something you can build in-house or buy from a horizontal platform.
Internal systems capture what you planned to do. FourKites captures what’s actually happening in the real world.
That’s the moat.
Which brings me to what we announced at Manifest: FourKites Loft.
Loft™ is our answer to the “15 systems duct-taped together” problem. It’s an AI-native agent development platform purpose-built for supply chain orchestration.
Here’s what makes it different:
Agent Operating Procedures (AOPs): When our AI agents execute work – reconciling PO discrepancies, routing supplier exceptions, managing warehouse capacity constraints, coordinating cross-functional approvals – we capture the decision trace. Not just what happened, but why it was taken, under what context, based on what precedent, with whose approval.
That decision trace doesn’t exist in your TMS or ERP. It lives in Slack threads, email chains, and people’s heads. Until now.
Digital Workers: Our AI agents don’t just surface insights. They execute work autonomously across your systems – the boring, repetitive, error-prone work that’s crushing your teams.
The Context Graph: Every decision captured becomes reusable precedent. Over time, this creates a context graph – the accumulated institutional knowledge of how your supply chain actually executes, becoming queryable and making your entire operation smarter with every decision.
Critically, we marry orchestration across your internal systems with real-time external data from our network.
When an AI agent needs to decide whether to escalate a supplier delay, it doesn’t just check your internal policies. It knows this supplier’s actual performance history across the network, sees real-time patterns from their other customers, understands precedents from similar situations, and has context no internal system can provide.
This is what “Internal Systems Meet External Reality” means in practice.
What most AI agent vendors don’t talk about is that deployment is just the beginning.
The dirty secret of AI agents is the ongoing maintenance burden. Managing model drift, evaluating new models, and improving performance are perpetual problems. Every subsequent improvement becomes another heavy engineering lift that deepens vendor lock-in.
Most solutions solve this by throwing engineering time at the problem. Which means either you’re building an internal team to maintain your agents (good luck hiring and retaining that talent), or you’re dependent on your vendor’s engineering capacity and roadmap prioritization.
We took a different approach.
Meet Sophie, our AI agent that builds and maintains AI agents.
Sophie is a developer agent within Loft. You describe a workflow in natural language – literally tell her your Standard Operating Procedure. Sophie then:
What used to take months and require dedicated engineering resources now happens in days or hours. And the maintenance burden? Sophie handles that too, continuously improving, adapting to model changes, optimizing performance.
This is how we answer “why not build in-house?” Because even if you build it, you’re signing up for perpetual maintenance. With Sophie, you get continuous evolution without the engineering tax.
I’ve been in this industry long enough to know the difference between “nice to have” and “mission-critical.”
Nice to have: A dashboard that shows you problems
Mission-critical: A system that solves problems autonomously before they impact customers
The way we become mission-critical is by mapping customer processes, identifying the manual work consuming hours every day, and using AI to eliminate it. Not by replacing systems – by orchestrating across them.
When operators can’t do their job without your platform, when the manual work is gone, when exceptions are handled automatically with the right context and precedent – that’s when you’ve moved from vendor to infrastructure.
At FourKites, we’ve spent a decade building the external reality layer – the real-time network intelligence that no one else has. Now we’re extending that into orchestration with Loft and Sophie.
We’re not just automating steps. We’re capturing how supply chains actually execute, turning institutional knowledge into durable precedent, and making every operation smarter with every decision.
If you’re attending Manifest 2026 in Las Vegas (February 9-11 at The Venetian), come see what we’re building. And if you’re an enterprise operations leader tired of duct-taping together 15 systems, let’s talk sooner.
The age of orchestration has arrived.
