By Rahul Kayala, CEO of Echelon.
The enterprises that succeed with AI will be the ones with the clearest picture of how their business, the processes and systems underpinning it, actually run.
I've spent my career deploying enterprise platforms to hundreds of companies. Watching this AI wave, I keep seeing the same pattern. The programs that stall rarely do so on the quality of the AI model. They stall on whether the AI has a true understanding into the systems it's deployed against.
There is no one place where you can see how work actually moves across the dozens of systems a process touches: every approval, exception and workaround along the way. You can't redesign a process you can't see, and you can't hand an agent a process no one can describe. That end-to-end understanding is the most valuable asset an enterprise can give to AI.
That's why we're launching the IT Process Graph: a living record of how your critical processes and systems work and why they were built that way.
To understand why so few enterprises have this picture today, you have to look at what happened to IT over the last thirty years.
Act one: IT as builder
In the 1990s and early 2000s, IT became one of the most strategic functions in a company. Emerging software and the internet enabled a business to differentiate with systems, but only if it could build and implement them. IT was where that happened. The people who ran IT knew how the business worked on a fundamental level because they had built the systems it ran on.
Act two: IT as operator
Over the last two decades, IT went from a transformation organization to a service management organization. That wasn't a failure of ambition. The process of how software was built & implemented moved, so IT evolved with it.
Systems exploded. A company that once ran a handful of core systems now runs hundreds of SaaS apps. Every integration adds dependencies & complexity.
The business started buying its own systems. SaaS allowed a sales ops or HR team to buy software. Today business units control most SaaS spend, and IT only directly manages a small slice of it. The people who know how work flows in those systems don't sit in IT.
Critical systems became a specialty. ERP, CRM, ITSM and HCM platforms are so configurable they need their own specialists and integrators. The deepest knowledge of how your most critical processes are configured live with a partner, not with you.
IT stopped building and started running things: access, security, audits, vendor management, uptime. Important work, but operating work. All with no knowledge captured.
Long story short, IT kept oversight but gave away the deep understanding of the business and its processes and systems.
Act three: IT as orchestrator
AI is forcing IT to change again. At first it looks like a return to building, since agents have to be designed, deployed, governed and maintained. But IT can't give up any of the operating work either. The result is a job that sounds impossible, because it nearly is.
Running IT today means juggling five competing demands at once:
Keeping the lights on. The average organization now runs & maintains 300+ applications, most bought by business teams rather than IT, and many overlap in function.
Keep up with upgrades. Vendors ship faster and supporting older versions for less time, so upgrades and catch up work never stops.
Delivering AI, and proving ROI fast. Execs want AI. Procurement teams want proof of ROI before they sign. Meanwhile, the options keep multiplying: agentic workflows on top of existing systems of record, AI features from current vendors, and new AI-native providers.
Defend against AI-enabled attacks. In one campaign last year, AI ran most of the attack against 30 companies, including large tech firms, banks, and chemical manufacturers.
Do it all without more people. IT budgets are growing, but most of the new money goes to AI, not to the teams who run everything else. IT leaders are being told they cannot hire.
The bottleneck is throughput. There isn't enough capacity to get through the list. So CIOs have to become orchestrators: running existing systems, building the AI layer on top, and using AI to do both.
Trying to be an orchestrator while blindfolded
A new CIO inherits a decades old stack, with little record of how or why it got that way. Yet they're expected to keep it running, make big bets on what comes next, and modernize it quickly. To deploy effective AI, they need that understanding of how the enterprise's processes run.
Any IT executive knows this story, whether they're in their first year or their thirtieth:
Undocumented work. People and consultants leave, and what they did leaves with them.
Sprawling point solutions. Every team bought its own tools, so the same job gets done three different ways.
Mountains of tech debt. Organizations spend about 30% of their IT budgets just managing tech debt.
Just finding out what exists means sending in an Accenture army, uncovering why each system exists, what processes run on it, and why it's configured the way it is.
But the "why" is exactly what leaders need to make good decisions & what AI agents need to reason effectively.
The missing piece: the IT Process Graph
Your systems record what is configured. They don't record why. The reasons behind each workflow, customization and integration were never captured as data. Foundation Capital calls these missing records "decision traces": the exceptions, overrides and precedents that live in Slack threads, escalation calls, emails and people's heads.
Here's what that looks like. Your ServiceNow instance has an approval rule that sends every EMEA hardware request over $2,000 to a second manager. It was added in 2019 after an audit finding. The finding was closed years ago, the SI team that built the rule has rotated off, and now the rule is blocking your upgrade. The instance shows you the rule. Nobody can tell you whether it's safe to remove.
The IT Process Graph captures that why. It connects the technical view of your stack (what exists and how it runs) with the business reasons behind it, recovered from system configuration and logs, emails, Slack threads, consultant handover decks and the SharePoint folders your partners left behind.
And it doesn't start from today. Most thinking on decision traces focuses on capturing them going forward. For IT, the past matters just as much, and it isn't gone, just scattered. The graph reconstructs it across two time horizons:
Recent past: how the business has used each part of the stack over the last few quarters.
Key turning points: moments when the stack changed significantly, such as a new implementation, an upgrade, or a migration.
All of the exceptions: moments where your users have logged tickets or found manual workarounds.
From there, the graph keeps capturing as your systems change, so it never goes stale again.
With it, IT leaders can finally answer two questions they rarely can today:
- What exists in my stack, and how do workflows run across it?
- Why was it built this way?
Echelon started by helping ServiceNow and SAP teams plan, build and test changes rapidly inside a single system. As customers pushed us onto workflows spanning many systems, speed wasn't enough. We needed context. The graph is that context. Together, they give IT something it hasn't had before: one system of action across all your fragmented systems, with the knowledge to understand them and the throughput to change them at the speed of AI.
What the IT Process Graph unlocks
Query your enterprise stack. Ask how a process really runs, where it breaks and what a change will affect, and get an answer grounded in data rather than intuition or a consultant's memory.
Own your knowledge. Capture what your systems know today, recover what they knew in the past, and keep capturing it going forward.
Transform your stack, fast. Consolidate a sprawl of SaaS the business bought over the years into a simpler stack. Merge or rip-and-replace systems after an acquisition. Work that used to take multi-year consulting engagements happens far faster and far more cleanly.
Keep systems clean for AI. Agents are only as good as the systems they act on. With the graph, you can see which customizations, workflows and integrations still serve a purpose and which are leftovers, so you clean up before you hand the stack to an agent, and stay clean as it changes.
Taking back the map
For twenty years, IT has held accountability for systems it no longer fully understood. AI makes that untenable, and for the first time it also makes it fixable.
The CIOs who win the next decade won't be the ones with the most AI pilots. They'll be the ones who build the deepest understanding of their business processes and systems.
We'll share more about the IT Process Graph next week.
.webp)