The Problem
3E's customers keep their compliance data in 3E, but they run their operations in SAP, and keeping the two in sync is work that almost nobody can do. SAP talks to the outside world through interfaces that only a small pool of specialists ever learned, so when the data drifts, the only real option is to write a check to a systems integrator. Timelines for this class of work get quoted in years. The customer on the other side of this particular integration was floating a 2028 target.
"SAP uses complexity as a moat. But it is a commodity when you use AI. Most people just can't get their head around that."
Marcus Daley, CTO, 3E
Marcus had seen the opportunity for a while. But an open-ended exploration of a system nobody at 3E could fully describe, with no guarantee on feasibility or timeline, is exactly the kind of project that lives permanently on the list of things worth doing that no one ever gets to. There is no clean way to scope it, no confident way to resource it, and no obvious moment when it becomes anyone's priority.
3E did not hire us to execute a defined spec. Marcus brought a project he believed in and sponsored it directly, outside the normal planning process, because what it needed was not more engineers.
"I need a team who can take the MCP and not be afraid of not knowing what they're getting into, because they'll lean into AI to go figure it out."
Marcus Daley, CTO, 3E
The Solution
We realized early that the bottleneck on legacy integration was never engineering effort. The bottleneck is that the ground truth about the data inside these systems cannot be fully understood by humans — nobody can afford to read every table of a thirty-year-old system. With an agent, that constraint no longer holds.
Our approach is what we came to call a walkabout. We adapted it here to SAP, but we think of it as a template for integrating with any system that is queryable, poorly understood, and standing between you and a product you want to build.
| Step | Move | Why it matters |
|---|---|---|
| 1 | Expose the system of record through MCP servers that speak its native protocols | Turns an opaque enterprise system into an addressable surface an agent can work against |
| 2 | Give an agent read-only credentials and let it explore the full surface | Discovery becomes cheap and exhaustive rather than scarce and sampled |
| 3 | Capture what it learns as skills | Each pass compounds instead of restarting from zero |
| 4 | Make it argue for every conclusion, then backtest those conclusions against real data at volume | Separates plausible inference from verified ground truth |
| 5 | Put your own experts on the output, not on the system | Scarce human judgment is spent on validation and product decisions, not on archaeology |
Employing this methodology, we built a working end-to-end prototype for 3E and demoed it to 3E's prospective customer. In four weeks, that engagement moved from assessing whether implementation was feasible to signing agreements for the product build.
The Impact
We are still early in the development process and 3E has not yet taken this to market. But the prototype has already changed what is on the table.
The integration is now a committed part of 3E's go-forward business plans. What had been a perennially deferred idea — technically unproven, unscoped, and unresourced — is now a validated capability with a signed customer behind it. The work moved SAP integration out of the category of speculative exploration and into 3E's product and commercial roadmap, where it anchors both the specific customer engagement it unblocked and the broader SAP-adjacent opportunity 3E can now credibly pursue. The prototype did not just answer a feasibility question; it converted a maybe-someday initiative into a planned line of business.
The economics of the surrounding market are the reason that matters. The SAP services and consulting market is valued on the order of $250B a year. Much of that spend exists because the knowledge is scarce, not because the work is hard. This project demonstrates 3E's ability to build systems that bypass the integrator and deliver products that solve the end customer's problem directly, with no middleman in between.
The capability transferred to 3E's own team. Scoped into the project from kickoff was using our approach and pace to show 3E's engineers how quickly someone can become expert at something they knew nothing about the day before. The effects of that lead-by-example approach are already visible: a 3E engineer, handed the same MCP and one worked example, independently built a functioning SAP integration for a separate product in roughly five hours. The pattern did not stay with us — it is now something 3E can run on its own, on the next system and the one after that.
Where This Applies
Every industry has systems guarded by scarcity of understanding, with entire economies of integrators priced on that scarcity. If ground truth about any queryable legacy system is now days away for a small team with an agent and read-only credentials, the moat around esoteric knowledge is gone. What remains is knowing what to look for, how to validate it, and what to build on top.
MoE is a forward-deployed engineering firm. We embed with clients to understand their business problems, then build the AI systems that solve them. If you have a project like the one above, let's chat.