Twenty-Two Years, Including the Parts That Didn’t Work

TLDR

Twenty-two years of enterprise technology work produced two things worth paying for — pattern recognition about what actually goes wrong, and the scars from the times we learned it the hard way. AWS independently validated the first. Their funding programs let us prove it at our expense before it becomes yours.

Every serious prospect eventually asks the same question, usually about forty minutes into the second meeting, and usually more politely than this: why should we believe you?

It is the right question. The market is full of firms with credible-looking decks and a confident answer to everything. So here is ours, in the order we think it should be weighed — starting with the part that took the longest to earn and ending with the part that costs you nothing to verify.

First, the knowledge — and where it came from

We have been doing this for twenty-two years, across manufacturing, healthcare, financial services, construction, hospitality, logistics, government, and education. Six hundred clients. Twelve hundred projects. Those numbers matter less than what they consist of.

Twenty-two years means we did not arrive with the current wave. We built on physical servers, then virtualized them, then moved them to the cloud, then optimized the cloud bill after everyone discovered that lifting and shifting is not a migration strategy. We integrated systems over flat files, then SOAP, then REST, and now MCP. We watched three generations of “this platform will unify everything” arrive and settle into being one more system that needs integrating.

What that buys a client is not nostalgia. It is pattern recognition. When a project is going wrong, it is usually going wrong in a way that has gone wrong before — and the tell is almost never the thing the status report is about. We have found the needle in the haystack often enough to know which haystacks tend to have needles. We have scaled platforms so that companies could scale, which mostly means we have been on the receiving end of what breaks first when volume triples, and it is rarely the part anyone hardened.

Second, the parts that didn’t work

We have also had setbacks and outright failures. We are willing to say so, and we think you should be suspicious of any firm our age that isn’t.

We have underestimated how much of a data migration is really a data-quality project, and paid for the difference. We have built exactly what was specified and watched adoption stall because the specification described the org chart rather than how people actually worked. We have taken on integrations against systems whose documentation turned out to be aspirational. Each one cost us something, and each one is now a question we ask early — which is the only real return on a failure.

This is not a confession for its own sake. It is the practical reason our assessments look the way they do. When we tell you a phase will be harder than it looks, it is because we have been wrong about that specific thing before, on our own money.

Third, someone other than us saying we know what we’re doing

Self-assessment has obvious limits, which is why third-party validation matters. Over twenty-two years we have built partnerships with more vendors than are worth listing — Microsoft, Google, Zoho, ServiceNow, and dozens of smaller ones. Three are worth naming specifically.

We are an AWS Advanced Tier Services Partner holding dual AWS AI Competency — both Agentic AI and Generative AI. Fewer than 65 partners worldwide hold both. That designation is not a logo you request; it is an audit. AWS examines delivered customer work, architecture decisions, and outcomes, and a competency in agentic AI specifically means they have reviewed how we handle the hard parts — governance, guardrails, human oversight, and what the system does when it is uncertain.

And this year, Anthropic invited us into its Claude Partner Program — a program they had just launched. We did not apply. Anthropic builds Claude, which is among the model families we deploy on Amazon Bedrock, so this is the company that makes the model saying we know how to build on it. Note what the AWS competency and the Anthropic invitation have in common, because it is the only reason either is worth citing: we could not have bought them.

We are also a ServiceNow Technology Partner, with an application approved for the ServiceNow Store.

We cite these for one narrow reason: none of them is a claim we made about ourselves.

Fourth, what we built with it

Knowledge that stays in people’s heads is a staffing model, not an asset. So we have been turning ours into products.

Datum CMDB is the clearest example. The configuration management database is the foundation every ServiceNow capability depends on — incident routing, discovery, license optimization, Now Assist, agentic workflows — and in most enterprises it is quietly broken while the health dashboard reports green. That gap is the single most common blocker we find sitting underneath a stalled AI initiative.

What that looks like in practice

In one enterprise assessment, the customer’s dashboard reported 87% CMDB health. Actual operational health was 34 out of 100 — because staleness thresholds had been relaxed to 360 days and the 100% compliance score was an unconfigured default rather than a measurement. Of 577,000 configuration items, 462,000 had no assigned owner. Eleven integrations were writing past the platform’s own reconciliation controls, generating roughly 28,000 duplicates.

Nobody was hiding this. The dashboard was measuring configuration, not reality — and no one had a way to tell the difference.

Datum CMDB is an agentic application on Amazon Bedrock that connects read-only to a ServiceNow instance and returns a complete, evidence-based health assessment in about thirty minutes. Zero writes to your CMDB. Fewer than ten thousand API calls. The data can be destroyed when we are done. It produces a health score by dimension, findings with confidence scores and the specific records behind them, a map of which licensed modules are currently blocked, quantified cost of inaction, and a phased remediation roadmap. The diagnostic tier is complimentary. Results vary by environment and are validated during the engagement rather than promised in advance.

Two other lines of research are further back but pointed at problems we keep meeting. The first is instrumentation for human–agent collaboration: as agents take on real work, the handoffs between people and systems become the least documented and least secure part of the operation, and we are building ways to document, measure, and secure that journey end to end. The second is software with no front end — interfaces generated dynamically at the speed and scale that would make the approach viable rather than merely clever. Both are research. We will tell you when they are products.

Fifth — the part that makes the rest testable

Everything above is an argument. Here is what turns it into a test.

AWS has been running funding programs designed to accelerate enterprise AI adoption, and they route them through a short list of partners. We are on it. In practice that means several distinct things, and the distinction matters more than most people realize:

● Funded discovery. AWS co-funds the initial AI assessment — the architecture, roadmap, and business case. The first real step costs you close to nothing.

● Funded labor. This is the one that gets misheard as “credits.” AWS pays us directly to build AI systems on your behalf. Not a discount on your bill — engineering, delivered.

● Migration funding. A separate program with its own budget, for moving workloads to AWS, which can run concurrently with the AI work.

● Infrastructure credits and outcome-based funding. Credits reduce your actual AWS bill; outcome-based funding pays us after measurable results are documented post-launch.

All of it is subject to qualification and approval, and we will not pretend otherwise. But the shape of the offer is unusual enough to state plainly: if you are not yet convinced we know what we are talking about, we will have AWS underwrite the work that demonstrates it. You get senior practitioners, a fixed scope at a fixed fee, and deliverables your team can use without us — at a fraction of what the same work costs unfunded.

We are aware of how that sounds. It is why we put it last. A funding program is not a substitute for judgment, and a firm whose main asset was access to someone else’s money would deserve the skepticism. The funding is the icing. The twenty-two years, the successes, the failures we will describe to you by name, and the products we built out of them are the cake.

So: what to do with this

Ask us for the assessment. Thirty minutes to establish whether you qualify, then a scoped engagement that AWS co-funds. If you have ServiceNow, start with the complimentary CMDB diagnostic — it is read-only, it takes about half an hour, and it will tell you something true about your environment whether or not you ever hire us.

Then hire us.