Clean Up the Brownfields Before You Deploy the AI

The world of technology is like a carnival merry-go-round. The lights are flashy, the music is loud, and people jump on and off looking for a quick escape. But it’s a dangerous ride if the machinery isn't maintained, the operator is asleep, or no one is supervising the chaos. Having spent years in this traveling technology road show, I’ve seen my share of accidents—some caused by pure carelessness, others by decisions made on completely different continents. On this blog, I want to map out both the promise and the peril of Artificial Intelligence. Because right now, the flashy ride is slowing down, and we are entering a gritty new phase in how this technology is actually deployed.

Artificial Intelligence is an incredibly powerful tool, but it is hitting a wall. It is massively expensive, drains power grids, and requires vast amounts of cooling water—leading to severe community pushback against new data center construction. Meanwhile, inside the corporate office, a different bottleneck is forming. This week, Celonis pointed out a massive disconnect in how AI is being deployed. Their data shows that 83% of business leaders already use AI tools in their daily routines, and 85% aim to transform into "agentic enterprises" within three years. AI is quickly becoming a utility as standard as spreadsheets or video calls. Unfortunately, most organizations are miles away from that seamless, agentic reality. The culprit isn't worker resistance; it’s poor organizational design and fractured legacy processes. The merry-go-round is broken.

In its report last year, MIT found that 95% of Artificial Intelligence projects fail to deliver a return on investment. As someone who works on the inside of the business, I have a pretty good idea why this is the case. The high cost of this technology makes it hard to scale systems across complex businesses. It is even harder when you have office politics, human judgment, and systems that do not collaborate, acting as the scaffolding that keeps the business operating.

Take the accounts receivable department at your firm. Checks and cash enter your organization in various ways. First, you send out invoices, and customers mail checks back. You could also set up ACH transactions to automatically charge the customer's account. Alternatively, you can create EDI systems to bill and credit customers in real time. What appears to be a straightforward process quickly becomes a tactical nightmare as the accounting system, inventory, and bill processing reside on three different computer systems, purchased over different decades. There is also a reconciliation spreadsheet on someone's desktop, with complex macros written by a consultant who left the business fifteen years ago, that runs as a nightly batch process. This is why you need six people to process accounts receivable: the sheer level of human judgment and manual interaction required. If it is that difficult for six experienced humans, imagine the struggle of trying to build an Artificial Intelligence agent to automate it.

Inefficiency like this did not happen all at once. It happened one small decision at a time, creating a situation where accounts receivable becomes its own dangerous merry-go-round requiring six people just to keep it from spinning off the tracks. Streamlining those systems is a massive undertaking that will require significant time and money. It is precisely why the agentic enterprise remains so elusive. Your organization has significant technical and structural debt, and historically, it has been cheaper to ignore it rather than address it.

Technology professionals are finally turning their attention to these situations, adopting a term from urban planning: "brownfield" development. In the physical world, a brownfield site is an abandoned commercial or industrial property where reuse is complicated by real or perceived environmental contamination. Because the land is contaminated and the cost of rehabilitation is prohibitive, these sites are often abandoned for decades. With the rise of Artificial Intelligence, corporations are finally being forced to address the metaphorical brownfield problems buried deep within their own organizations.

Brownfields are the new frontier of IT and Artificial Intelligence. Agile professionals and project managers are beginning the heavy lift of repairing these dangerous merry-go-rounds so that executive dreams of agentic enterprises can actually come true. It is expensive, time-consuming, and it requires frank, uncomfortable conversations about organizational structure. It also requires people willing to stick around for the long haul to take these efforts to their conclusion—because when someone in accounts receivable retires or quits, the checks start bouncing.

The brownfields in your business explain why 95% of your projects lack a clear return on investment. It is why Celonis notes that only a fraction of organizations are truly positioned to upgrade their operations for AI successfully. It is also why less experienced employees are struggling so much in today's job market: organizations don't need theoretical builders right now. They demand seasoned, experienced professionals who know how to clean up toxic brownfields.

If we are lucky, we can fix our organizations and implement Artificial Intelligence without the corporate merry-go-round hurting someone. Otherwise, we are all in store for a very dangerous ride.

Until next time.