Explore the challenges of AI adoption in organizations and learn how to unlock its true value through effective architecture and governance strategies.
... Returning ROI, that's the hard part -
Halfway through a recent conversation on The Chemical Show, host Victoria Meyer paused to take notes. We had drifted, without planning to, into the part of the AI story that most executives I meet are quietly stuck on. Not whether to adopt it. They already have. What to do now that they have.
Almost every chemicals and materials company I talk to brought AI on board this past year. Pilots, working groups, a firewalled document archive that everyone was told to go use. That was the right instinct. When a capability arrives this fast, getting your organization's hands on it is a reasonable first move, and the companies that did it were reading the moment correctly.
But adoption and return aren't the same thing, and the gap between them is where this year's real work sits.
Why the Chat Prompt Is a Ceiling
The pattern is familiar. A company opens a chat tool to the whole organization and waits for the good ideas to surface from the ground up. Some do. People make individual tasks faster. What almost never emerges from that is anything at scale, because the chat prompt is a ceiling, not a staircase.
As long as your people are working one prompt at a time against whatever they can paste into a window, the size of the problem they can solve stays capped by that window. The interesting opportunities live somewhere else.
The progression that opens them up runs in three steps. Chat is the first: one person, one question. The second is pointing AI at a folder, twenty or fifty or a hundred documents, and asking it to work across all of them at once. The third is connecting AI to the systems where your data actually lives, the CRM, the ERP, the contract repository, so it can work across all of them together. That last step is where the scale is. It is also where most organizations have not gone.
The Architecture Underneath
Getting there is not a matter of buying a better tool. The AI that returns real value does not sit inside any one system. It sits next to your organization and does the cross-system work your people already do by hand: pull this from here, that from there, reshape it for the situation in front of them. Recreating that well takes an architecture, and in the conversation three parts of it came into focus.
Governance comes first, because connecting systems means giving a new kind of user access to data that used to sit in separate rooms. That has to be designed deliberately, not opened up and hoped over.
Connected systems come second. Value shows up when AI can reach across your data, not when a single vendor's tool has a chatbot bolted onto it. Those embedded assistants don't live in the workflow where the cross-system work actually happens.
Contextual layers come third, and they're the part almost nobody is talking about. At scale you're not prompting AI. You're giving it a well-built body of your organization's context and a set of operating instructions. Built properly, that layer carries your best practice inside it, stays under your control, and is worth protecting. It's closer to a competitive advantage than to a piece of software, and the firms capturing the most value have started to treat it that way.
Actioning the Insight
Most of what gets called an AI risk, hallucination especially, is an artifact of the chat prompt: too little context, so the model fills the gap and fills it wrong. A well-built system with its context supplied and protected doesn't behave that way. Getting out of chat is a quality decision as much as a scale one.
If you want to know whether an AI initiative is real, ask three questions. What is the methodology behind it. How was the architecture built. How was it validated. An initiative that can't answer them is an experiment, however it's dressed up, and experiments are the first line a budget review cuts.
I worked through all of this with Victoria Meyer on The Chemical Show. If you're trying to find where AI actually earns its return in a materials business, that's the conversation we have with clients every week.
Until next week,
Kendall -
Find me on LinkedIn or Book a 1:1 call
Not a subscriber yet? Subscribe here
