By Anupam Anand, AI Leader
Walk into any boardroom today, and everyone is throwing around “AI”. But here is what nobody says out loud: most of it is going nowhere. We see expensive demos and pilot projects that die after three months. Leadership gets excited for an afternoon, then reality hits and nothing changes.
I see this every day, and honestly, I am tired of the hype. Look, if AI isn’t saving you money, speeding something up, or helping someone make a better decision. Why are we doing it? I don’t care how fancy the model is.
The Thing Nobody Wants to Fix First
Most companies do this backwards. They pick a model, they pick a vendor, and they start building. Then, surprise—their data is a complete mess.
Enterprise data is never clean. It is trapped in old spreadsheets, buried in emails, or scattered across three different CRM systems that do not talk to each other. Someone is inevitably keeping critical information on a shared drive from 2017. If you feed that chaos into a language model, good luck. Without context, AI is just guessing. And I don’t know about you, but I don’t want my compliance system guessing.
That is why my team and I focus on the boring stuff first: data context and metadata. We figure out what things actually mean in the specific context of your business. This shift,while not glamorous or trendy, is exactly what saves projects from flaming out.
Where It’s Actually Working
When looking at what is real beyond the polished case studies, we can see genuine impact across several industries:
Healthcare: Claims processing used to take forever due to endless manual back-and-forth. We built a solution that automates parts of the process while keeping it entirely explainable. For us, ensuring that auditors can trust the output matters far more than raw speed.
Financial Services: Compliance is a universal nightmare. The tools we deploy help banks onboard customers faster by minimizing false alerts and avoiding flagging innocent transactions. This means significantly less grunt work for human analysts.
Supply Chain: Instead of reacting when something breaks, our systems flag potential disruptions early, before the damage happens. It sounds simple, but it isn’t.
At the end of the day, nobody remembers what model you used. They remember if you saved them three weeks of work.
What’s Next—and Why It’s a Little Scary
I believe generative AI is just the appetizer; the main course is Agentic AI. This involves multiple small AI agents working together, one plans, one checks facts, and another takes action, all operating within the guardrails you set. It functions less like a chatbot and more like a junior employee who never sleeps.
However, autonomy without accountability is dangerous. You cannot just let AI run loose in a bank or a hospital. We need strong governance from day one, not retrofitted later. Most people skip this step because it is boring, but it is the exact boundary between being useful and being reckless.
What Actually Keeps Me Up at Night
It is not the technology that keeps me awake; it is the people. I have heard the fear and resistance—the phrase *”this thing is going to replace me”*—a thousand times.
My approach to overcoming this is simple: bring people in early and let them help build it. When a team co-creates a solution, they stop seeing AI as a threat and it becomes their tool. When people help build something, they trust it. That sounds obvious, but most companies still roll things out from the top down and wonder why nobody uses it.
My Direct Advice to Leaders
If you are a CXO trying to figure out your next move, here is my unfiltered advice:
1. Start with a real problem: Do not start with a cool model. If you cannot explain the problem you are trying to solve in two sentences, go back to the drawing board.
2. Build governance on day one: Do not wait until month six. Establish your guardrails immediately.
3. Be honest about what is working: Some use cases are never going to scale. Kill them and move on.
4. Get ready for AI agents, not just copilots: The way your teams work is going to change faster than you think.
The next 12 to 18 months are going to move fast. The winners won’t be the ones using the fanciest AI; they will be the ones who know exactly where AI belongs—and where it doesn’t.