But this time it is different, really! We hear it at every keynote, read it in every vendor deck, and see it in most board updates on AI right now. The models are smarter, the agents can act on their own, and the money behind them is larger than anything the industry has seen. All of that is true, and I've heard the same line about every wave that came before.
Every few years a new wave of technology arrives with the same promise: this time, the business will finally get what it asked for. Mainframes, enterprise software, the web, cloud, mobile, and now AI and agents. I've spent several decades on both sides of that promise, selling it and living with what came after.
Here's what I keep noticing. The complaints I hear from business leaders about AI today are the same complaints I heard about every wave before it. The pain is old. What's new is the speed and the scale at which it shows up.
So here are ten places businesses are hurting right now. Every one of them has been around since computers were invented.
1. “That's a data issue.”
Data quality has always been a problem. For as long as I can remember, software engineers have looked at a bad result, called it a data issue, and moved on. The business, for its part, has always said some version of “Don't believe what I say. Believe what I mean.” The data gets created somewhere between what was said and what was meant, and nobody owns that gap.
AI inherits the gap and runs with it. A model grounded in data nobody trusts produces answers nobody trusts, only faster and in more places at once. The problem keeps getting pointed at a different set of people, the outcomes keep failing to arrive, and the next tool picks up exactly where the last one left off.
2. Faster in the wrong direction
When technology doesn't reflect how the business works, it locks the mismatch in. Systems built on an outdated picture of the business have always made the wrong things happen more efficiently. The org chart changes, the pricing model changes, the way customers buy changes, and the systems keep running the old rules as if nothing happened.
Agents raise the stakes. An agent will carry out that outdated picture tirelessly, at machine speed, across every transaction it touches. A person who hits a rule that no longer makes sense will usually stop and ask. An agent keeps going. Speed amplifies whatever it's pointed at, and too often it's pointed at a version of the business that changed years ago.
3. “I need a human to talk to.”
Connecting the dots has always been hard. Then came interactive voice response, the phone menus that ask you to press 1 for billing. Contact center automation followed, then AI, then agents. Each layer was meant to get the customer to an answer faster. Each layer also put more distance between the customer and a person who could see the whole picture.
Now agents are talking to agents. One system asks another a question, gets a partial answer, and passes it along. Everyone in the chain holds a piece. Nobody holds the picture, and the customer still ends up asking for a human, only now there are more layers to get through before they find one.
4. Technology creating work for technology
The promise was always the same: automate the routine and free people up for the work that matters. What many people feel instead is that their work has grown. Look closely and a lot of that new work exists only to serve technology. Reconciling outputs between tools. Checking what the AI produced. Fixing what one system did to another. None of it shows up in the business case that justified the tool.
That work shows up in unexpected places and at a scale we haven't seen before. It also moves around. Tasks that used to sit with the people who understood them now land on people who don't, while the people who understood them are busy feeding the machine.
5. Transformation fatigue
People are tired of transformation, which keeps arriving faster and costing more. In 2025, MIT's Project NANDA, a research initiative studying how companies adopt AI, reported that U.S. companies had put $35–40 billion into generative AI and that 95% of the organizations it studied had seen no measurable return.
Medicine lived through this a decade earlier. When hospitals adopted electronic health records, doctors suffered. A 2019 report from Harvard's School of Public Health and the Massachusetts Medical Society declared physician burnout a public health crisis and named those systems a major cause, with doctors spending two hours on screens for every hour with a patient. So who gained? The report notes the systems were built to capture the measures insurers and regulators demanded. Doctors did the typing and carried the cost.
Through every round, people worry about their jobs and their stability, even as many suspect the job they have now is the most exciting one they'll ever get. Still, nobody wants to wake up every morning strapped into a roller coaster.
6. Every new tool needs its own plumbing
Each new AI product has to connect to the systems the business already runs: the customer database, the finance and operations system, the data warehouse, the ticketing tool. When every tool connects to every system on its own, the connection work grows much faster than the number of tools. Ten tools wired individually to ten systems can mean a hundred connections to build and maintain.
This has been true since the first two computers were wired together. AI makes it worse because companies are buying and building AI tools faster than they ever bought software, and each one arrives expecting its own pipes. Each connection is one more thing to secure, monitor and repair whenever a system on either end changes.
7. “Did it work?”
Leaders approve AI budgets and then struggle to trace the spend to a business result. Measuring the value of technology has always been weak. Plenty of companies still can't say with confidence what their last major system upgrade returned, years after the project closed and the team moved on.
Agents make the answer harder to find. Their work is spread across many systems and many small steps, so when revenue rises or a cost falls, separating what the agent contributed from everything else turns into guesswork. The vendor has a number, the finance team has another, and the business owner has a feeling. Without a clear answer, the next budget conversation runs on faith.
8. “I don't trust that number.”
Business users have always second-guessed reports they couldn't trace back to a source. The classic response is to quietly rebuild the number in a spreadsheet before the meeting. Every company has a few people whose real job is reconciling the official number with the one leadership actually believes. AI hands them a lot more to reconcile.
An agent that answers with confidence and shows no reasoning deepens that distrust. When people can't see where an answer came from, they check it by hand, which brings us right back to number four: technology creating more work for people. The more decisions an agent touches, the more of that checking piles up, and the less time anyone saves.
9. The bill nobody expected
Technology costs have always surprised somebody. Most AI is priced by usage, so costs climb as people use it, often before anyone has set limits. Axios reported in May 2026 on a company that ran up roughly $500 million in AI usage in a single month with no spending controls in place.
As a technology matures, finance starts booking it as infrastructure, a utility the business runs on, and it moves into committed spend. Every new wave upends the CFO's spreadsheet of technology costs, and spend that looked optional turns out to be committed sooner than anyone planned. AI adds another complication: the cost of each query depends not just on the model, but on how data is retrieved, processed, and repeatedly fed into it. The Modern Data Company's AI Cost Calculator helps businesses estimate these hidden costs and understand where better data architecture could reduce spending. By the time the invoice arrives, the usage is already built into how teams work, and pulling it back means taking away something people depend on.
10. What the business knows lives in people's heads
The meaning behind the numbers has always been tribal knowledge. Someone knows that “active customer” means something different in one region. Someone knows which field in the old system nobody fills in correctly. That knowledge rarely gets written down, because the people who hold it are too busy using it. New hires inherit the systems without the explanations.
Restructuring sends it out the door. Every round of reorganization takes a few of those people with it, and the knowledge leaves with them. Agents need exactly that context to do anything useful, and once the person who carried it is gone, the agent has nothing to work from. The Modern Data Company's Context Gap Calculator helps companies assess how much of that knowledge is missing from their data, from business definitions and relationships to the rules behind key metrics. The agent will still produce an answer. It just won't necessarily be the right one.
What ties the ten together
Look across all ten and the same thing shows up. Tools, software and utilities have always been there, and they'll keep getting better every year. What decides whether a business gets value from them is the speed at which it can absorb technology into how it actually runs.
That takes operatorship: a management discipline for deciding how technology gets adopted, who owns the outcome, how the business's meaning gets captured, and how much change the organization can take on at once without being knocked over by it. My hope is that enterprises build that discipline deliberately, through their leadership, so the next wave lands better than the last ones did.
Which of these ten is hurting your business most right now? And which ones did I miss?
Sagar Paul

