The most valuable advice an AI consultant ever gave a client was not about AI.
It was six words: “Don’t use AI for that.”
A retail chain operating across South India came to us with a clear, well-defined ask. They wanted to automate customer complaint resolution using an AI chatbot. Reduce the load on the call center. Cut costs. Improve response times. Tick the box on digital transformation. Simple.
They already knew what they wanted built. They’d seen demos. They had a budget. They were ready to sign.
We asked for two weeks before writing a single line of code. They agreed, somewhat impatiently.
In those two weeks, we did something that most technology vendors never do: we mapped their complaint patterns. Not the categories listed in the ticketing system, but the actual patterns underneath. What customers were really complaining about. How frequently. From which geographies. Linked to which touchpoints.
What we found stopped the project in its tracks.
Sixty eight percent of all incoming complaints traced back to exactly three root causes. First: delivery time estimates shown in the app were systematically inaccurate, creating expectations the operations team couldn’t meet. Second: a billing glitch affecting one specific payment gateway was generating duplicate charges for a subset of customers who didn’t even know they were being double-billed. Third: a packaging issue at the Coimbatore warehouse was causing product damage in transit at a rate that was invisible in aggregate but catastrophic to the customers experiencing it.
None of those problems needed a chatbot. They needed fixes.
The AI solution they’d come to us for would have been genuinely impressive. Faster responses. Friendlier tone. Available at 2 AM. Multilingual. It would have scored beautifully on satisfaction surveys in the first month. And underneath all of that, the disease would have continued getting worse. Because an AI that handles complaints better doesn’t eliminate the conditions producing them.
Customers would have gotten smoother apologies for problems that should never have existed. And because those apologies were coming faster and more gracefully, the signal in the complaint data would have gotten quieter, making it even harder to spot the root causes that needed attention.
This is the thing that good AI consulting forces into the open: the difference between automating a process and fixing it. Between wrapping a problem in technology and actually solving it. It’s a distinction that nobody selling software has an incentive to make, because making it sometimes means recommending a smaller project or no project at all.
Real strategic consulting starts with the diagnosis, not the prescription. It asks whether this is actually an AI problem or a process problem wearing AI’s clothes. Whether this workflow is worth automating or worth eliminating. Whether deploying AI here creates genuine leverage in the business or simply creates a more expensive, harder-to-debug version of the broken thing you already have.
These are uncomfortable questions. Clients don’t always want to hear them. But they are the questions that separate technology implementations that actually generate returns from the ones that generate case studies and then quietly get switched off.
For that retail chain, we ended up building something considerably smaller than what they initially asked for, and considerably more valuable.
An AI system that monitored incoming complaint patterns in real time, identified spikes, and automatically traced them to operational root causes. When complaints about delivery accuracy crossed a threshold in a specific geography, the system didn’t route them to a chatbot. It flagged the logistics team and surfaced the delivery estimate discrepancy in a structured diagnostic report. When billing complaints clustered around a specific payment gateway, it triggered an automatic audit and escalation to the payments team before the issue appeared on social media.
Leadership got 48-hour early warning on systemic issues before they became PR disasters. The operations team stopped playing whack-a-mole with symptoms and started seeing causes. And the chatbot never got built, because by the time the system was running, the complaint volume had dropped enough that it wasn’t needed.
That’s what good AI consulting actually looks like. Not the biggest system. Not the most impressive demo. The right solution for the real problem, even when the real problem turns out to be smaller and stranger than anyone expected.
At Evvo Technology, we ask the hard questions before building anything, because we’ve seen too many organizations invest heavily in AI that solved the wrong problem brilliantly. The right solution starts with the right diagnosis. That’s where we start. And we’d like to start there with you.
And if you’re wondering what happens when AI doesn’t just execute workflows but actively improves the quality of business decisions, read our blog: “The Meeting That AI Attended Better Than Anyone Else” It explores how AI can surface critical insights before the meeting even begins, helping organizations make faster, smarter, and more confident decisions.
