Cutting Through the AI Hype

This piece was written by Chip Albright, an Advisor Team Leader here at Opkalla. Tales from the Field is a recurring series on what the IT advisory landscape actually looks like from the field.

--

Not long ago, I sat in on two vendor demos within a single week, both pitching an AI fix for the exact problem one IT director was trying to solve. The reactions couldn't have been further apart.

The first vendor opened with agentic, autonomous, and proactive intelligence inside the first two minutes. Big claims, bigger fonts, a keynote-ready deck. By minute ten, the IT director's arms were crossed and their eyes were on their phone.

The second vendor barely used a buzzword the whole hour. They demoed something that sounds almost boring on paper: a tool that segments a company's environment so IT can see who's using what, including the AI tools employees quietly signed up for on their own. About eight minutes in, the IT director started nodding in agreement and didn't stop.

The room, the week, and the problem never changed. One pitch landed. The other didn't. I've been chewing on why ever since.

Every vendor pitch this year seems to come with one word attached to it somewhere: agentic. Before that it was proactive versus reactive. Before that it was AI voice. Scroll back further and you'll find a familiar pattern with different paint. Synergy. The cloud. Big data. Blockchain. Digital transformation. Old shape, new coat.

I don’t mean to knock on AI itself. Rather, it’s a knock on how it gets sold. And if you run IT, you're the one fielding every one of those pitches, and the one who has to support whatever gets bought.

At Opkalla, we sit across the table from a lot of vendors and a lot of IT leaders trying to figure out which of those pitches are real and which ones are marketing outrunning the product. Here are the four filters we run every pitch through, the same ones you can take into your next vendor meeting.

Filter 1: Can the Rep Explain It Better Than the Marketing Does?

Ask the rep to explain, in plain terms, what the product does for your team. If they can't say it as clearly as the ad in your inbox did, that's your first red flag. When the marketing shows up faster than a rep who can explain the product, that's usually a sign there's more hype budget than engineering budget behind it. If a company can describe what they do better in a LinkedIn post than in a live demo, that's worth noticing.

Right now almost everyone claims to be agentic. Not everyone is. Watch for the gap between the term and what the product actually does when you put it in front of one of your team's real workflows. The first vendor in that demo I mentioned never closed that gap. Ninety minutes of terminology, and I still couldn't tell you what the product did differently than the tools that IT team already owned.

Filter 2: Does the Team That Would Run It Lean In?

Vendor pitches are performances. Your team's reactions aren't.

When you're in the room and the people who would actually run it (your engineers, your service desk lead) start nodding, you have a signal worth some merit. Not the vendor's claims, but your team's reaction to a specific use case. What problem did it solve? What couldn't the old technology do? What changed? That's the real evaluation, and it happens on your side of the table, not the vendor's.

That second demo is the one I keep coming back to. Nobody in that room got excited about a slide. They got excited because someone showed them, in real terms, where their own blind spot was.

Filter 3: Is the Pricing Built for a Company Your Size?

Some of these AI tools are genuinely impressive. They solve complex business problems and they demo beautifully. Then you find out the pricing was built for the Fortune 1000, and your 800-person company is priced like an afterthought to some enterprise logo the vendor is chasing.

A great tool with the wrong pricing model for your business isn't a great tool for your business, and you're the one who has to defend that line item at renewal. That distinction gets lost in a lot of AI conversations right now.

If a tool solves your problem but the price doesn't fit, don't walk away from the problem. Use the demo as your spec. Write down exactly what it fixed and the capabilities your team found value in, then take that list to the market. Ask the original vendor whether a mid-market tier, a smaller module, or a shorter initial term exists, because enterprise list pricing is often a starting point, not a final number. Check whether a platform you already pay for, like your Microsoft 365 licensing or your security stack, covers part of the need. Then compare vendors built for companies your size against that same spec. A tool that does 80% of the job at a price you can defend at renewal usually beats the perfect one you can't.

Filter 4: Does It Fix Something That's Already Broken?

Ask what specific problem the tool solves today, in your environment. If the answer is a vision instead of a problem, keep moving. The tools that land usually aren't flashy, and they fix something that's already costing your team time or exposure.  

That second vendor I mentioned had built something that logically segments a company's environment so you can finally see what's running inside it. It's not going to headline a keynote. But shadow AI, employees quietly using their own AI tools without any oversight, is a real and growing problem for IT teams, and it ends up on your desk whether you approved those tools or not. Governance tools that address it are landing with IT leaders precisely because they solve something painful and unglamorous.

That's usually where you find real signal - in the tool that fixes something that's broken, without overpromising to change everything.

This is also part of why we run technical comparisons across multiple vendors before we land on a recommendation. You see the good, better, best options at the end. You don't see the other seven we vetted and ruled out to get there. That legwork is what Opkalla does so thoroughly, and it enables us to determine which of those demos deserved that IT director's time.

So, Should You Be Implementing AI Right Now?

Yes.

Don't let perfection get in the way of progress. AI has already shown real, measurable advances in fields like healthcare and life sciences, where the results speak for themselves. The spokespeople for this industry have not done it any favors. The rhetoric about ending humanity or replacing every job in a year has done more to create fear and confusion than clarity.

Strip that away and the practical question is simple. Are you at least evaluating what AI could do for your environment and the teams you support? If you aren't, someone in the business already is, likely on a credit card you've never seen. At worst, you find out it doesn't fit right now and you've lost a little time. At best, it changes how efficiently your team operates and how well IT serves the rest of the business. Either outcome is worth the exploration.

I think back to those two demos more than I'd like to admit, because the difference between them was never how much AI was in the product. One vendor sold a vocabulary. The other solved a problem the IT director had and showed it in terms their team could see. Which is exactly the difference the filters are built to catch.

Just run it through the filters first.