I lost count of how many emails last month started with “as your trusted AI partner.” Most of those senders had never seen our clients’ environments, which is a funny definition of partnership if you think about it for more than five seconds. Partnership, it turns out, can be a font choice.
That noise is why independence matters more than it did when we were arguing about servers in a closet. Not because AI is magic, but because the sales motion got faster and the vocabulary got fuzzier at the same time, and fuzzy vocabulary is where bad decisions hide.
Incentives don’t disappear because the product is smart
If a firm makes margin on seats, consumption, or implementation hours tied to one stack, their “best practice” will tilt. Sometimes the tilt is subtle—a larger package, a longer term, a module you’ll never configure—and sometimes it’s a full-court press with countdown clocks. We’re not pure, because nobody is, but we are not paid to move a particular cloud, license, or agent platform. When we recommend something, the question is supposed to be whether it fits this business’s risk, skills, data, and budget, or whether we should say no. That last option is underrated. “No” is a deliverable.
Wrong choices compound faster now
Legacy IT mistakes were expensive. AI mistakes can be expensive and fast. An agent with broad permissions doesn’t wait for a change window to make a weird decision, and a multi-year platform deal freezes strategy while the market keeps moving. I don’t say this to scare people off AI. We’ve helped clients use tools carefully and get real value. I say it because the cost of a bad technology decision used to take months to entrench, and now it can entrench in a sprint with a success story nobody wants to re-open.
Same words, different products
Transformation, autonomy, intelligence, governance—everyone uses them, and differentiation collapses into relationships and FOMO. Independent advisors exist to translate the pitch back into English: what works today, what’s experimental, what needs supervision, and what simply isn’t ready for your shop. A useful question I keep recycling is “show me the failure mode.” If the room goes quiet, you’re not buying technology so much as confidence, and confidence is a terrible sole requirement for a purchase order.
Independence without competence is just opinion
I’ve met “independent” consultants who couldn’t tell you what a restore test is. Independence is not a personality. It’s a posture plus scars. Our guidance sits next to managed IT for a reason—monitoring, security, compliance support, hosting, continuity, custom web work when off-the-shelf is the wrong fit. AI advice that ignores those fundamentals is incomplete. Tools don’t erase patching, identity, or ownership. They make the gaps louder.
How we use independence without creating chaos
We’re not trying to second-guess every ticket. Independence earns its keep at decision points: shortlist, pilot design, contract, renewal. Between those points, operations should run. That’s the complete IT department model—day-to-day reliability plus judgment when money and lock-in show up. Sometimes independence means “not yet,” because data quality or identity hygiene is a mess, and AI on top of disorder multiplies disorder. The advisor who says so may sound less exciting than a vendor. Six months later, when the budget is still intact, that dull sentence looks pretty good.
If you’re choosing between vendors, shortlisting AI tools, or staring at a renewal that feels inflated, get a review that isn’t funded by the sale. That’s not luxury. In this market, it’s risk control. I’ll end with something unfashionable: independence can mean recommending a competitor’s tool, or recommending nothing this quarter. If that threatens a relationship, the relationship was a sales funnel. Nearly three decades of watching funnels masquerade as partnerships will make a person picky, and picky is how budgets survive the AI gold rush.