TL/DR: Treat generative AI like the late-90s internet bubble: useful and here to stay, yet massively overfunded right now, with high burn rates, unresolved legal risk, uneven accuracy, and real backlash on the horizon. Use it sparingly as a brainstorming accelerator, not as the backbone of your business, not as a replacement for humans.

The AI Bubble Logic: growth first, profits later

The current AI race is a classic saturation play. Leaders are scaling usage while tolerating extraordinary losses and cash burn to “win the platform” before worrying about profits. Reporting shows OpenAI expects to remain cash-flow negative until 2029 and is burning billions annually to fund chips, data centers, and model training; that mirrors dot-com era logic where market share trumped unit economics, and bubbles eventually reset valuations even if the underlying tech endured.

Central banks and economists are openly debating “AI bubble” risk, drawing explicit parallels to dot-com and telecom booms. The Richmond Fed shows AI’s investment arc identical to the 1990s infrastructure surge, while global forums and business media track the same pattern of massive capital pouring in ahead of proven returns, which can work until sentiment turns and when it turns it turns fast.

What this means for you: avoid deep vendor lock-in on core operations. If your workflow depends on models that get repriced after the land-grab, you inherit the bill later and you have no leverage when costs change.

AI Accuracy Reality: “predictive text on steroids”

Generative AI does not “know” things or is even intelligent. It predicts words simplified as tokens using patterns in data and context windows. U.S. guidance like NIST’s AI Risk Management Framework stresses reliability, testing, and risk controls for safety-critical uses which is a polite way of saying do not use general-purpose generative models where you need high, auditable accuracy like finance, medical, or regulated form completion. The FTC has also stepped in against deceptive AI accuracy claims… exposing your company to legal liability.

If your threshold is “we must be right more than nine times out of ten,” then humans and traditional software with deterministic checks are the way to go… and yes that’s slower sometimes but it is WAY cheaper than remediating errors at scale later.

AI Legal and IP Risks are Real

Copyright fights are not theoretical. Major publishers, media companies, and creators are suing model makers over training data and output use and judges are allowing significant portions of these cases to proceed which creates uncertainty about future rules and licensing costs. There is also an active test of whether purely AI-generated works are even copyrightable in the United States. Until the dust settles, heavy reliance on AI for creative output exposes you to takedowns, rework, or even damages.

Practical takeaway: keep final creative assets human-made. Use AI for ideation, mood boards, or rough comps, then have a human designer execute the production work so your chain of title is clean.

The AI Reputational Risk: customers notice and some boycott

Consumer sentiment toward AI-made ads and imagery is mixed at best. Surveys show a large share of U.S. adults distrust brands that use AI in advertising and recent campaigns using synthetic creatives have drawn public backlash and mockery these are not edge cases… they are warning shots. If your brand leans on AI art or auto-generated social, expect segments of your market to actively avoid you. Selling merch with AI? Get ready for backlash including being kicked out of marketplaces and dealing with the police.

That is why our advise is simple and a little old-school: keep your brand voice human, keep your visuals human, and keep your community management human because trust is slow to earn and fast to lose.

A Balanced AI Adoption Plan for Real Businesses

Use AI where it speeds up thinking without owning critical outcomes. Brainstorming, competitive scans, outline drafts, research synthesis, code scaffolding, and internal notes are fair game. Keep humans accountable for decisions, numbers, signatures, and creative deliverables that ship to customers. Build your stack so you can swap vendors or revert to non-AI workflows if pricing spikes or terms change, because they will change.

Bottom line for Ancilia clients: AI is valuable as an accelerant, not as an engine. We integrate it cautiously to reduce costs today without creating expensive liabilities tomorrow which is how you benefit from the technology without becoming dependent on the hype cycle that will eventually break.

If you want specific help with AI and your business, contact us!