If you follow AI in the news, you could come away believing two opposite things at once: that AI is about to take every job, and that it is a useless machine that makes things up. Neither is a good basis for running a business.
We have been a technology company for 36 years. We heard dramatic predictions about the PC and about the Internet, too. The useful response then was the same as now: look at what the evidence actually says, and make practical decisions. Here are six common claims, checked against the sources.
Myth 1: “AI is already wiping out jobs across the economy”
What the evidence says: so far, not at the level of the whole economy. The Budget Lab at Yale compared how the mix of jobs in the US has changed since ChatGPT’s release with earlier technology shifts. In its October 2025 analysis, it found that the broader labor market had not experienced a discernible disruption in the 33 months since ChatGPT launched. The Lab now tracks the question with regular updates, and its latest, on September 15, 2026, using August 2026 survey data, reached the same conclusion: no clear evidence of labor market disruption associated with AI. Churn across occupations, AI exposure among the unemployed and usage data all remained flat, within historical ranges or on pre-AI trends.
The researchers are careful to say this is not a prediction; their analysis looks backward, and past workplace shifts have tended to unfold over decades rather than months. That is a fair, balanced place to stand. AI will change how many jobs are done. The evidence doesn’t support panic.
Myth 2: “AI is always right” (or its twin, “AI is always wrong”)
What the evidence says: AI assistants are often right and sometimes confidently wrong. The industry calls the wrong answers “hallucinations”. In a September 2025 research paper, OpenAI explained one reason they happen: the way models are trained and tested tends to reward a confident guess over admitting uncertainty, much like a student guessing on a multiple-choice exam.
The practical takeaway is simple. Use AI for drafts, summaries and research, and have a person who knows the subject check anything that matters. That is how we use it at FTI: AI is far from perfect, and it takes experienced people to direct the work and review the results.
Myth 3: “Everything you type into AI is used to train it”
What the evidence says: it depends on the plan. Free consumer accounts and business accounts often have different terms. As of September 2026, OpenAI states on its business data privacy page that it does not train its models on business customers’ data by default. Microsoft states in its Microsoft 365 Copilot privacy documentation that prompts, responses and the organizational data Copilot accesses are not used to train its foundation models.
So the fear isn’t baseless; it is a reason to choose the right account. Use business plans for company work, read the data terms, and never paste passwords or confidential customer or financial information into a tool your business hasn’t approved. We cover plan choices in AI search in 2026.
Myth 4: “AI is going to use all the electricity”
What the evidence says: AI’s energy use is real and growing, but the headline versions often lose the scale. According to the International Energy Agency’s Energy and AI report, published in 2025, data centers used around 1.5% of the world’s electricity in 2024. The IEA’s base case projects that roughly doubling to around 945 terawatt-hours by 2030, just under 3% of global electricity use.
That is a serious planning issue for utilities and governments, and it deserves attention. It is also a long way from “all the electricity”. Both things can be true.
Myth 5: “AI is only for big companies with big budgets”
What the evidence says: the everyday tools are available to anyone. ChatGPT, Microsoft Copilot and Claude all offer business plans, and Copilot works inside the Microsoft 365 apps many small offices already use. You don’t need a data scientist to draft a letter, summarize a report or clean up meeting notes.
What small businesses do need is a little structure: a business plan rather than personal accounts, a one-page policy on what can and can’t be pasted in, and someone checking the output. That is well within reach of a five-person office.
Myth 6: “AI can run things on its own now”
What the evidence says: AI can take on more of the routine work, but it still needs people in charge. Given Myth 2, that should be no surprise. A tool that sometimes guesses confidently is a great assistant and a poor unsupervised manager.
That is how we use it in our managed IT service. An AI layer watches monitoring, security logs, tickets and documentation around the clock and puts the full picture in front of our engineers. The AI assists; engineers decide and act.
How to read the next AI headline
- Look for the source. Is the claim from research, a vendor’s own terms, or an opinion piece?
- Check the date. AI changes quickly, and last year’s limitation may be fixed, or last year’s promise may not have arrived.
- Ask “compared to what?” A big number without context tells you little.
- Bring it back to your office. Does this change what your team should do on Monday? Usually, the answer is a small practical step, not a revolution.
How FTI helps
We use AI every day in how we design, build and support technology, and we help clients adopt it calmly: choosing business plans, setting sensible rules, and fitting AI into the Microsoft 365 tools they already use.
