ONLINEAGENT_OPS 2026.Q3 HOME HUBSMODELSBUILDINGCOMPANIESFILMMAKINGSECURITYLIVING WITH ITSKILLSALL HUBS, ONE TERMINAL → ARTICLES CRAFT RECORD BLOG TOOLS SKILLS SEARCH /
HOME›ARTICLES›PROS AND CONS OF AI: A BALANCED, SOURCED LIST
ARTICLES · EXPLAINERBEGINNER

Pros and Cons of AI: A Balanced, Sourced List

The pros and cons of AI in plain English, side by side: real gains in work, science and health, and real costs in errors, bias, scams, privacy and energy.

READ10 min
WORDS2,063
SECTIONS6
TYPEEXPLAINER
CHECKED5 OCT 26

The main pros of AI, so far, are speed on routine thinking work, the most help for the least experienced, and real advances in science and medicine. The main cons are confident mistakes, unfair results for some groups, cheaper scams, privacy and energy costs, and disruption for some workers, especially people starting out. Both lists have evidence behind them. Which side wins depends on the task, and on whether a person checks the work.

◈ IN PLAIN TERMS

AI is like a very fast, very confident assistant who has read everything and checked nothing. Hand it a first draft or a pile of data and you save hours. Hand it a decision about someone’s health or money and walk away, and its confidence becomes the problem. The pros and cons below are mostly two sides of that one fact.

◈ THE QUICK LIST: 16 PROS AND CONS, ONE LINE EACH
  1. Pro: Speeds up routine work drafting, summarising and rewriting go faster; support agents in one study resolved more issues an hour.
  2. Pro: Helps beginners most the least experienced workers gained most, with little change for experienced staff.
  3. Pro: Speeds up science AlphaFold predicts protein structures, and a research group has used it to look for treatments.
  4. Pro: Helps doctors catch more AI-supported breast screening found more cancers, with radiologists still reading the scans.
  5. Pro: Broadly capable, widely used passes professional exams and writes working software, yet struggles with some simple-seeming tasks.
  6. Pro: Finds patterns in data older, predictive AI quietly forecasts outcomes from data every day.
  7. Pro: Can run on your own device keeps your data local, at the cost of a smaller, less capable model.
  8. Con: Confident mistakes a wrong answer looks exactly like a right one.
  9. Con: Not always faster experienced developers in one trial were slower with AI, though they felt faster.
  10. Con: Can be unfair face analysis erred most on darker-skinned women, and detectors wrongly flagged non-native writers.
  11. Con: Cheaper scams fake voices and video calls are part of the AI-related fraud reported to the FBI.
  12. Con: Your words go to a company many chatbots keep what you type and, on consumer plans, may train on it when a setting is on.
  13. Con: Lost work, especially at the start employers cite AI in job cuts, and young workers in exposed jobs struggle to find work.
  14. Con: Electricity use one answer uses little, but billions add up, and data-centre use is projected to grow.
  15. Con: Flattery and over-reliance people trust a flattering AI more, and more confidence in AI went with less critical thinking.
  16. Con: Power in a few hands the UK competition regulator warns that a few big tech firms already hold market power.

Pros and cons of AI, side by side

The plain point: for almost every gain there is a matching cost, and each row links to the evidence below.

AreaProCon
WorkPro: faster routine work; in one large study, newer staff gained most.Con: not always faster for experts; fewer first jobs in exposed fields.
KnowledgePro: broad, fluent help on almost any topic.Con: confident mistakes, and a pull towards agreeing with you.
Science and healthPro: protein structure prediction; more cancers caught in screening.Con: errors in health settings carry high stakes.
FairnessPro: spots patterns in data, such as forecasts, quickly.Con: can work worse for groups missing from its training data.
Safety and trustPro: can run on your own device, keeping data local.Con: cheap fake voices and faces; chats sent to companies.
Society and planetPro: most AI researchers surveyed expect more good than bad.Con: rising electricity use; power held by a few firms.

The pros, with the evidence

The plain point: the strongest evidence for AI’s benefits comes from controlled studies at work and from science and medicine.

It speeds up routine thinking work

In a study of 5,179 customer support agents, economists found that an AI assistant “increases productivity, as measured by issues resolved per hour, by 14% on average”.1 Drafting, summarising and rewriting are the everyday versions of the same gain.

It helps the least experienced most

The same study found “a 34% improvement for novice and low-skilled workers but with minimal impact on experienced and highly skilled workers.” Its authors suggest the AI spreads “the best practices of more able workers”, and that it also “improves customer sentiment, increases employee retention”.2

It is speeding up science

The 2024 Nobel Prize in Chemistry went partly to Demis Hassabis and John Jumper for AlphaFold, an “artificial intelligence (AI) protein structure prediction model”. The research group DNDi (Drugs for Neglected Diseases initiative) says it has used it to look for treatments for Chagas disease and leishmaniasis.3

It can help doctors catch more

In the MASAI trial in Sweden, with about 106,000 women, AI-supported breast screening found “29 percent more detected cancers”, cut the radiologists’ reading workload “by 44 percent”, and led to “12 per cent fewer interval cancers” (cancers found between screenings).4 Radiologists still read the scans: the AI sorted them and “marked suspicious findings” for the radiologists.5

It is broadly capable, and widely used

The 2026 International AI Safety Report says leading systems “pass professional licensing exams in law and medicine, write functional software when given simple prompts”, and that “At least 700 million people now use leading AI systems weekly.”6 The same report adds that they still “struggle” with some simple-seeming tasks, which is where the cons begin.

It finds patterns in data

Older, less famous AI does quiet work every day. IBM: “Predictive AI blends statistical analysis with machine learning algorithms to find data patterns and forecast future outcomes.”7 Types of AI explains how this differs from the chatbots.

It can run on your own device

Not all AI has to send your data away. IBM says running a model on your own device “theoretically provides the greatest possible user privacy.”8 The trade is a smaller, less capable model. See local AI meaning.

The cons, with the evidence

The plain point: the strongest evidence against careless use comes from errors, bias, fraud and the cost of running it all. Is AI dangerous? sorts these risks by how sure anyone can be.

It makes confident mistakes

NIST, the US government’s standards agency, names the risk: “The production of confidently stated but erroneous or false content”.9 A wrong answer looks exactly like a right one. Why AI makes things up explains why.

It is not always faster

In a 2025 randomised trial with 16 experienced open-source developers, “allowing AI actually increases completion time by 19%”, even though the developers estimated afterwards that it had saved them 20%.10 Gains depend on the person and the job; feeling faster is not the same as being faster.

It can be unfair

A 2018 study of face-analysis systems found error rates “of up to 34.7%” for darker-skinned women against “0.8%” for lighter-skinned men.11 AI-writing detectors wrongly flagged “more than half” of essays by non-native English speakers.12 See AI bias meaning.

It makes scams cheaper

The FBI received “more than 22,000 complaints reporting AI-related information” in 2025.13 Fake voices and video calls are part of it; how AI fraud works and deepfake self-defence show what to watch for.

Your words go to a company

Many chatbots can keep what you type and, on consumer plans, use it for training when a setting is on; Anthropic’s terms, for example, cover “Free, Pro, and Max accounts when this setting is on”.14 What happens to what you type lists the settings.

Some people lose work, especially at the start

US employers have cited AI in “184,538 job cut announcements” since 2023, and the Dallas Fed finds young workers in exposed jobs held back by “a low job finding rate”.15 National figures barely move so far; which jobs are safe from AI? holds both facts.

It uses a lot of electricity in total

The IEA projects data-centre electricity “roughly doubling from 485 TWh in 2025 to 950 TWh in 2030”.16 One answer is small; billions are not. What AI actually runs on gives both scales.

It can flatter you, and you can lean on it too much

People “trusted the sycophantic AI model more” in one study, and a Microsoft Research survey found “higher confidence in GenAI is associated with less critical thinking”.17 See sycophancy meaning.

Power sits with a few firms

The UK’s competition regulator has warned of “a small number of incumbent technology firms which already hold positions of market power”.18 Who builds AI names them.

TAKEAWAY

The pros are strongest where a person stays in charge and checks the result. The cons are worst where AI output is trusted without checking, or where it decides something about a person.

So is AI good or bad overall?

The plain point: nobody can total it up, and the experts are split but leaning hopeful.

In the largest survey of AI researchers, 68.3% “thought good outcomes from superhuman AI are more likely than bad”, yet nearly half of those optimists, 48%, still gave at least a 5% chance of extremely bad outcomes.19 For you, the useful question is smaller: for this task, does the time saved outweigh the checking needed?

How to get the pros without the cons

The plain point: most of the cons shrink with a few habits.

  • Use it for first drafts, not final answers. Then read and fix what it wrote.
  • Give it the source. Paste the document you want summarised instead of asking from memory.
  • Check the specifics. Names, numbers, dates, quotes and links are where mistakes hide.
  • Keep people in charge of decisions about people. Hiring, grades, loans and health need a human who can be asked why.
  • Keep secrets out. No passwords, health details or other people’s information.
  • Verify any request for money on a channel you already trust.

Questions people ask

What are the main advantages of AI?

Speed on routine work, the biggest gains for less experienced people, and advances in science and medicine. In one study, customer support agents resolved 14% more issues an hour, and novices 34% more.2

What are the main disadvantages of AI?

Confident mistakes, bias against some groups, cheaper scams, privacy and energy costs, and job disruption for some. Is AI dangerous? sorts them by how well each is measured.

Does AI make people lazy?

There is early evidence it can reduce effort when people over-trust it: in a Microsoft Research survey, more confidence in AI went with less critical thinking. It is a survey of 319 workers, so it shows a link, not proof.20

Is AI good or bad for society?

Both, depending on how it is used and who checks it. Most AI researchers surveyed lean hopeful, while still taking serious risks seriously.

Will AI replace humans?

So far the measured effect is on some tasks and some jobs, mainly fewer openings for newcomers, not on people wholesale. Which jobs are safe from AI? has the detail.

QUICK CHECK · 3 QUESTIONS

Pick an answer to see if you got it. Nothing is recorded.

1. According to the page, where are the pros of AI strongest?

  • Where a person stays in charge and checks the result
  • Where AI makes decisions about people on its own
  • Where the output is used without a review
Show the answer

Answer: Where a person stays in charge and checks the result. The page says the pros are strongest where a person checks the result, and the cons worst where output is trusted unchecked or decides something about a person.

2. In the page’s study of customer support agents, who gained most from an AI assistant?

  • The most experienced staff
  • Managers rather than the agents
  • The least experienced workers
Show the answer

Answer: The least experienced workers. The page reports the biggest improvement for novice and low-skilled workers, with little change for experienced staff.

3. Which habit does the page suggest for getting the pros without the cons?

  • Ask it from memory rather than pasting the document
  • Give it the source, then check names, numbers, dates and links
  • Let it make hiring and loan decisions to save time
Show the answer

Answer: Give it the source, then check names, numbers, dates and links. The page says to paste the document you want summarised and check the specifics, because names, numbers, dates, quotes and links are where mistakes hide.

SOURCES
  1. 1Brynjolfsson, Li & Raymond, Generative AI at Work, NBER Working Paper 31161, April 2023, read at source 5 Oct 2026“we study the staggered introduction of a generative AI-based conversational assistant using data from 5,179 customer support agents. Access to the tool increases productivity, as measured by issues resolved per hour, by 14% on average”. ↑
  2. 2Brynjolfsson, Li & Raymond, Generative AI at Work, read at source 5 Oct 2026“including a 34% improvement for novice and low-skilled workers but with minimal impact on experienced and highly skilled workers. We provide suggestive evidence that the AI model disseminates the best practices of more able workers” · “AI assistance improves customer sentiment, increases employee retention, and may lead to worker learning.” ↑a“by 14% on average, including a 34% improvement for novice and low-skilled workers”. ↑b
  3. 3DNDi, Nobel prize for chemistry awarded for AlphaFold, 10 Oct 2024, read at source 5 Oct 2026“Demis Hassabis, John Jumper and David Baker for being awarded the 2024 Nobel Prize in Chemistry” · “artificial intelligence (AI) protein structure prediction model” · “DNDi teams and partners have used the game-changing technology to find innovative ways to accelerate drug discovery for Chagas disease and leishmaniasis.” ↑
  4. 4Lund University, AI support in breast cancer screening: fewer missed cancer cases, 30 Jan 2026, read at source 5 Oct 2026“106,000 women participated in the MASAI trial.” · “29 percent more detected cancers, mostly small, lymph node negative invasive cancers.” · “The screen-reading workload for radiologists was reduced by 44 percent.” · “Overall, there were 12 per cent fewer interval cancers with AI-supported screening compared to standard screening (82 compared to 93).” ↑
  5. 5Lund University, AI support in breast cancer screening, read at source 5 Oct 2026“In the MASAI trial, AI was used to triage mammograms to single or double reading based on AI risk scores. AI was also used as detection support for the radiologists where it marked suspicious findings in the mammogram.” · “In Sweden, women between the ages of 40 and 74 are regularly invited to mammography screening.” ↑
  6. 6International AI Safety Report 2026, publication page, 3 Feb 2026, read at source 5 Oct 2026“Leading general-purpose AI systems now pass professional licensing exams in law and medicine, write functional software when given simple prompts, and answer PhD-level science questions as well as subject-matter experts.” · “At least 700 million people now use leading AI systems weekly.” ↑
  7. 7IBM, Generative AI vs. predictive AI, 9 Aug 2024, read at source 5 Oct 2026“Predictive AI blends statistical analysis with machine learning algorithms to find data patterns and forecast future outcomes.” ↑
  8. 8IBM, What is AI inference?, updated 13 Jan 2026, read at source 1 Oct 2026“On-device deployment is simple and secure, and theoretically provides the greatest possible user privacy.” ↑
  9. 9NIST, AI 600-1, Generative AI Profile, July 2024, read at source 1 Oct 2026“The production of confidently stated but erroneous or false content (known colloquially as ‘hallucinations’ or ‘fabrications’)”. ↑
  10. 10Becker et al., Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity, arXiv:2507.09089, July 2025, read at source 5 Oct 2026“16 developers with moderate AI experience complete 246 tasks” · “After completing the study, developers estimate that allowing AI reduced completion time by 20%. Surprisingly, we find that allowing AI actually increases completion time by 19%”. ↑
  11. 11Buolamwini & Gebru, Gender Shades, PMLR 81, 2018, read at source 1 Oct 2026“darker-skinned females are the most misclassified group (with error rates of up to 34.7%). The maximum error rate for lighter-skinned males is 0.8%.” ↑
  12. 12Liang et al., GPT detectors are biased against non-native English writers, Patterns, July 2023, read at source 17 Sep 2026“they incorrectly labeled more than half of the TOEFL essays as ‘AI-generated’”. ↑
  13. 13FBI Internet Crime Complaint Center, 2025 IC3 Annual Report, read at source 17 Sep 2026“In 2025, IC3 received more than 22,000 complaints reporting AI-related information. Adjusted losses of these complaints exceed $893 million.” ↑
  14. 14Anthropic, Updates to Consumer Terms and Privacy Policy, August 2025, read at source 11 Sep 2026training applies to “Free, Pro, and Max accounts when this setting is on”. ↑
  15. 15Challenger, Gray & Christmas, Job Cut Announcement Report, 6 Aug 2026, read at source 23 Sep 2026“it has been cited in 184,538 job cut announcements.” · Federal Reserve Bank of Dallas, Young workers’ employment drops in occupations with high AI exposure, Jan 2026, read at source 9 Sep 2026: “not due to layoffs but to a low job finding rate for young workers entering the labor force”. ↑
  16. 16IEA, Key Questions on Energy and AI, April 2026, read at source 17 Sep 2026“roughly doubling from 485 TWh in 2025 to 950 TWh in 2030, accounting for around 3% of global electricity demand by that date.” ↑
  17. 17Cheng et al., preprint arXiv:2510.01395, submitted 1 Oct 2025, read at source 23 Sep 2026“participants rated sycophantic responses as higher quality, trusted the sycophantic AI model more, and were more willing to use it again”. · Lee et al., Microsoft Research, The Impact of Generative AI on Critical Thinking, CHI 2025, read at source 5 Oct 2026: “higher confidence in GenAI is associated with less critical thinking, while higher self-confidence is associated with more critical thinking.” ↑
  18. 18UK Competition and Markets Authority, CMA outlines growing concerns in markets for AI Foundation Models, 11 Apr 2024, read at source 5 Oct 2026“a small number of incumbent technology firms which already hold positions of market power in many of today’s most important digital markets.” ↑
  19. 19Grace et al., Thousands of AI Authors on the Future of AI, arXiv:2401.02843, revised 8 Oct 2025, read at source 5 Oct 2026“While 68.3% thought good outcomes from superhuman AI are more likely than bad, of these net optimists 48% gave at least a 5% chance of extremely bad outcomes such as human extinction”. ↑
  20. 20Lee et al., Microsoft Research, The Impact of Generative AI on Critical Thinking, read at source 5 Oct 2026“We survey 319 knowledge workers”. ↑

Becker et al., Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity, July 2025, read at source 5 Oct 2026.

Liang et al., GPT detectors are biased against non-native English writers, July 2023, read at source 17 Sep 2026.

Cheng et al., arXiv:2510.01395, 2025, read at source 23 Sep 2026.

Lee et al., Microsoft Research, The Impact of Generative AI on Critical Thinking, April 2025, read at source 5 Oct 2026.

Grace et al., Thousands of AI Authors on the Future of AI, revised 8 Oct 2025, read at source 5 Oct 2026.

Sources read before 5 Oct 2026 were read for this site’s linked detail pages and are re-used here with the date they were read. The comparison on this page (the fast, confident assistant), the side-by-side table’s groupings and the habits in “How to get the pros without the cons” are this site’s own, not claims from these sources.

Free printable AI cheat sheets, and AI news every week.

Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.

SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK
aipromptgeneer://$ about this siteAI PROMPT GENEERAI explained in plain English. Every claim links to its source, with the date it was checked.NEWSLETTERWhat changed in AI each week, in plain English.SUBSCRIBE FREE ↗AI PROMPT GENEER · MACHINE ASSISTANCE DISCLOSEDEVERY ARTICLE CARRIES ITS OWN CHECKED DATE