AI-first company

Claim check

“85% of AI projects fail (Gartner)”

Gartner published an 85% figure about AI projects in February 2018. It was a forecast that “through 2022, 85 percent of AI projects will deliver erroneous outcomes due to bias”, not a count of failed projects. Repeats turned wrong results into failure and the forecast into a fact, and later added causes and dates that are in no Gartner text. The deadline passed in 2022, and Gartner has not published whether the forecast came true.

By
Last reviewed
Corrections
Report an error
Verdict Needs correcting

The source exists, but the claim gets the year, the population or the wording wrong.

Kind of figure Forecast

A prediction about the future.

Holds for AI projects giving wrong results because of bias in data, algorithms or teams, up to 2022. It is what Gartner expected in February 2018, not a count.

The claim

85% of AI projects fail, according to Gartner.

How the claim travelled

Each step is a publication that passed the claim on, oldest first, with its exact words and what changed.

  1. Press release

    Gartner · archived copy

    “Gartner predicts that through 2022, 85 percent of AI projects will deliver erroneous outcomes due to bias in data, algorithms or the teams responsible for managing them.”

    What changed: Origin. Gartner forecasts that 85% of AI projects “will deliver erroneous outcomes due to bias”, for the years up to 2022. The release gives no method, no sample and no meaning of “erroneous outcomes”. It points Gartner clients to a paid report.

  2. Vendor

    BMC Software (Chrissy Kidd) · archived copy

    “Earlier this year, industry research firm Gartner made an audacious prediction: that 85 percent of AI projects won’t deliver for CIOs. That means in 20 artificial intelligence projects, only three will succeed – 17 of them will fall short.”

    What changed: “Will deliver erroneous outcomes due to bias” becomes “won’t deliver for CIOs”. Bias and the 2022 deadline are dropped. The headline adds “up to”, and a subheading asks why so many AI projects “fail”. The figure is still called a prediction.

  3. Blog

    Analytics Vidhya (DataHack Summit 2019) · archived copy

    “Last year, industry research firm Gartner made an audacious prediction: that 85 percent of AI projects don’t deliver. That means in 20 artificial intelligence projects, only three will succeed – 17 of them will fall short. With all the buzz and talk about AI technologies, that’s an interesting and surprising fact.”

    What changed: The text copies BMC’s paragraph almost word for word. “Won’t deliver” becomes “don’t deliver”, in the present tense. The prediction is called a “fact”. The panel’s title makes it “Why do ~85% of AI Projects Fail?”.

  4. Vendor

    TinyFish · archived copy

    “Gartner pegs GenAI project failure at 85%, mostly from inadequate testing.”

    What changed: The figure becomes a failure rate of generative AI projects. The 2018 forecast was about AI projects in general. A cause is added: “inadequate testing”. No Gartner document is named or linked.

  5. Vendor

    Folio3 AI · archived copy

    “85% of AI project failures trace back to poor data quality. (Gartner, 2025)”

    What changed: The 85% becomes a share of failures, not of projects, and a cause is added: “poor data quality”. The figure is dated 2025 and linked to a Gartner statement of February 2025 that has no 85% figure.

What the primary source says

The source is a Gartner press release of 13 February 2018, Gartner Says Nearly Half of CIOs Are Planning to Deploy Artificial Intelligence. Gartner sells research and advice to IT leaders, the chief information officers (CIOs) the release speaks to. The release also advertised Gartner’s Data & Analytics Summit.

How it was done. The release does not say. Its main news is Gartner’s 2018 CIO Agenda Survey, in which 4% of CIOs said they had put AI in place and 46% said they had plans to. The 85% is not tied to that survey. It sits in a section that begins “Conversations with Gartner clients reveal…”. For more, the release points Gartner clients to a paid report, “Lessons From Early AI Projects”, which this check could not read. The release gives no sample, no count of projects and no meaning of “erroneous outcomes”.

Where the 85% appears. Under the heading “3. Plan for Knowledge Transfer”: “Gartner predicts that through 2022, 85 percent of AI projects will deliver erroneous outcomes due to bias in data, algorithms or the teams responsible for managing them.”

What it says. It is a forecast for the years up to 2022. It is about wrong results caused by bias. It is not about projects that were stopped, went over budget or made no money.

What is wrong with the claim

  • It is a forecast, not a count. Gartner wrote “will deliver”, for the years “through 2022”. Repeats say that AI projects “fail”, in the present tense. A conference page in 2019 called the prediction a “fact”.
  • “Erroneous outcomes due to bias” became “fail”. Gartner wrote about wrong results caused by bias in data, algorithms or teams. A blog post by BMC, which sells IT software, turned that into projects that “won’t deliver for CIOs”. A subheading in the same post, and later repeats, turned it into failure.
  • The deadline has passed. The forecast covered the years up to 2022. Gartner has not published whether it came true, and this check found no count of AI projects that gave wrong results because of bias.
  • Later repeats add causes and dates that are in no Gartner text. One vendor applies the 85% to generative AI and blames “inadequate testing”. Another says that 85% of AI project failures “trace back to poor data quality”, dates the figure 2025 and links it to Gartner’s February 2025 statement on data for AI. That statement does not contain an 85% figure.
  • A second 85% from Gartner is about big data. In November 2017, a Gartner analyst, Nick Heudecker, told a TechRepublic writer that the failure rate of big data projects was “[C]loser to 85 percent”. That remark was about big data projects, not AI projects, and gave no method.

What other sources found

This check found no source that counted how many AI projects gave wrong results because of bias. The nearest figures are Gartner’s own surveys on how many AI projects make it into production, that is, into everyday use. In a survey of 699 respondents in the US, Germany and the UK, run from October to December 2021, Gartner found that “on average, 54% of AI projects make it from pilot to production”. In a survey of 644 respondents in the same three countries in late 2023, “on average, only 48% of AI projects make it into production”. These are what respondents told Gartner, not counts of projects. Not reaching production is not the same as giving wrong results, but neither survey comes near 85%.

Gartner’s 2018 release is worth citing for what Gartner expected in 2018 about bias in AI projects. It is not a source for the share of AI projects that fail. For measured results from AI at work, see the studies on this site.

What to cite instead

In February 2018, Gartner predicted that “through 2022, 85 percent of AI projects will deliver erroneous outcomes due to bias in data, algorithms or the teams responsible for managing them”. Gartner gave no method for this forecast, and it is not a count of failed projects. In Gartner’s own later surveys, respondents said that on average 54% (late 2021) and 48% (late 2023) of their AI projects made it into production, that is, into everyday use.

Sources

  1. Gartner (13 February 2018). Gartner Says Nearly Half of CIOs Are Planning to Deploy Artificial Intelligence. Press release. gartner.com
  2. Kidd (18 December 2018). Why does Gartner predict up to 85% of AI projects will “not deliver” for CIOs? BMC Blogs. bmc.com
  3. Analytics Vidhya (2019). Panel Discussion: Why do ~85% of AI Projects Fail? DataHack Summit 2019. analyticsvidhya.com
  4. TinyFish (21 January 2026). Testing Infrastructure Prevents Most Failures. Current newsletter. current.tinyfish.ai
  5. Folio3 AI (24 April 2026). AI Project Failure Rate in 2026: What the Data Shows. folio3.com
  6. Asay (10 November 2017). 85% of big data projects fail, but your developers can help yours succeed. TechRepublic. The separate big data figure. techrepublic.com
  7. Gartner (26 February 2025). Lack of AI-Ready Data Puts AI Projects at Risk. Q&A with Roxane Edjlali. gartner.com
  8. Gartner (22 August 2022). Gartner Survey Reveals 80% of Executives Think Automation Can Be Applied to Any Business Decision. Press release. gartner.com
  9. Gartner (7 May 2024). Gartner Survey Finds Generative AI Is Now the Most Frequently Deployed AI Solution in Organizations. Press release. gartner.com