A single figure shows up in dozens of small-business advice articles published this year: 67% of customers would rather solve a problem themselves than talk to a person. It traces to a survey Zendesk ran in 2013, back when self-service mostly meant a static FAQ page and a phone tree.
That number is not an outlier. Across the business-optimization advice genre, precise and dramatic percentages get repeated years past their expiration date, while the real, sourced 2026 data on how small businesses actually use automation tells a slower, messier story than any ten-step checklist admits.
A Number Older Than Most Customer Service Chatbots
The 67% figure keeps circulating as though it were fresh. A Forbes Technology Council piece traced it to two specific studies: a Zendesk survey run in 2013, and a 2017 Aspect Software survey, passed around through HubSpot, that put a related preference at 73%. Neither has been repeated at that scale since.
What is stranger is where the number gets re-attributed. One customer-experience blog credits it to older research from Nuance and Coleman Parkes. Another names HubSpot as the source, but for a narrower claim about peer support forums. A third simply calls it “a recent study” with no name attached at all.
The software companies repeating the figure are not neutral messengers. Tidio sells chatbots. Document360 sells knowledge-base software. Raffle sells AI site search. All three build blog content around the same 67%, each pointing customers toward the exact category of tool that number is used to justify.
The 70 Percent Failure Line Has the Same Problem
Business-optimization content leans on a second zombie statistic even harder than the first: the claim that 70% of digital transformations fail. It shows up in consulting decks, LinkedIn posts, and process-improvement listicles as settled fact.
A citation-history analysis published this month traced the 70 percent line’s actual paper trail and found it does not hold up. McKinsey has used the figure in change-management writing, but one of its own articles cites no study, and a later one footnotes back to a Forbes column and that same unsourced piece. A separate researcher who traced the broader “70% of change efforts fail” claim across five published sources found no empirical study behind any of them.
McKinsey’s own measured research says something different. Its 2018 Global Survey of 1,793 executives found that just 16% of digital transformations both improved performance and sustained the gain, with a further 7% improving without holding onto it. Across four survey years going back to 2012, success never once topped 26 percent. Boston Consulting Group’s separate analysis of 850 companies put the share hitting its targets at roughly a third.
- What we know: McKinsey’s own surveyed number is 16% for full, sustained success, and its broader transformation research has never measured above 30% since 2012.
- What we know: Boston Consulting Group’s 850-company analysis found close to two thirds of transformation programs fall short of their goals.
- Unconfirmed: The specific “70% of transformations fail” line has no named underlying study; McKinsey’s own change-management writing footnotes it to a Forbes column and an earlier unsourced McKinsey piece.
- Unconfirmed: The broader “70% of change fails” claim traces to a 1993 reengineering assertion its own author later walked back.
That does not mean transformation and automation projects usually work. It means the specific number everyone quotes was never actually measured. Those are different problems, and business advice content treats them as the same one.
Where Do Small Businesses Actually Stand on Automation?
Nowhere near where the marketing suggests, and nowhere near total failure either. Government transaction data puts real day-to-day operational use of AI tools around 17% to 20%. Self-reported surveys claim more than half of small firms have tried something. Only about a quarter of those ever change a workflow enough to notice a difference.
Four different ways of measuring the same question produce four different answers, and the gap between them is the real story.
| What Was Measured | Source | Reported Figure |
|---|---|---|
| AI used in day-to-day production operations | US Census Bureau, May 2026 | 17% to 20% |
| AI integrated into operations, transaction-based data | JPMorgan Chase Institute, December 2025 | 17.7% |
| Firms under 50 employees using at least one AI tool | 2026 small-business adoption analysis | 58%, up from about 23% in 2023 |
| Adopters whose workflow “meaningfully changed” | Same 2026 adoption analysis | 22% to 31% |
Salesforce’s sixth-edition SMB Trends Report found 91% of small businesses already using AI say it lifted revenue, and 90% say it improved efficiency. Those numbers are real, but they only describe people who stuck with a tool long enough to answer a survey about it. Businesses that tried and quit are not in that sample.
The Gap Between Trying a Tool and Trusting It
Small businesses are not skipping automation out of stubbornness. The barriers that keep tools from becoming habits are specific and well documented, and cost is rarely the top one.
- Understanding the benefit, cited by 62% of small businesses as their reason for not moving past a first try, according to the Small Business Administration’s Office of Advocacy Research Spotlight from September 2025.
- In-house resources, cited by 60% in that same research, who say they lack the staff or time to implement and maintain a tool once it is installed.
- Confidence and skills, named by 44% of UK small business owners as their top barrier in 2026 survey data, ahead of cost at 31% and output-quality worries at 19%.
- Choosing the right tool, a struggle for 47% of small business AI users who told Goldman Sachs they cannot tell which platform to prioritize.
Every one of those is a skills-and-confidence problem, not an access problem. The tools are mostly free or cheap already. What is missing is the time and expertise to make one stick.
Who Gains When Nobody Checks the Math
Vague, dramatic percentages are not accidents. They are useful to specific people, and it is worth naming who.
Content sites built around search traffic profit from round numbers that make a checklist sound authoritative without requiring a reader to click through to a primary source. Software vendors profit twice: first from the traffic the statistic pulls in, then from the sign-ups it is written to justify. Neither has much incentive to note that a figure is thirteen years old or was never actually measured.
The same dynamic shows up outside statistics, in the tools themselves. In small-business bookkeeping, AI systems sold to eliminate mistakes have instead been shown to create a sneakier, harder-to-catch error that surfaces later, often at tax time. The promise and the reality are not the same size, and the gap tends to land on whoever adopted the tool in good faith.
What Actually Predicts Whether Optimization Works
The research that holds up under scrutiny points somewhere unglamorous: culture and follow-through, not the software itself. Analysts who dug into McKinsey’s own transformation data found that organizations investing seriously in cultural change and training saw roughly 5.3 times higher success rates than those that just bought and installed a new platform.
McKinsey’s Global Institute separately found that businesses that do successfully automate a process see productivity gains of 20% to 30% within the first year, which is a real number, just a smaller and slower one than “cut response times by 70%” promises.
The honest version of the advice in most listicles is duller than the marketing version. Pick one process. Measure it before changing anything. Give the new tool a full quarter, not a week, before judging it. Small businesses that treat automation as an ongoing habit, adjusting tools as the business changes rather than installing them once, are the ones that show up in the successful minority rather than the try-once majority.
Frequently Asked Questions
How Can I Tell if a Business Efficiency Statistic Is Reliable?
Check for a named study, a date, and a sample size before repeating any percentage. A red flag is a stat that shows up worded almost identically across unrelated software companies’ blogs, each attributing it to a different or unnamed source, which is exactly the pattern behind the widely repeated 67% self-service figure.
What Do Small Businesses That Succeed at Automation Actually Save?
Among small businesses that report real results, 66% say they save between $500 and $2,000 a month, and 58% save more than 20 hours a month, according to 2025 survey data from Thryv. That is a genuine gain, though it applies only to businesses that got a tool fully working, not everyone who tried one.
Which Automation Use Case Pays Back the Fastest?
AI-powered customer messaging shows the clearest return among small businesses, based on 2026 adoption research. A Kantar and Meta study spanning 22 markets found 73.3% of consumers prefer messaging over other contact methods, 72.4% say they are more likely to buy from a brand that offers it, and 66.8% report frustration when it is missing.
Will Self-Service Tools Replace Customer Service Jobs?
Not entirely, but the labor market is shifting. One frequently cited projection puts the decline in customer support employment at roughly 4% between 2021 and 2031, a gradual shrinkage rather than a sudden replacement, as routine inquiries move to chatbots and portals while complex cases still go to a person.
How Long Does It Take to See Real Automation ROI?
Plan on a full year, not a pilot sprint. McKinsey’s Global Institute ties its 20% to 30% productivity gain figure to a business’s first full year running a process with automation in place, well past the point most small businesses stop to judge whether a new tool is worth keeping.
