Ford Motor Company laid off hundreds of experienced quality engineers. Then it replaced them with AI-powered cameras and automated inspection systems. The AI missed things the veterans would have caught. Defects slipped through. So Ford did what any honest company would do: it admitted the mistake and rehired more than 300 of those inspectors. 1
That story should hit every MSP owner who has been pitched an “AI-powered” solution in the last 18 months. Not because AI does not work. It does. But because the gap between what vendors claim and what their tools actually do has become a chasm. And most buyers in the MSP channel do not have the technical depth to test the claims before they sign the contract.
This is AI washing. It is the practice of overstating the role of artificial intelligence in a product or service to make it seem more capable than it actually is. The term was coined by the AI Now Institute at New York University in 2019. 2 It works the same way greenwashing does: slap the label on the box, charge a premium, and hope nobody looks under the hood.
Why the MSP Channel Is a Target
Managed service providers are caught in a squeeze. Clients are asking about AI. Competitors are advertising it. Vendors are promising that their platforms will “leverage AI” to reduce ticket resolution time, predict failures, automate documentation, and half a dozen other things that sound game-changing in a demo.
The problem is that the MSP channel has always been susceptible to vendor hype. The RMM space is full of tools that promised “set it and forget it” monitoring. The PSA platforms claimed they would “automate your entire workflow.” The backup vendors said their products would “eliminate data loss.” Some of those claims were mostly true. Others were mostly marketing. The same pattern is repeating with AI, except the buzzwords are more impressive and the buyers are more anxious to believe.
The Washington Post defined AI washing in 2026 as “a trend for bosses to blame layoffs on the productive capabilities of AI and its ability to replace workers, even when job cuts may have little to do with the technology.” 2 That definition cuts both ways. Vendors use AI to justify replacing human technicians. Then when the automation fails, the MSP is left holding the bag with a client who expected human-level judgment.
The Five Claims That Should Make You Ask Questions
Not every vendor claiming AI is lying. Some have built genuinely useful tools. But there are five specific claims that should trigger deeper scrutiny before you commit budget.
Claim one: “Our AI resolves tickets automatically.” Ask what percentage of tickets actually get resolved without human intervention. If the vendor cannot give you a number, they do not know. If the number is below 15 percent, you are buying a tool that handles the easy stuff and leaves your technicians with the same complex work they always had. That has value. Just do not pay a premium price for “automation” that only works on password resets.
Claim two: “Our AI predicts failures before they happen.” Ask what the false positive rate is. Every predictive model generates false alarms. If the vendor says they do not have false positives, they are not measuring. A prediction engine that cries wolf 40 times a day gets ignored by your technicians on day three. The useful question is not whether it predicts failures. It is whether your team will still be paying attention when the real one comes.
Claim three: “Our AI writes documentation for you.” Ask to see examples from environments like yours. AI-generated documentation is only as good as the data it was trained on. If the vendor trained their model on clean, well-structured ticket data from a 500-seat environment, and your clients are 15-person shops with flat networks and no documentation culture, the output will be generic at best. Documentation tools work. They just work better when someone with context reviews the output.
Claim four: “Our AI replaces the need for a vCIO.” This is the most dangerous claim. A vCIO does not just produce reports. They sit across from a business owner and explain why the ERP migration matters to their bottom line. They handle politics. They know when a client is about to make a bad decision and how to redirect without damaging the relationship. No AI does that. If a vendor is telling you their tool replaces strategic advisory services, they have never actually done vCIO work.
Claim five: “Our AI is proprietary.” Ask what model they are running under the hood. Many “AI-powered” MSP tools are thin wrappers around OpenAI or Anthropic APIs with a custom prompt layer. That is not necessarily bad, but you should know what you are paying for. If the vendor’s entire AI advantage is a prompt template, you could build the same thing yourself for the cost of an API key.
The Ford Lesson
Ford’s Charles Poon put it directly: “Artificial intelligence is a fantastic tool, but it’s only as good as the information you use to train it.” 1 That sentence should be printed on a card and handed to every vendor who pitches AI to your company.
The second half of his statement matters just as much: “Over prior years, we didn’t pay as much attention as we should have to the experience of our most knowledgeable engineers that have been with us through many product cycles.” Ford made the same mistake MSPs make when they buy AI tools. They assumed the technology could replace institutional knowledge. It cannot. It can augment it. But augmentation requires the institutional knowledge to already be there.
The MSPs that get genuine value from AI are the ones that already have disciplined processes, clean data, and experienced technicians. The AI makes those teams faster. The MSPs that buy AI hoping it will fix a broken operation end up with expensive tools producing expensive nonsense.
The Slop Problem
There is a deeper issue beneath AI washing. Even vendors who genuinely believe in their AI may be selling tools built on contaminated foundations. The training data that powers large language models has been doubling every nine months since 2010. 3 The internet is running out of clean, original content to feed the models.
So AI companies started paying contractors to generate fresh training data. And those contractors, working for low wages on short contracts, started using other AI chatbots to produce the data they were being paid to create. One contractor told New Scientist: “It’s very widespread. Every company I’ve worked for has had explicit guidelines around it and they clearly do try to catch people out, so I think they do care. But I don’t think they can stop it.” 3
This is AI eating AI. Experts have long warned that this kind of feedback loop destabilizes models. When training data is itself AI-generated output, the models drift. They become less reliable, more prone to hallucination, more likely to reproduce the patterns of other models instead of ground truth. The contractor’s summary: “If these companies want quality data, then they should offer quality contracts.”
For MSPs, this matters because the AI tools you are evaluating may have been trained on data that was itself AI-generated. The vendor is not lying to you on purpose. They may genuinely believe their model works. But if the training data was slop, the output will be slop too. And your technicians will be the ones sorting through it.
How to Evaluate AI Claims
Before buying any AI-powered tool, run it through three filters. First, ask for a proof of concept in your own environment. Not a demo. Not a sandbox. Your tickets, your data, your clients. If the vendor refuses, they are selling confidence, not capability.
Second, ask what happens when the AI is wrong. Every model makes mistakes. The question is whether the tool catches its own errors or whether your technicians become the safety net. If the answer is “our technicians review everything,” you are paying for AI and still doing the work.
Third, ask what the vendor’s roadmap looks like for the next 12 months. AI capabilities are moving fast. If the vendor cannot articulate where their product is going, they are riding a hype cycle, not building a platform. You do not want to bet your operation on a company that is chasing buzzwords.
Where This Leaves You
AI washing is not going away. The FTC has issued policy statements warning against it. 4 The SEC has fined companies for overstating AI capabilities to investors. But regulation moves slowly. In the meantime, the burden falls on buyers to separate substance from marketing.
The next time a vendor tells you their product is “AI-powered,” ask them what that means in practice. Ask for numbers. Ask for references from MSPs your size. Ask what happens when it fails. If the answers are vague, the AI probably is too.
Ford figured out that 300 veteran engineers were worth more than 900 cameras running software that did not understand what it was looking at. The lesson is not that AI is useless. It is that experience still matters. The MSPs that understand that distinction will be the ones who actually get value from AI. The ones who do not will be rehiring their technicians in two years.
Sources
1 Liv McMahon, “Ford rehires human engineers after AI fails to match quality checks,” BBC News, 2026.
2 Wikipedia, “AI washing,” 2026.
4 Federal Trade Commission, “FTC warns against AI washing,” 2024.
