Anthropic surveyed 80,000 people across 159 countries about how they use AI today, what they hope it could do for them, and what genuinely worries them. The participants came from 70 languages and every kind of industry. It is one of the most detailed public snapshots we have of how people are actually fitting AI into their lives.
Most AI discussions get stuck on the dramatic end of the spectrum: job loss, existential risk, AGI timelines. This study does something more useful. It asks simpler, harder questions: what does "AI going well" actually feel like for the people using it right now? Where is it working? Where is it falling short? And what do people most want it to do next?
The answers are worth reading carefully — especially if you are running a sales or engineering function in a manufacturing company.
The Number One Thing People Want From AI Is Not Magic. It's Relief.
When Anthropic asked people what they most wanted from AI, the answers were grounded in a way that surprises people who spend too much time reading tech coverage.
The top responses were not about superintelligence or automation replacing whole departments. They were about reducing mental overload, doing better work with less friction, learning faster, managing life logistics, and getting time back.
That is an exhaustion wish, not a technology wish.
Most people are not asking AI to become something futuristic. They are asking it to help with things they already need to do — but with less cognitive load, less time lost to repetitive tasks, and less of that low-grade pressure that comes from having more to handle than hours to handle it.
For manufacturing companies, this matters because the same dynamic plays out at the sales and engineering level. Your senior technical people are exhausted. Not from lack of effort — from carrying knowledge that the rest of the organization can't access without routing a question through them. Every configuration inquiry, every feasibility question, every RFP that lands in the inbox adds to a queue that moves slower than your customers want and faster than your team can sustain.
The relief people are asking for from AI is exactly the relief that AI sales engineers for manufacturers are designed to deliver.
81% Said AI Had Already Helped — But How It Helped Is the Real Story
Anthropic found that 81% of respondents said AI had already moved them closer to something they wanted. That is a meaningful number. But the pattern underneath it is more useful than the headline.
The strongest responses did not come from people who said AI made them faster. They came from people who said AI gave them access to something that previously felt out of reach — not because the information didn't exist, but because it was too expensive, too intimidating, too judgment-heavy, or too time-consuming to navigate without help.
Someone learning to code despite a learning disorder. Someone asking medical questions they were embarrassed to raise with a doctor. Someone getting through documentation that was eating their evenings.
The pattern is consistent: AI creates the most real value when it removes a barrier to access, not just a barrier to speed.
In manufacturing sales, that barrier is technical knowledge. Your distributors and channel partners carry your products alongside ten others. They do not have the depth to answer detailed configuration questions on the first call. Your newer sales reps cannot match what a twenty-year engineering veteran knows about product limits and application edge cases. The knowledge exists inside your organization — but it is locked behind a queue, a relationship, or a meeting that has not been scheduled yet.
That is an access problem. And it is exactly the kind of problem AI solves well.
An AI knowledge base for manufacturers does not just make your team faster. It makes technical knowledge accessible to people who previously could not reach it without calling the one engineer who knows.
The Most Common Fear Is Not Existential. It's That AI Sounds Confident When It's Wrong.
The public debate around AI risk tends toward dramatic framings. But the number one concern in Anthropic's survey was far more practical: AI produces confident-sounding answers that are subtly or significantly wrong. In low-stakes contexts, that is a minor annoyance. In high-stakes contexts — law, medicine, finance, engineering — it is a serious problem.
Manufacturers who have experimented with generic AI tools for technical sales support have run into this directly. A system trained on general web content cannot reliably answer questions about your specific product configurations, tolerances, and application history. It will produce answers that sound plausible while being wrong in ways that matter to a customer or an engineer.
This is why AI RFP automation for manufacturers that is grounded in your own technical documentation produces fundamentally different results from using a general-purpose AI tool. The output is accurate because the system is trained on what your company actually builds and how it has responded to similar requests before — not on a generalized approximation of manufacturing knowledge.
Reliability is not a feature layer you add later. In technical sales, it is the product.
The Same Capabilities People Love Are the Ones They're Anxious About
One of the most honest sections of the Anthropic report is what they call the "light and shade" of AI — the way that the same capabilities people value most are often the ones that worry them most.
AI helps people learn faster, but they worry it makes them think less deeply. It saves time, but they worry it just raises baseline expectations. It handles repetitive tasks, but they worry it creates dependency on something they do not fully control.
Most people are not optimists or pessimists about AI. They are both at once. That is a more accurate picture than most coverage allows.
For manufacturers considering AI in the sales process, this tension shows up in a specific form. The concern is usually something like: if we automate the technical response process, do we lose the engineering judgment that makes our answers trustworthy? Does speed come at the cost of accuracy?
The answer depends entirely on how the system is built. AI that replaces engineering judgment is the wrong model. AI that makes engineering knowledge more accessible — while flagging the requests that genuinely need human review — is a different thing. The goal is not to remove engineers from the equation. It is to remove them from the front of the queue for every question that does not genuinely require their expertise.
Partner enablement tools for manufacturers work the same way. The goal is not to turn channel partners into engineers. It is to give them enough technical grounding to have the first conversation confidently — and to know when to escalate rather than guess.
The Regional Dimension: Emerging Markets See AI as a Ladder, Not a Threat
Anthropic's survey found a consistent geographic split: users in lower- and middle-income countries were significantly more optimistic about AI than users in wealthier regions. In many emerging markets, AI is framed less as a threat and more as a ladder — a way to start businesses, access education, reach customers, and overcome infrastructure gaps that have historically limited what was possible.
For manufacturers with distribution in Southeast Asia, Latin America, or other developing regions, this is relevant context. Your channel partners in these markets are often eager to adopt tools that help them compete with larger distributors. An AI-powered system that helps a regional distributor answer technical questions credibly — without a deep engineering background — is not replacing something they had. It is giving them something they never had access to before.
That is the access dynamic from Anthropic's survey playing out at the channel level.
What This Means for Manufacturing Sales Teams
Anthropic's study is useful because it reframes the question. Most conversations about AI in manufacturing get stuck on process automation: what tasks can be automated, what roles might change, what the technology roadmap looks like. Those are real questions. But they miss what people in the study kept circling back to.
People do not primarily want AI to make them faster. They want it to make their work less draining, more accessible, and more sustainable. They want to spend less time on the repetitive and more time on the work that actually requires their judgment.
In a manufacturing sales context, that looks like this:
A sales engineer who can answer technical questions on the first call instead of the third, because the product knowledge they need is immediately accessible rather than locked in documentation they have not had time to read.
A channel partner who can respond to a customer inquiry without calling your inside team, because the AI system gives them a technically accurate answer based on your actual product history.
An RFP response that takes hours instead of days, because the system pulls from your existing technical documentation and past proposals rather than requiring someone to build the answer from scratch.
A feasibility assessment that happens in minutes rather than days, because the AI can draw on your engineering history to answer "can we build this?" without routing the question through your most overcommitted people.
Those are not futuristic capabilities. They are available now, and manufacturers who deploy them are competing differently than those who are still routing every technical question through the same engineer queue they used ten years ago.
The Practical Takeaway
Anthropic's survey is a useful corrective to both the hype and the fear that dominate AI coverage. The reality, as 80,000 people described it, is quieter and more specific.
People want AI to remove friction from work they already need to do. They want access to knowledge that was previously hard to reach. They want the repetitive cognitive load off their plate so they can focus on the work that actually requires them.
For manufacturing companies, that is a precise brief. The question is not whether AI belongs in your sales process. The question is where the friction is highest — which queue is slowest, which knowledge is most locked up, which partner or sales rep is losing deals because they cannot get a technical answer fast enough — and starting there.
That is where the value is. Not in general AI adoption, but in solving a specific problem that currently costs you deals, time, or engineering capacity.
Source: Anthropic, "What Do People Want From AI?" — global survey of 80,000+ users across 159 countries and 70 languages.
https://www.anthropic.com/research/what-people-want-from-ai
