83% of Black Respondents Know AI — So Why Aren’t More Using It to Build?

AI in the Black community is not starting from zero. In a 2025 Jobs for the Future survey, 83 percent of Black learners and workers surveyed said they were familiar with artificial intelligence, and more than half reported using AI tools daily or weekly.

That finding matters because it challenges an easy assumption. In fact, the problem is not simply that Black workers have never encountered AI. Many already have.

But familiarity is only the first layer.

Knowing an AI tool exists is different from knowing how to direct it. Likewise, using a chatbot is different from redesigning a workflow. And generating an answer is different from producing work that creates measurable value.

Saving twenty minutes is useful. But turning that efficiency into stronger capability, better income, a business asset, or greater ownership is something else.

Familiarity without application is not literacy.

So the real opportunity for AI in the Black community is to move from awareness to deliberate use: solving problems, producing value, expanding capacity, and building something that lasts beyond a single prompt.

What 83% Tells Us About AI in the Black Community

Specifically, the 83 percent figure comes from a Jobs for the Future survey of 2,754 learners and workers, including 529 respondents who identified as Black or African American.

Among those Black respondents, 83 percent said they were familiar with AI, compared with 75 percent of the full sample. And 53 percent said they used AI tools daily or weekly, compared with 39 percent across the overall survey.

So the evidence does not support a story in which Black communities are simply standing outside the technology looking in.

A significant share is already inside the door.

AI Use Is Not the Same as Economic Position

Even so, frequency alone does not tell us what the technology is doing for the user.

Two people can use AI every day and build very different forms of capability.

One person may use it to summarize articles, rewrite text, answer questions, or generate entertainment. By contrast, another may use the same class of technology to analyze customer data, automate a recurring workflow, produce client deliverables, prototype a service, improve a sales process, learn a technical skill, or operate a small business with less overhead.

Both are AI users.

Yet the economic consequences are different.

That distinction matters because adoption statistics can make a community look technologically prepared while hiding a deeper question: what kind of use is compounding?

The New Digital Divide Is About Capability

The first digital divide was largely about access. Who had a computer? Who had broadband? Who could get online?

Access still matters. But AI adds another layer.

People can now have access to similar tools while receiving very different returns from them.

As a result, the gap increasingly sits between those who can ask AI for something and those who can integrate AI into a larger system of judgment, domain knowledge, workflow, verification, and production.

That is a capability divide.

The Capability Gap in AI in the Black Community

Prompting matters, but prompt engineering has been oversold as the central AI skill.

A clever prompt can improve an answer. But it cannot decide whether the answer solves the right problem.

It also cannot supply domain expertise you do not have. Nor can it determine whether a source is trustworthy, understand every organizational constraint, or decide how much risk you should accept.

Most importantly, it cannot own the consequence of the decision.

So durable AI in the Black community takes more than access to a chatbot.

It requires a system around the tool.

Why the Prompt Is Only One Layer

Without those supporting layers, AI can make consumption faster without making the user materially stronger.

Better prompts may produce cleaner output. But if the user cannot verify the result, connect it to real work, or improve a recurring process, the capability remains shallow.

By contrast, a person who combines strong prompts with domain knowledge and judgment can use AI as part of a larger operating system.

That is the difference between tool familiarity and capability.

Five Levels of AI Capability

A more useful way to read AI adoption is as a progression rather than a binary choice between user and non-user.

Level 1: Familiarity

You know the tools exist. You understand generally what generative AI can do, and you may have tried one or more systems.

This matters because unfamiliarity creates friction. Still, familiarity by itself creates little economic advantage.

Level 2: Experimentation

Next, you begin testing the technology against real tasks.

You might use AI to draft an email, explain a concept, summarize research, compare options, generate ideas, or organize information.

At this stage, the goal isn’t mastery — it’s pattern recognition.

Along the way, you begin learning what the tool handles well, where it fails, what requires verification, and which tasks improve when the human provides better context.

Level 3: Application

Application begins when AI becomes attached to a recurring problem rather than occasional curiosity.

For example, a recruiter may use it to organize candidate notes before human review. Similarly, a tradesperson may use it to turn field notes into cleaner estimates or customer explanations. Meanwhile, a nonprofit worker may accelerate grant research, and a small-business owner may structure customer follow-up or analyze recurring questions.

The important shift is from What can this tool do? to Which part of my work should this tool improve?

Level 4: Production

Production means the tool is helping you create useful output consistently.

That output could be analysis, proposals, code, designs, training material, customer support, documentation, marketing assets, research, financial models, operating procedures, or products.

At this level, AI becomes part of capacity.

But increased production raises another question: who captures the value?

Level 5: Leverage

Leverage appears when AI allows the user to do something economically larger than before.

Perhaps one employee can manage a broader scope. Similarly, a freelancer can serve additional clients without reducing quality, and a business owner can test ideas with lower upfront cost. Or a worker may pick up an adjacent capability faster than they otherwise could.

This is where the technology becomes strategically consequential.

The goal is not simply to use AI more. Instead, the goal is to move upward through the levels.

What It Means to Build With AI

“Build” can sound like entrepreneurship language, but the concept is broader.

You are building whenever AI leaves you with something more durable than the conversation itself.

Build a Repeatable Workflow

If you perform the same information-heavy task every week, document the process.

First, identify what requires human judgment. Then figure out what can be accelerated, what needs verification, and what should never be delegated.

After that, build a repeatable workflow rather than improvising prompts every time.

The workflow becomes an asset because it preserves learning.

Build a Portable Skill

Do not learn only how one employer uses one AI product.

Learn the underlying capability instead.

That capability might be research synthesis, data interpretation, workflow automation, structured writing, quality assurance, customer operations, knowledge management, or AI-assisted analysis.

If the tool changes, the capability should still travel.

Build Evidence

AI literacy becomes more valuable when somebody else can see what you can do with it.

Keep examples of improved work. Also measure time saved when possible. Then document the before-and-after process, and create a portfolio example that does not expose confidential information.

“I know AI” is a weak career signal.

By contrast, “I redesigned this process and cut three hours of repetitive work while adding a verification step” is evidence.

Build an Asset You Control

This is the highest-leverage move.

Use AI to help create something that remains useful after the session ends: a training system, template library, research database, curriculum, operating manual, small software tool, intellectual property, customer process, service offering, or body of published work.

The asset does not have to become a startup.

It simply needs to compound.

Consumer Use and Builder Use in AI in the Black Community

The distinction is not moral. There is nothing wrong with using AI for convenience.

Sometimes convenience is exactly what a person needs.

But if the question is economic mobility, convenience can’t be the only metric.

Consumer Use Solves the Immediate Task

Consumer use asks:

What can AI do for me right now?

That may mean rewriting a message, summarizing an article, generating ideas, or speeding up a one-time task.

Those uses are valuable, especially when they reduce friction.

Builder Use Creates Repeatable Capability

Builder use asks a different question:

What can I become capable of doing repeatedly because this tool exists?

That second question is more demanding, because it keeps the human at the center of the system. AI doesn’t become the capability. Instead, the person does.

AI Fluency Should Reduce Dependency

That principle matters.

Good AI use should help you understand the work better, make decisions faster, test ideas more cheaply, or create stronger output.

On the flip side, bad use can create the opposite effect. If AI performs every difficult cognitive step, the user may become faster while understanding less.

So keep certain functions human:

• defining the real problem;
• choosing the objective;
• evaluating tradeoffs;
• checking consequential claims;
• understanding context;
• accepting responsibility for the final decision.

That is not resistance to AI. It is disciplined use.

Where AI Leverage Could Matter Most

The most important promise of AI may not be replacing highly paid specialists.

Instead, it may be reducing the cost of capabilities that previously required more money, more staff, or more institutional access.

Research assistance can become cheaper. So can drafting. Basic technical prototyping is getting easier to justify, too, and translation, analysis, tutoring, design iteration, and administrative support may also become more accessible.

For people historically operating with less capital or thinner institutional infrastructure, lower capability costs could matter enormously.

Lower Cost Does Not Guarantee Equal Outcomes

Still, lower cost does not automatically produce equal results.

People need reliable access, digital literacy, time to experiment, quality training, domain knowledge, and enough trust to use the technology without surrendering judgment.

Without those supports, nominal access can hide major differences in practical capability.

The Question Is Who Converts Access Into Capacity

This is where the next divide may form around AI in the Black community.

If one group uses AI mainly to consume faster while another uses it to increase output, learn valuable skills, build businesses, improve workflows, and create assets, similar access can still produce widening economic differences.

That is why the 83 percent figure should be treated as encouraging but incomplete.

Awareness is not the finish line. Instead, it is starting infrastructure.

A 30-Day AI Capability Build

You do not need to become an AI specialist.

Instead, choose one real problem.

Week 1: Observe

Track repetitive information work for five days.

Notice what consumes time: research, drafting, formatting, summarizing, scheduling, organizing, comparing, documenting, or communicating.

Then choose one task.

Week 2: Experiment

Test AI against that task three different ways.

Then compare the output with your normal process.

Finally, record where the tool helps, where it introduces errors, and what context improves the result.

Week 3: Systemize

Turn what worked into a repeatable process.

Then write the inputs, the human review points, the verification rule, and the final output standard.

At that point, you have more than a prompt. You have a workflow.

Week 4: Prove

Finally, measure one result.

Did the process save time? Improve quality? Increase output? Help you perform a task you could not perform before?

Document the evidence, because that evidence is the beginning of capability.

The Groundwork: AI in the Black Community Beyond Familiarity

The encouraging story about AI in the Black community is that familiarity is already substantial among the Black learners and workers JFF surveyed. Regular use is also not unusual.

So the strategic question is what comes next.

Do we remain consumers of systems other people build? Do we use AI only to make existing tasks a little faster? Or do we convert access into skills, workflows, evidence, intellectual property, businesses, stronger careers, and institutions with greater capacity?

The Goal Is Not to Turn Everyone Into an AI Expert

There is no reason every person needs to become a founder, programmer, or AI specialist.

That is not the objective.

Instead, the objective is to become harder to separate from the value being created.

So learn the tool. Understand its limits. Attach it to a real problem. Build a repeatable capability. Preserve the evidence. Then decide where that capability can create leverage.

Familiarity gets you into the room. Capability changes what you can do once you are there.

Groundwork Daily Builder

Jordan Avery

Jordan Avery builds Future Literacy, examining how people can recognize change early, think clearly under uncertainty, understand emerging systems, and build capabilities that remain useful as conditions evolve.

Meet Jordan Avery →