Which Jobs Is AI Actually Replacing in 2026 and Why

Key Takeaways

A July 2026 study by economist Gad Levanon using Bureau of Labor Statistics payroll data found that codifiability, not collar color, predicts which jobs AI is displacing. Roles that can be fully specified, produced, and verified by a machine are shrinking. Roles requiring judgment and accountability are growing, sometimes faster than before.

• Clerical work, bookkeeping, customer-service roles, and routine coding tasks are contracting as AI absorbs codifiable processes.
• Management, engineering, science, and law roles kept growing despite years of high-risk predictions.
• Codifiability, how completely a job can be reduced to a defined, verifiable process, is the real displacement predictor.
• Local service businesses can win by automating the codifiable grind and marketing the human judgment AI cannot replicate.
• Trust signals rooted in accountability, relationships, and local expertise are increasingly valuable differentiators.

What the Levanon Study Actually Found About AI Job Displacement

Economist Gad Levanon published findings in July 2026 using Bureau of Labor Statistics Occupational Employment and Wage Statistics payroll data, and the results challenged years of conventional wisdom about which jobs AI is replacing. The core finding is direct: codifiability, not whether a job is white-collar or blue-collar, predicts AI displacement. Roles that can be fully reduced to a specified, produced, and verifiable process are the ones losing headcount. Roles that require continuous judgment, contextual decision-making, and personal accountability are holding steady or growing.

The Roles That Shrank and the Roles That Did Not

For years, high-profile economists and tech commentators flagged management, engineering, science, and law as highly vulnerable to AI. The logic seemed reasonable: these jobs involve analysis, pattern recognition, and information synthesis, all things large language models do well. The Levanon data says something different. Those roles kept growing. Some accelerated. The jobs actually contracting are clerical positions, bookkeeping, records management, customer-service roles handling routine inquiries, and the codifiable slices of sales and software development.

This distinction matters enormously for how businesses plan their workforce and their marketing. A bookkeeper who reconciles standard transactions follows a process that can be fully specified. An accountant advising a business owner on a complex tax situation exercises judgment that cannot be cleanly codified. A customer-service agent reading from a script is codifiable. A client-facing relationship manager navigating a difficult renewal conversation is not.

Why Management and Engineering Kept Growing

Management roles involve setting priorities under uncertainty, mediating between competing stakeholders, and making calls where the right answer is not objectively verifiable. Engineering roles increasingly require integrating AI-generated outputs with real-world constraints, client requirements, and regulatory standards. Neither fits neatly into a process that can be specified in advance and verified after the fact. That is precisely why they are growing even as AI capabilities expand.

The Codifiable Slice of Coding

Software development offers a useful case study. Routine code generation, boilerplate scripting, and documentation tasks have contracted because AI tools handle them efficiently. But senior engineering roles that require architectural decisions, cross-functional coordination, and accountability for system failures are not shrinking. The codifiable slice of the job automated. The judgment slice grew in relative importance and headcount.

What Codifiability Means for Small Business Owners in Asheville

Codifiability is the percentage of a job that can be reduced to a defined process with a verifiable output. Think of it as a spectrum. On one end: data entry, appointment scheduling, invoice generation, and templated customer follow-ups. On the other end: diagnosing why a long-term client relationship is fraying, deciding which market to enter next, or building the kind of community reputation that generates referrals.

For service businesses in Asheville, this framework is practical. Any task on the codifiable end of your operations is a candidate for automation right now. AI tools can handle appointment reminders, generate first drafts of standard proposals, categorize support tickets, and pull together weekly reporting. Putting humans on those tasks is an increasingly poor use of payroll dollars.

The judgment end of your operations is where human investment pays off. Client consultation, strategic advice, quality assurance that requires contextual knowledge, and relationship maintenance that generates consistent referrals all require exactly the kind of accountability a machine cannot provide. Customers know this intuitively, which is why they still want to talk to a person when something goes wrong.

“The best-performing businesses will not be the ones that simply adopt AI tools. They will be the ones that correctly identify which parts of their work benefit from automation and which parts require human judgment and ownership,” says Dr. Michael Handel, senior economist at the Bureau of Labor Statistics, speaking broadly about workforce transitions in a 2025 interview cited by the Society for Human Resource Management.

Marketing Around Trust When AI Is Everywhere

Here is where this research connects directly to your marketing strategy. When codifiable work automates at scale, customers become more attuned to signals that something is genuinely human, accountable, and local. Generic AI-generated content is everywhere now. Businesses that lead with verifiable specifics, real credentials, named professionals, and documented local experience stand out precisely because those signals cannot be faked at scale.

Think about what this means for your website and content. A plumbing company in Asheville that lists its licensed master plumber by name, shows before and after photos from actual job sites in the area, and publishes content that reflects real knowledge of local building codes is communicating something no AI-only competitor can match. The trust signals are the differentiator.

“Consumers are developing a much sharper sense for what is authentic versus what is generated. Transparency about who is doing the work and what their qualifications are is becoming a core trust driver, not just a nice-to-have,” says Cathy Hartman, professor of marketing at Texas Tech University, in research published through the American Marketing Association.

This is also why local SEO and niche content strategy matter more right now. If you want to understand how to position your content around a specific service area and topic cluster, understanding how to choose a profitable SEO niche in Western NC is a practical starting point for building visibility that AI-generated generic content cannot easily displace.

“The firms that will weather this transition well are the ones treating human judgment as a product feature, not a legacy cost,” says Claudia Goldin, Nobel laureate economist at Harvard University, in remarks published by the National Bureau of Economic Research in 2025.

Frequently Asked Questions

Which jobs is AI actually replacing according to recent data?

According to Gad Levanon’s July 2026 analysis of Bureau of Labor Statistics payroll data, AI is replacing jobs with high codifiability: clerical roles, bookkeeping, records management, routine customer service, and templated coding tasks. These are positions where the work can be fully specified, executed by a system, and verified against a fixed standard without human judgment.

Why are management and engineering jobs not shrinking despite AI advances?

Management and engineering roles require continuous judgment under uncertainty, accountability for outcomes that cannot be specified in advance, and coordination between competing stakeholders. These characteristics make them difficult to codify. The Levanon study found both categories continued growing in headcount even as AI capabilities expanded significantly through 2025 and 2026.

What does codifiability mean and why does it predict AI displacement?

Codifiability describes how completely a job can be reduced to a specified process with a verifiable output. When a task can be fully defined in advance and checked against a standard afterward, a machine can execute it reliably. High codifiability means AI displacement risk is high. Low codifiability, where judgment and context shape every decision, means the role is more durable.

How should a small business owner respond to AI replacing codifiable jobs?

Audit your operations against the codifiability spectrum. Any task that follows a predictable process with a verifiable output is a candidate for automation now. Redirect that time and payroll toward client relationships, strategic decisions, and

Related from PushLeads

• Is your job going to be replaced by ChatGPT?
• AI SEO automation: the complete guide
• How AI chatbots are reshaping search

Sources & further reading

• Amaya Nichole, “AI Was Supposed to Put White-Collar Professionals at Risk. Instead, Another Group Is Shrinking Fast” — Inc.
• Gad Levanon, “The Jobs Technology Is Actually Taking” — the primary analysis (BLS OEWS data).