AI Can Do Your Job “Good Enough.” Here’s Why That Should Change How You Job Search.

A new MIT study tested 40 AI models against 11,500 real workplace tasks and had 17,000 workers in those fields evaluate the results. The headline finding: AI can now complete about 65% of text-based work tasks at what researchers called a “minimally sufficient” level. Not great. Not impressive. Good enough to pass.

That’s a weird place for AI to land, and it has direct implications for how you should be running your job search. If AI can do a passable version of most knowledge work, then AI can also do a passable version of most job applications. Which means personalized job outreach, the kind that shows a real human did the thinking, is now the clearest way to separate yourself from an increasingly AI-generated applicant pile.

What “minimally sufficient” actually means

The MIT researchers weren’t testing whether AI could do work well. They were testing whether it could do work well enough that a coworker wouldn’t need to redo it. The bar was “acceptable output, no edits required.”

By that standard, AI hit 50% in 2024 and 65% in 2025. The researchers project it’ll reach 80-95% of text-based tasks by 2029. But here’s what matters: even at 65%, the work coming out of these models is consistently described as adequate, not good. The Fortune coverage of the study used the phrase “good enough to pass but not good enough to impress,” which captures it well.

Performance varied dramatically by field. Legal work had the lowest AI success rate at 47%. Legal tasks require precision, judgment, and the kind of strategic thinking that current models handle poorly — you can’t have an AI that’s “pretty close” on a contract clause. Installation, maintenance, and repair tasks scored highest at 73%, mostly because AI can automate the administrative paperwork that surrounds manual work without touching the wrench-turning itself. Managerial tasks landed at 53%. AI did decently on planning and writing but fell apart on coordination and decision-making, which is basically the whole job description for most managers.

The study’s framing was deliberate: AI’s impact on work is a “rising tide,” not a “crashing wave.” Jobs won’t vanish overnight in specific sectors. Instead, work will change broadly and gradually across most of them. For job seekers, this means the ground is shifting under every role, but not fast enough to make any single job obsolete tomorrow. The uncertainty itself is the problem.

The gap between what AI can do and what companies are actually doing

This is where it gets interesting for job seekers.

AI can theoretically handle a lot. Companies aren’t actually using it to replace that much. According to Fortune, only 9% of companies surveyed said AI had fully replaced certain roles. Another 45% said it had partially reduced the need for new hires. That leaves nearly half of companies reporting no meaningful headcount impact from AI at all.

The implementation gap is real and well-documented. Integrating AI into workflows is expensive, slow, and unreliable enough that most organizations are still in pilot mode. MIT Technology Review reported in January that the era of “flashy AI” is ending, with AI embedding into everyday workflows but doing so gradually. The biggest promise is in agentic AI — systems that can work independently — but reliability remains the central unsolved problem.

What this means practically: companies are reorganizing around AI but not yet executing on the reorganization. They’re restructuring teams, redefining roles, and in many cases freezing or slowing hiring while they figure out which positions are genuinely redundant. Nearly 60% of hiring managers surveyed by Resume.org said they plan layoffs in 2026, and AI or automation was the most-cited reason, according to CBS News. But the jobs aren’t disappearing because AI replaced the workers. They’re disappearing because leadership read the same MIT headlines you did and decided to preemptively cut headcount.

The result is a job market that’s tightening because of the idea of AI, not the reality of AI. Companies are in a holding pattern. Hiring is slow. The “low-hire, low-fire” dynamic that Indeed Hiring Lab described in their March 2026 report applies here: organizations are afraid to add headcount, and workers already employed are afraid to leave. Both sides are waiting for clarity that isn’t coming soon.

What this means for the job application response rate

So what does any of this have to do with finding a job?

The same AI that produces “minimally sufficient” work in offices also produces minimally sufficient job applications. Tools that auto-generate resumes, cover letters, and application responses have flooded the market. According to Fortune, 38% of job seekers now use AI tools to mass-apply. When you look at what that output actually looks like, it makes sense. The applications are competent. They hit the right keywords. They’re formatted properly. They sound exactly like every other AI-generated application in the pile.

The job application response rate has dropped roughly 3x since 2021, according to data from Upplai. That collapse corresponds almost perfectly with the rise of AI-powered mass applications. More applications per posting, fewer humans reviewing them, lower quality signal for everyone involved.

If AI produces “good enough” work, and a third of applicants are using AI to produce “good enough” applications, then “good enough” is now the floor, not the ceiling. The bar for standing out has shifted from “submit a competent application” to “demonstrate that a human being, specifically you, put actual thought into this.”

Hiring managers are starting to notice. Recruiters report that many applications now read identically — same structure, same buzzwords, same polished-but-hollow feel. When the pile is full of applications that all look like they were generated by the same model, the one that doesn’t look generated stands out immediately.

Personalized job outreach is the one thing AI can’t fake

The MIT study found that AI struggles most with tasks requiring judgment, coordination, and context-specific thinking. Those are the same skills you demonstrate when you research a specific hiring manager and write them a message that couldn’t have come from a bot.

When you research a hiring manager, understand what their team is working on, and write a message that connects your experience to their situation, you’re doing something AI can’t fake at a passable level. It takes judgment to pick which roles to pursue and what to say about them. A bot can’t demonstrate that it read someone’s recent LinkedIn post about a product launch and connected it to a specific skill on your resume. You can.

The data backs this up. Recruiter-sourced candidates are 8x more likely to be hired than job board applicants, according to Upplai. Direct sourcing is 2.5% of applications but accounts for nearly 10% of hires. The reason is partly structural, you bypass the ATS, but it’s also about signal quality. A personalized outreach message to a hiring manager says “I chose you specifically” in a way that a well-formatted application through a portal never can.

In a market where AI makes the average application indistinguishable from every other average application, the advantage goes to the candidates who do something demonstrably non-average. And the most accessible version of “non-average” is writing directly to the person doing the hiring.

How to do personalized job outreach without burning out

The process takes more time per role than clicking “Apply.” That’s the whole point, and also the reason most people skip it.

Start with the job posting. Use it for what it’s good at: understanding what the company needs and what team the role sits in. Then stop. Don’t click “Apply.”

Find the hiring manager. LinkedIn is the fastest route. Look for the person who manages the department or function described in the posting. Check their recent activity — what they’ve posted about, who they’ve recently hired, what challenges they seem focused on. This tells you whether the role is real and gives you material for your message.

Write something specific. Not a cover letter. Not a template. A short message — maybe 100 words — that references something real about their team or their situation and explains, briefly, how your background connects. End with a specific ask: a 15-minute conversation this week. Make it easy to say yes.

Follow up once if you don’t hear back. Then move on. Personalized job outreach works because of the accumulated pattern: 20-30 well-researched contacts over a few weeks beats 200 auto-applied listings every time. The volume is lower. The conversations are real. The response rates are in a different universe.

The math bears repeating: direct outreach candidates convert at 4-8%, compared to 0.1% for mass-apply tools. When your competition is submitting AI-generated applications by the hundreds, spending 20 minutes on research and a real message is the closest thing to a cheat code this job market has.

The “good enough” economy and what it means for your search

AI is making work “good enough” across the board. That includes job applications. In a market where minimally sufficient is the default, the candidates who land offers are the ones who do something that can’t be generated by clicking a button.

Personalized job outreach is that something. It’s harder than mass-applying. It takes more time per role. But the conversion rates are 40-80x better than auto-apply tools, and the conversations it starts are with actual decision makers, not ATS filters.

The hard part is the research: finding the right person, understanding their context, writing something that sounds like it came from a human who cared enough to do the homework. FoxHire.AI automates that research-to-outreach pipeline so you can focus on the conversations that actually matter, not the application grind that AI has made pointless.

AI Can Do Your Job "Good Enough." Here's Why That Should Change How You Job Search.