AI in the Inbox: One in Four Workers Now Write with Chatbots as Managers Automate Reviews and Layoffs

Inside offices across the United States, the inbox has become a shared space between humans and machines. A recent ZeroBounce survey of a thousand professionals shows that roughly one in four employees now use AI tools every day to draft or polish their emails. Among technology workers, that number rises to about one in three.

What began as a way to fix grammar and tone has become something larger. More than half of all employees say AI makes them feel more confident in their writing. Yet that comfort often turns into reliance. Around eight percent admit they struggle to write emails without help, and fourteen percent have sent sensitive messages copied directly from AI-generated text without editing a word.

Automation Creeps into Management Tasks

Managers are no exception. Forty-one percent say they have used AI to draft or revise performance reviews. Seventeen percent admit they have relied on it when preparing layoff notifications. The trend appears strongest in marketing and technology departments, where digital tools are deeply embedded in daily operations.

On average, managers estimate that about sixteen percent of the messages they send are written by AI. A smaller group, roughly one in twelve, say half or more of their correspondence now originates from a chatbot. The speed and polish are tempting. The result, however, is that formal communication (once built on personal judgment) has started to sound uniformly synthetic.

Workers Notice the Shift in Tone

Employees are growing wary of how automated their offices have become. A quarter suspect they have already received an AI-written performance review. Among tech employees, that suspicion jumps to thirty-seven percent. Sixteen percent of those who have been laid off believe the email ending their job was generated by AI, and nearly a fifth of them said the experience brought them to tears.

Even when emotions are not at stake, many notice the sameness in tone. One in five employees say they have seen identical AI-generated emails sent by different coworkers. Seventeen percent feel more anxious when writing without AI than when using it. That anxiety is highest among healthcare workers and millennials, groups often pressured to maintain professional polish under time constraints.

Confidence, Dependence, and the Disappearing Human Voice

While forty percent of employees believe AI should never be used for sensitive messages, more than half think it can improve clarity if paired with genuine human oversight. The division reveals how workplace communication is entering a new gray zone, where efficiency and empathy often compete for space.

AI’s impact goes beyond time-saving convenience. It reshapes how people feel about their own ability to communicate. For some, automation eases the fear of misphrasing or sounding unprofessional. For others, it dulls emotional honesty, creating a kind of linguistic distance between sender and recipient. When a carefully worded review or farewell note arrives, few can tell whether it came from a person or a prompt.

A Cultural Turning Point for Office Communication

The growing use of AI in professional writing marks a cultural shift rather than a passing experiment. The corporate inbox has become a test site for how far automation can stretch before sincerity breaks. What once relied on human judgment is now managed by tools that optimize for readability and tone but lack intuition.

AI may continue to refine the language of work, but it cannot replace the nuance of real empathy. Used responsibly, it can polish sentences and reduce anxiety. Used without care, it risks turning vital moments into transactions. The ZeroBounce findings suggest a workforce learning to balance convenience with conscience, one email at a time.

Notes: This post was edited/created using GenAI tools.

Read next: How Entrepreneurs and Creators are Shaping Their Own Brands Without Design Degrees
by Irfan Ahmad via Digital Information World

Making Data Work Across Systems Without Opening Attack Vectors

Most modern stacks aren’t built to operate around one big app anymore. It’s more like a patchwork of apps, services, scripts, and third-party tools that are all stitched together. To make them functional, APIs feed data to dashboards, webhooks trigger jobs, and cloud services sync with internal systems.

All of this makes today's systems extremely fast and flexible. But it can also get very messy, especially when it comes to security.

As data moves between environments and vendors, the risks shift from obvious break-ins and brute-force attacks to more subtle problems like misconfiguration, inconsistent policies, shadow integrations, and access patterns that no one is watching closely.

You don’t need a headline-grabbing breach to get hurt. One overly permissive key or a forgotten token can be enough.

Image: freepik

The Hidden Cost of Connected Systems

Interoperability is both a huge benefit and a liability at the same time (as with most new tech advancements we have all come to enjoy).

The more systems you connect, the more credentials you create, and the more assumptions you make about trust. A backend service calls an external API and stores the results in a shared database. A third-party tool gets token-based access to internal data. An ETL job transfers information between regions on a scheduled basis. Every one of those links widens the attack surface.

The tricky part is visibility. In a distributed setup, it’s easy to lose track of who can see what and which pieces can communicate with each other. That’s where simple mistakes become expensive.

Architecture is the Real Security Layer

Good security is not just a list of tools. It is the way your systems are assembled and how the whole infrastructure supports itself. It’s important to remember that everything is interconnected these days. For better or worse, the old idea of a single perimeter fell apart some years ago.

This means you need to control access at the junctions, not only at the edge.

That is why more teams lean toward modular, distributed designs that embed security into the data path. Some organizations borrow ideas from cybersecurity mesh architecture (CSMA) , which treats each system and identity as its security boundary and verifies trust continuously across a fragmented environment.

You don’t have to go all in on a new framework to benefit from the mindset. Even small shifts toward identity-aware controls and local enforcement reduce exposure.

Mistakes That Turn Integrations Into Vulnerabilities

Most teams move fast, and it's usually for a good reason (think growth, market opportunity, or simply to serve customers better). Still, a few patterns keep popping up when things go wrong.

Over-permissive credentials

It is tempting to hand out broad access just to make something work, then never tighten it later. However, long-lived admin tokens and unrestricted service accounts that have long overstayed their welcome turn a small compromise into a significant incident.

The problem worsens when credentials are shared across multiple systems or hardcoded into applications, making it nearly impossible to rotate them. What starts as a quick fix to unblock a deadline becomes a permanent backdoor that attackers would love to exploit.

Flat access across environments

If dev and staging can reach production, a breach in one environment can spread to the others. Environmental boundaries are not red tape. They are blast-radius control.

When developers can accidentally push test code that hits live databases, or when a compromised staging server can pivot into production systems, you've essentially removed one of your most important safety nets. Proper network segmentation means that even if an attacker compromises your development environment, they hit a wall when trying to reach anything that matters.

No visibility into cross-system traffic

If services exchange data without logs, alerts, or audit trails, you are flying blind. You won't know something is wrong until users do. This blind spot becomes especially dangerous when integrations involve sensitive data or critical business processes.

Without proper monitoring, you can't tell whether an integration is being abused, leaking structured or unstructured data, or simply broken and failing silently. Good logging captures not just what happened, but who initiated it, what data was accessed, and whether the request was legitimate.

Shadow integrations

A well-meaning script goes straight to a production database. A new SaaS tool gets connected outside the review process. These shortcuts are complex to monitor and easy to forget. They often start as temporary solutions that become permanent fixtures, but without undergoing proper security reviews or documentation.

The person who built the integration might leave the company, taking all knowledge of how it works with them. These hidden connections create security gaps that don't appear in your official architecture diagrams, making them prime targets for attackers who've learned to look for the informal pathways that security teams don't know exist.

Principles for Secure Interoperability

Security that works at scale is boring on purpose. It relies on small, consistent rules that stack up into a mighty fortress of defenses. Here are the key principles you need to keep in mind (and action) if you want to make data work across systems safely.

Use short-lived credentials with tight scopes: prefer tokens that expire quickly and grant only the minimum level of access required. Reading a single dataset is not the same as owning the account. Treat scopes accordingly.

Authenticate and authorize every request: Internal does not mean trusted. Verify who or what is calling and whether the call is permitted. OAuth, signed JWTs, and mTLS are all useful here. Pick one that fits your stack and be consistent.

Segment by environment and by service: Separate dev, staging, and prod. Isolate services so that only the pairs that must talk can talk. The goal is containment. When something goes wrong, you want it to stop quickly.

Add observability to data flow: Centralize logs, track access, and alert on unusual behavior. You can’t fix what you can’t see, and you can’t learn from incidents without real traces.

Put gateways in front of sensitive systems: Expose services via an API gateway or a reverse proxy. Enforce auth, rate limits, schema validation, and logging in a single layer. It makes the safe path the easy path.

Final Word

Moving data across systems is how modern web applications and business solutions work today. If you want to make these systems secure, you need to do so without slowing them down (as much as you can). The point is to make the safe way also the easy way, so that the architecture does most of the security work for you.

You do not need a brand new platform to get there. You need fewer permanent keys, more identity in the request path, boundaries that mean something, and enough visibility to spot trouble early. Borrow ideas that fit your stack and ship them in small steps.

Most incidents are not master plans by brilliant attackers. They are ordinary accidents in complicated systems. Design yours so that accidents stay small. Then keep shipping.

by Web Desk via Digital Information World

Instagram’s New Competitor Insights Tool Raises Value and Privacy Questions

Instagram has expanded its professional dashboard with a new feature called Competitive Insights, designed to help business and creator accounts compare their activity with similar profiles. The tool allows users to select up to ten other accounts and view side-by-side information such as posting frequency and follower growth across Reels, feed posts, and ads.

The comparison appears straightforward, giving only numerical indicators of how often an account posts and how its follower count changes. While the feature also reveals engagement data from individual posts, even when like counts are hidden, it stops short of providing deeper breakdowns such as click-through rates, overall engagement graphs, or group comparisons. Users can only view one-to-one data contrasts, which limits broader performance analysis.

For marketers, the new view could offer a quick sense of how active competitors are and whether their own posting rhythm aligns with audience growth patterns. Yet some social media professionals have pointed out that follower counts alone rarely reflect campaign effectiveness. In an environment where discovery is increasingly shaped by algorithms rather than user follows, reach, shares, and clicks tend to matter more than raw growth numbers.

The update may serve as a helpful reference point for agencies or brands tracking their industry peers, but its practical value for creators appears mixed. Some have reacted cautiously, suggesting that such comparisons might encourage unhealthy competition or unnecessary pressure among individuals who already face constant performance tracking. "This is totally against the mental health care", noted one user under Sarah.roizman's post. Rickwulfk95 claims, "Another useless feature, instead changing back the algorithm to allow higher reach to the target, they try to start a war between poors, let's abandon IG for Reddit and Discord, where creators stick together instead of competing". Others have questioned the ethics of making aspects of post-level performance visible to outsiders, arguing that it introduces unwanted transparency into data that many consider private.

Instagram has not detailed whether further metrics will be added, though the current design suggests it could evolve toward broader benchmarking tools. For now, the feature remains basic and informative rather than strategic. It shows how much one posts and how followers respond, but it leaves out the indicators that tie content directly to audience action.

As social media measurement shifts toward engagement and conversion rather than follower tallies, the new dashboard option may feel like a step backward in understanding what drives meaningful results. For creators, it might prompt more anxiety than insight; for businesses, it may offer context but not clarity.

Notes: This post was edited/created using GenAI tools.

Read next:

• How Entrepreneurs and Creators are Shaping Their Own Brands Without Design Degrees

• Router Neglect: How a Simple Setting Puts Millions at Risk
by Irfan Ahmad via Digital Information World

How Entrepreneurs and Creators are Shaping Their Own Brands Without Design Degrees

With design tools more accessible than ever, a new breed of entrepreneurs are skipping the formalities of learning graphic design and teaching themselves. From content creators to solopreneurs, self-taught designers are building their businesses and brands from scratch, one social post at a time.

A new report from Adobe Express illustrates how this DIY design revolution is changing the face of small businesses. Based on a survey of 454 U.S. entrepreneurs and creators, the report tells how those with no formal design training are handling their branding tasks, often learning on the fly, with the assistance of AI tools and putting in hours of extra work to have professional looking content.

Getting the Job Done Without the Financial Blanket

Nearly 6 out of 10 (58%) said they do all their own designing. This includes their logos, graphics for their sites, and social posts. They handle all aspects of their branding work by themselves, often with little budget and no design knowledge.

Their two main means of learning? Trial and error (64%) and YouTube (54%). Only 16% indicated they had any formal design education. Don't slight TikTok, though. 1 in 4 turn to this platform for design tips.

Completing all of their own work, this group is highly dedicated, both time-wise and financially. Entrepreneurs are spending an average of $249 per year on design tools, while over 1 in 5 spend more than $1,000. Also, nearly 10 hours per week are spent on designing, with half stating that this time involved is the biggest obstacle of the whole process.

The amount of success that comes from these creators is especially impressive due to the amount of time, money, and lack of formal training that's involved. They often work outside regular working hours to get their branding efforts where they want them to be. Most often, they’re juggling day jobs, client work and personal obligations. For many, it’s not about the money, it’s about holding on to the creative vision and for them breaking those boundaries to establish something that feels authentic. This means each design choice, from font changes to color palette selections, is done from a personal perspective, based on what they think is good.

Design Burnout, Meet Imposter Syndrome!

The greatest barriers to success aren’t always technical; some are psychological. Many non-designers say they have problems with confidence in their images. Almost 1 in 5 (19%) confirm that their confidence is shaky regarding the quality of their brand assets. Over 50% have put off launches and posts because they did not appear polished enough to them.

What would help? A greater grasp of design (41%), more time (37%), and better quality templates (33%). The lack of confidence is compounded even with the growing popularity of AI tools. 71% say they use AI tools to help them create. 48% of them confirm they are feeling confident with their designs compared to the 42% who are not using AI.

Self-doubt is more than just a botheration; it has serious business consequences. 20% of respondents confirm they have received negative comments pertaining to their visuals. About a quarter admit they have copied or made substantial use of other brands styles.

It's very easy to think you should just do what others are doing when they seem successful, especially with so many design ideas going around. Doing this will generally cause what we call brand dilution. It's much better for entrepreneurs and creators to take their own personalized road, separating themselves from everyone else to show authenticity. Even when things aren't produced perfectly, it's better than copying everything you see. It almost always brings in more attention than a basic cliche template.

Time Dedication and Pressure to Keep Up

Creating visual content is not a one-time task, it’s a big weekly responsibility. Respondents report they spend the most time designing for TikTok (an average of 9 hours a week). Instagram is next with 8 hours, followed by business web pages and Etsy shops with 6 hours.

Spending this amount of time on content sometimes leads to endless revision. More than half of the respondents revise their graphics once or twice before posting. 15% report making three to five revisions, while 11% revise widely enough to lose track of the number.

For many, the solution is recycling. Nearly 3 in 4 respondents report using earlier content for new visual products. Others dream of outsourcing, with web design, video editing, and brand identity at the top of the list if budget weren’t a factor.

Even so, creators are working out their own shortcuts and routines for handling the pressure. Some batch content, some use templates and drag-and-drop tools to maintain consistency. The emergence of apps that help to resize, animate, and publish across channels has alleviated some of the pain. The pressure to remain visible and relevant is nevertheless high on fast-moving platforms such as TikTok.

Making DIY Design Work

Although they often face challenges, many creators feel good about their finished products. 79% say their branding work has enhanced the way in which their business is perceived for credibility. 16% even report that people have confused their brands with businesses that are larger or more established.

Instagram remains the best-performing platform for visuals, followed by Facebook, business websites and TikTok. 40% confess to being afflicted with a feeling of perfectionism, 35% say they suffered from stress, and 30% have confessed that they are subject to indecision. Imposter syndrome still hangs on for 19%. It is not all just stress and burnout, though. Almost half (49%) of creators have reported that designing is satisfying and a creative outlet for them.

Nevertheless, the ability to create visuals that register, without relying on others to help to get them, is a proud point. Many of these creators refer to this as a milestone in their growth. Each iteration and revision, or simply a design incident, inspires the creator to greater fluency in visual storytelling. Even though we live in a digital world that’s saturated with content, a brand’s biggest asset can be authenticity.

The New Approach to Branding

DIY branding offers something different and bigger. It constitutes a movement. While non-designers find it difficult still to become masters of instruments in the process of design, many have proved that creativity in combination with persistence and resourcefulness, does go a long way.

For the small businessman, the side hustler, or the creator who is starting from scratch, branding has become another skill in the toolbox. It is less a matter of polish and more about progress. As AI tools, platforms and learning resources continue to expand, so will the confidence and capabilities of these creators.

Read next:

• Gen Z Won't Wait Around for Bad Onboarding

• AI’s Limits Exposed: New Study Finds Machines Struggle With Real Remote Work

• How Social Media Loyalty Loops Make Us Copy the Worst Behavior from Our Own Group
by Irfan Ahmad via Digital Information World

Router Neglect: How a Simple Setting Puts Millions at Risk

A new study has revealed that millions of home internet users could be leaving their routers open to cyberattacks simply because they never change the default settings.

The research, carried out by Broadband Genie with technical input from McAfee, found that a large share of households continue to overlook one of the most basic forms of digital protection.

The survey gathered responses from more than 3,200 participants, analysing their habits and awareness of home network security. Although the findings show small improvements since 2024, they paint a picture of widespread complacency around router configuration. Nearly half of all users, around 47 percent, have never adjusted any of their factory settings. For a country with roughly 28 million broadband connections, that amounts to nearly 12.7 million routers still using default options that could be exploited by attackers.

Router settings are often left untouched for years after installation. The study showed that 81 percent of users have not changed their administrator passwords, and 85 percent continue to use the network names set by manufacturers. Almost seven in ten have never replaced their Wi-Fi passwords, and 84 percent have never updated their router’s firmware. While each of these figures has improved slightly from previous years, the overall trend suggests a slow pace of change.

Many users simply do not understand why router adjustments matter. Nearly three quarters of respondents said they saw no reason to change the default settings, and about a fifth admitted they did not know how to make those changes. This awareness gap is most visible among older participants. More than six in ten of those aged 65 or above said they had never opened their router’s settings, compared with less than a third among 18 to 24-year-olds. Even so, younger groups were far from fully secure.

The lack of attention to router configuration leaves an easy pathway for criminals. Devices that run on factory credentials can be accessed remotely, giving intruders the chance to read network traffic, intercept data or spread malicious software across connected devices. A single compromised router can expose personal files, home surveillance feeds, and even financial transactions. Despite growing awareness of cybersecurity risks, home routers remain one of the least protected points in the network chain.

Experts note that a few small adjustments are often enough to block these threats. Changing the administrator password and Wi-Fi key are the most effective first steps. Renaming the network adds another layer of difficulty for anyone attempting to identify the router model and exploit its known weaknesses. Regular firmware updates, meanwhile, close vulnerabilities that hackers may try to use. Many newer routers now install updates automatically, but older ones still depend on manual checks from users.

The figures collected by Broadband Genie also reflect a slow shift in attitudes. The share of users who changed their router settings rose by five percentage points compared with 2024, but that still leaves millions exposed. Firmware updates, which play a key role in preventing attacks, remain neglected despite years of public awareness campaigns. Most users continue to focus on software or mobile security, rarely considering that the router is their main entry point to the internet.

While users bear responsibility for their own security, the report points out that providers and manufacturers could make the process easier. Many routers are shipped with complex interfaces or generic instructions that discourage engagement. Simplifying these menus and sending clearer reminders could help users apply stronger settings without needing technical knowledge. Security measures could also be prompted during setup, much like operating systems require password creation on first use.

The study underlines how cybersecurity is no longer limited to experts or large companies. Every household network forms part of the wider digital environment, and a weak link in one home can affect others through compromised devices or data leaks. Taking a few minutes to update a router’s settings is not only a matter of personal protection but a contribution to broader online safety.

The 2025 results suggest the consumers are moving in the right direction, but progress remains slow. With more connected devices entering homes each year, the router continues to act as a digital front door. Keeping it secure requires awareness, routine updates, and small habits that, over time, close off easy routes for cybercriminals. For most people, the simplest step, changing a default password, remains the one that could make the biggest difference.

Notes: This post was edited/created using GenAI tools. 

Read next: New Study Shows Which Countries Use VPN Most and Least
by Asim BN via Digital Information World

Gen Z Won't Wait Around for Bad Onboarding

First days at work should be thrilling. But for the majority of Gen Z new hires in recent years, that inaugural week is a warning sign rather.

The latest SoftwareFinder poll queried over 500 Gen Z workers who started new positions in the previous two years. What did they discover? Nearly 1 in 5 ghosted throughout training. Another 22% had considered peacing out prematurely, and 8% actually left within their initial three months.

That ain't turnover. That's exit-running on a grand scale.

Off to a Bizarre Start

First impressions matter. And Gen Z isn't making them. They were most commonly reported to feel confused, disoriented, and as if an afterthought when getting onboarded. More than half did not even feel invited to ask basic questions, like who to talk to, or how policies were being implemented.

One described it as such: "It was like I went to a party that nobody invited me to."

They turned to the internet instead of direct guidance. 77% said they had to Google not-a-covered onboarding problem. 35% said they visited ChatGPT instead of their real manager. Not good.

And it's not a matter of getting the fix to some technical glitch or HR process. It's a trend. If new hires are left to figure out the basics, they start to doubt whether the company is not ready for them, or worse, is not interested.

Not Asking for Much

The vast majority of Gen Z hires aren't asking to be babysat. They just need:

• Clear pay information (90% said that does make a difference)
• Transparency of remote/hybrid policy (74% would prefer upfront)
• Sane career path (66%)
• Actual mental health care (60%)
• More than lip service DEI practices (54%)

Not too much to ask. Nobody's asking for meditation pods or catered lunches. Just truth and direction.

And that word, direction, is called upon a great deal. The majority of the study subjects said they had no way of knowing what their first week was going to be like. No schedules were given to them. No clearly stated goals. No individual to check in with.

Still Getting Decks and Dead Links

Most onboarding remains the same old playbook: 50-slide powerpoints, overly complicated documentation, maybe some pre-recorded videos. Only 34% said they had a peer mentor specifically assigned to them. Those that did? 79% said it was of huge help.

One said they picked up more from a Slack message with a colleague than from all training documents. Another mentioned that their mentor was "the one person who ever actually told me how things really worked."

Mentorship isn't credit extra. It's just plain old assistance.

And yet, there are still a few companies that seem to play down the extent of influence peer bonding has. People don't require shiny decks, they require someone to ask the stupid questions to.

Same Problems, Different Address

In-person or virtual didn't seem to make a difference. Onboarding was cringeworthy all round. Remote staff were a bit more likely to consider leaving early, but in-person staff complained too

One reported being taken to a desk, given a login, and dismissed for two days. Another was asked to "be creative" on a project before they found out what the company does.

Another reported, "I didn't even get to meet my team until day four. I thought maybe I was in the wrong Slack channel."

These are not aberrations. They're trends. The outcome? People wondering why they took the deal in the first place.

The Attention Span Myth

There are a few who like to spin the tale of how Gen Z is lacking in attention span. That's a cop-out. They're just accustomed to content being simple and uncluttered.

63% said that onboarding videos longer than 15 minutes were too long while they were being surveyed. 75% admitted scrolling through repetition or irrelevant parts.

Not because they are unmotivated, because they respect their time. And they understand when something is wasting unnecessarily.

Give them a five-minute explain vid. Text documents with plain text. Loom tutorials. Give them anything that has some respect for their time and simplicity.

When the First Week Is a Red Flag

There were some amazing tales. One manager never appeared all week, one had to pen a poem on corporate values, and one left two weeks later.

That's absurd. But it did happen. And they aren't typical.

Gen Z is listening. If the firm says they are "flexible" but rewards someone to eat their lunch out of the office, something is wrong. If the job description promises teamwork but the orientation is mysterious and lone-ranger, trust is easily lost.

And once lost, hard to restore.

What Works (to Them)

The solutions aren't hard:

• Keep meetings short and sweet
• Provide a clear map or plan
• Have a friend, and not a supervisor, call to follow up
• Give real answers about pay, policy, and opportunity
• Call to follow up, ask them how they are, and be interested in the response
• Make it easy to return to again so they won't have to wonder twice

That's not new. Just human.

Onboarding Isn't Just Admin Tasks

It's the company's opportunity to show that what they said they would have in the interview actually exists. And Gen Z is watching. Up close.

They don't care about leaving early. And certainly not about telling them why. A source described how they left a Glassdoor review of their same-day resignation. Another complained on Reddit that they were "left completely alone" for a week.

Others taught friends not to apply. And those friends complied. A recruiter admitted that when their rejected candidate called with a question, the candidate responded, "I heard your onboarding is a disaster."

That's not the sort of PR that can be fixed with a little branding Band-Aid.

It's Fixable

Gen Z isn't waiting for improved systems. If onboarding is fragmented or tone-deaf, they'll jump ship.

But that ain't no big secret. Optimizing onboarding isn't about hip-ifying it. It's about smart-ifying it. Give recently hired individuals some structure. Give them the keys to contact actual flesh-and-blood human beings. Be honest about what the work is, what the culture is, and what they're getting themselves into.".

If the first week is neat and respectful, they know they belong. If it's sloppy or cringeworthy, they don't. And if that's made clear, it's difficult to undo.

Onboarding is the company's first actual conversation with a new hire. It can build trust, or shatter it.

Gen Z has made one thing clear which one they will put up with.

Read next:

• New Study Shows Which Countries Use VPN Most and Least

• Android’s AI Shields Outperform iPhone as New Studies Highlight Scam Protection Gap
by Irfan Ahmad via Digital Information World

AI’s Limits Exposed: New Study Finds Machines Struggle With Real Remote Work

For years, discussion around artificial intelligence has centered on whether machines could eventually replace human jobs. That question has become sharper with the growth of remote work, where tasks are handled entirely online and often require a mix of technical and creative ability. Yet a new study from the Center for AI Safety and Scale AI provides a clearer picture of what AI can actually do in those settings. The findings show that, despite steady progress in reasoning and automation tools, today’s AI systems can complete only a small fraction of real freelance projects at human quality levels.

The study, called the Remote Labor Index (RLI), represents one of the most detailed attempts so far to measure AI’s performance on practical digital work. It focuses on tasks that mirror real online freelancing jobs rather than theoretical tests or benchmark problems. Researchers collected 240 completed projects from professional freelancers working through platforms such as Upwork. Each project included the original brief, all input materials, and the final deliverable that a client had accepted. These projects came from 23 categories of work, including product design, animation, architecture, game development, and data analysis. Together they covered more than 6,000 hours of paid labor valued at about $140,000.

Six advanced AI agents were then tested on the same projects. The systems included Manus, Grok 4, Sonnet 4.5, GPT-5, ChatGPT agent, and Gemini 2.5 Pro. Human evaluators compared the AI results to the professional standards of the original deliverables. The measure used was called the automation rate, defined as the percentage of projects that an AI completed to a standard that would be acceptable to a reasonable client.

The overall results placed current AI performance close to the bottom of the scale. Manus achieved the best outcome, with a 2.5 percent automation rate. Grok 4 and Sonnet 4.5 followed at 2.1 percent, while GPT-5 and ChatGPT agent reached 1.7 and 1.3 percent. Gemini 2.5 Pro finished last at 0.8 percent. In effect, even the strongest model could only complete two or three projects successfully out of every hundred. These numbers confirm that most paid remote work remains well beyond the reach of today’s AI systems.

To understand why, the study reviewed where and how the models failed. Nearly half of the AI outputs were judged to be of poor quality. About 36 percent were incomplete, and 18 percent contained technical errors such as corrupted or unusable files. Many tasks broke down before completion, with missing visuals, truncated videos, or unfinished code. Others showed inconsistency between design elements, such as an object changing shape between different 3D views. These errors highlight that even powerful models lack the internal verification ability that human workers apply when checking and refining their own results.

The researchers also noted that remote projects typically combine several layers of skill. A single job might involve writing, coding, design choices, and client-level presentation. While current AI models can produce functional text, basic graphics, or snippets of code, they often fail to align all these elements into a coherent, professional output. The lack of integrated quality control leads to results that are close to correct in parts but unsatisfactory as complete deliverables.

Some narrow areas showed stronger AI performance. Tasks involving short audio clips, simple image generation, or data visualization were occasionally completed at human level. In those cases, the systems benefited from established generative tools that already handle single-format media. The study used an additional metric, known as an Elo score, to track relative progress between different models. Although none matched the human baseline, newer models did show measurable improvement compared with earlier versions, suggesting steady if limited advancement.

Economically, the gap between potential and reality remains wide. When translated into market value, the highest-earning model, Manus, produced accepted work worth only $1,720 out of a total pool of nearly $144,000. This indicates that the contribution of current AI tools to freelance productivity is still marginal. The data also show that AI has not yet achieved meaningful cost deflation in remote labor markets, since most tasks still require full human oversight or redo.

For professionals who depend on online freelance income, the study’s conclusions provide some reassurance. Remote workers, especially in design, architecture, and multimedia fields, remain largely irreplaceable at present. The same applies to roles that involve judgment, error correction, and visual or interactive quality checks. However, the results also point to a gradual path of improvement. As AI models gain better multimodal reasoning and tool-use capacity, they may begin to handle larger portions of complex tasks under supervision.

The authors acknowledge that the benchmark does not cover all types of remote jobs. Work involving direct client communication, teamwork, or long-term project management was excluded. Even so, the Remote Labor Index represents the broadest test so far of AI’s real automation capacity in economically meaningful work. Its value lies in offering empirical measurement rather than assumption. By grounding AI evaluation in actual freelance projects, it shifts the conversation from hypothetical capabilities to demonstrated performance.

The findings suggest that the path to full automation of digital labor remains long. While AI can now assist with smaller creative or technical steps, it still struggles with the coordination, judgment, and quality assurance that professional work requires. Future updates to the RLI may help track whether ongoing model improvements translate into practical economic performance. For now, the study confirms that artificial intelligence, though advancing quickly, has yet to match the reliability and completeness of human remote workers.

Image: Yasmina H / Unsplash

Notes: This post was edited/created using GenAI tools.

Read next: Carnegie Mellon Study Finds Advanced AI Becomes More Self-Interested, Undermining Teamwork as It Gets Smarter
by Irfan Ahmad via Digital Information World