Your team is quietly losing nearly a full working day each week to “botsitting” AI systems. According to a 2026 Glean report, white-collar workers now spend an average of 6.4 hours every week supervising, error-checking, and debugging AI tools instead of doing work that actually moves the needle. The result is tedious, often invisible labor that drains energy and never shows up in performance metrics.

This hidden workload is fueling real job frustration, and the Glean study found it is also pushing your best people out the door. If your organization relies on AI but no one is tracking the cost of constant oversight, you are paying for it in wasted hours and morale. This post explains how you can spot unnecessary botsitting, and gives you practical steps to cut the drag from your operation, for real productivity gains, and better retention.

AI Was Supposed to Save Time, Now It’s Creating Hidden Labor Costs

The promise of AI was simple: automate routine tasks and free up employees for work that matters. Instead, Glean’s new report reveals that the reality is more complex and frustrating. Large-scale adoption brings hidden labor, as staff spend hours “feeding context, checking outputs, debugging mistakes, and cleaning up errors.” This invisible work sits outside official job roles and is rarely measured or acknowledged.

Much of this botsitting isn’t just tedious, it’s a drag on morale. Rebecca Hinds, head of the Work AI Institute at Glean, describes it as “not rewarded and it’s not appreciated or tracked or measured and certainly not incentivized within the organization.” Workers are becoming unofficial mediators between AI tools that can’t connect or share accurate data, turning technical progress into practical bottlenecks. Productivity gains promised on paper vanish in the churn of fixing AI behind the scenes.

What Botsitting Actually Involves: Real Worker Tasks and Hidden Human Labor

Feeding AI context and instructions

Supervising AI starts with employees having to supply essential context the system should already know. Workers manually input background details, tweak prompts, and repeat instructions so the AI tool understands the specifics of their process or problem. This is not high-value creative work. It’s a repetitive, administrative task that pulls talent away from actual production and improvement. Glean’s report calls it “feeding it context”, which quickly becomes tedious, especially as employees fine-tune raw data and clarify ambiguous requirements again and again.

Cleaning up outputs and fixing errors

AI tools rarely deliver perfect results. Every week, employees sift through generated outputs, hunting for mistakes and correcting them line by line. This includes fixing misclassifications, patching formatting errors, and adjusting data that has been wrongly interpreted. It adds up to hours spent on quality assurance instead of advancing new initiatives. Rebecca Hinds of Glean describes this work as “not rewarded and it’s not appreciated or tracked or measured.” For operations leaders, failing to recognize this clean-up process means ignoring a substantial hidden cost within their teams.

Debugging mistakes between disconnected systems

Another major drain is acting as the go-between for multiple AI tools that do not talk to each other. Employees are tasked with searching for inconsistencies, troubleshooting integration issues, and manually moving information between platforms. Many become informal IT specialists, chasing bugs and bridging gaps between disconnected systems. The result is slow workflows, repeated errors, and rising job frustration as digital tools that should work together instead require constant human intervention.

Why Botsitting Is Exhausting: Morale, Recognition, and Exit Risk

Botsitting is not recognized or rewarded

Botsitting AI quickly wears down employees because the work is invisible and ignored. According to Rebecca Hinds, head of the Work AI Institute at Glean, this labor is “not rewarded and it’s not appreciated or tracked or measured and certainly not incentivized within the organization.” Workers get stuck doing tasks that do not count toward goals or performance reviews. No one is thanked for catching errors or debugging bot mistakes. This lack of recognition amplifies fatigue and frustration.

High botsitting hours correlate with increased job searching

The hidden labor eats away at retention. Glean’s survey of 6,000 full-time employees found a clear link: those who spend more time botsitting are 73% more likely to be actively looking for another job. Routine supervision and cleanup sap engagement. Employees who are ignored for bearing this extra workload grow exhausted. Instead of feeling valuable, they feel expendable. This sets up a talent drain and increases employee exit risk.

Employees forced to automate meaningful work

Botsitting cuts deeper when AI replaces tasks employees actually enjoy. The Glean report notes that some staff are now asked to automate the parts of their jobs that give them purpose, such as customer-service roles built on relationships. As Rebecca Hinds put it, “That is very dangerous.” Sidelining meaningful work for tedious AI supervision leads to resentment. The result is declining morale and a slow erosion of culture.

The Productivity Paradox: Why Company-Wide Performance Isn’t Improving

87% of workers report using AI

AI adoption is widespread. According to Glean’s Work AI Institute report, 87 percent of surveyed employees say they use AI tools in their daily work. This level of penetration suggests organizations have invested heavily in digital transformation. Employees are engaging with platforms, automating tasks, and interfacing with bots. In theory, more AI use should drive faster production and higher quality outcomes across the board.

Only 13% see substantial organizational performance improvement

The results tell a different story. Just 13 percent of respondents said their organization was performing “significantly better” because of AI. It’s a glaring gap. Productivity enhancements at the individual level do not translate into measurable gains for the business. As the Glean report notes, “much of the missing productivity is being consumed by work employees never expected to do.” The labor doesn’t move KPIs, and management is left wondering why efficiency targets remain unmet despite near-universal AI rollout.

Disconnected AI tools increase labor burden

One driving factor is tool fragmentation. Employees spend valuable hours moving information between separate systems, cleaning up errors, and providing extra context. Rebecca Hinds, the head of Glean’s Work AI Institute, highlights this in the report: workers are “effectively becoming the go-between for technologies that don’t work well together.” Poor integration compounds AI hidden labor, making botsitting tedious and duplicative instead of productive. Unless companies tackle this structural problem by consolidating tech stacks or improving data flow, AI will keep draining rather than delivering results.

Practical Steps to Break the Botsitting Cycle in Manufacturing Operations

Establish clear ownership and reward systems for botsitting tasks

Most manufacturing teams have no formal ownership for botsitting labor. Assign accountability for supervising AI outputs, error-checking, and context feeding. Make this a tracked task with explicit recognition. Build it into performance reviews rather than letting it slip into hidden AI labor that breeds resentment. Consider spot rewards or shift differentials for time spent on repetitive AI supervision. When tracking is visible, fatigue and frustration drop, and so does employee exit risk.

Integrate AI tools to reduce manual context-switching

The Glean report highlights the drain from moving information between disconnected AI systems. Consolidate platforms wherever possible. Choose interoperable tools from vendors such as Microsoft, Siemens, or GE Digital that minimize “go-between” work. Map out workflows to spot the context handoffs and manual data entry slowing teams down, then build integrations that eliminate handoffs altogether. When every AI solution speaks the same language, you cut hours of wasted supervision and manual cleanup.

Focus AI on eliminating mundane work, not replacing meaningful tasks

Use AI where it reduces routine checks, audits, and process documentation. Avoid automating tasks employees actually value, such as customer relationship management or creative troubleshooting. As Rebecca Hinds from Glean stresses, forcing employees to supervise bots instead of doing work that gives them “joy and meaning” risks damaging morale. Prioritize AI deployments that relieve monotony, not those that push staff away from the parts of their job that matter most. Scrutinize every implementation for its impact on engagement and real productivity.

Looking Ahead: Rethinking AI Implementation to Prevent Hidden Labor

Audit hidden AI labor before scaling deployments

Leaders need visibility before expansion. Audit where botsitting tasks show up in daily workflows, and pinpoint which AI tools generate the most manual supervision. The Glean report identifies common pain points: “feeding it context, checking outputs, debugging mistakes, and cleaning up errors.” Capture this data. Roll up the hours lost per function, not just by role. If you ignore hidden AI labor, you compound the employee exit risk. Tallying up these tasks gives you leverage for renegotiating vendor contracts or redesigning processes, before you pile more bots onto existing platforms.

Develop frameworks for measuring true productivity ROI

Generic metrics won’t catch invisible labor. Build frameworks that compare intended automation savings with actual time spent botsitting. Record every step required to supervise, correct, or supplement your AI’s output, especially those that zero out supposed gains. Get granular: count context feeding, error reconciliation, and manual integration between tools. Benchmark what performance looks like when botsitting is minimized. Only a practical ROI calculation reveals the productivity paradox that the Glean study uncovered: heavy AI adoption on paper versus weak operational improvement in reality.

Design AI solutions around operational realities

Stop assuming generic AI models will fit the needs of manufacturing and quality teams. Design deployments for the actual workflows and the people running them. Consider tasks employees enjoy and want to retain, as well as where error risk is highest. According to Rebecca Hinds on the “Cognitive Revolution” podcast, jobs are at risk when employees must supervise AI agents instead of focusing on what brings them meaning. Build solutions that respect the knowledge and preferences of your workforce, rooting out botsitting before scaling AI deployments further.

Source: businessinsider.com