The rise of “botsitting” and the new hidden tax on AI adoption
A growing share of the modern workweek is being quietly reallocated to a task few job descriptions mention: “botsitting”—the hands-on labor of supplying context to AI systems, debugging their outputs, and correcting mistakes. A study from Glean’s Work AI Institute, based on a survey of 6,000 full-time employees across the US, UK, and Australia, puts a hard number on that invisible workload: 6.4 hours per week on average.
That figure matters because it reframes the prevailing narrative around enterprise AI. Adoption is high—87% of workers report using AI tools—and perceived personal benefit is real, with 75% citing individual productivity gains. Yet the organizational payoff remains elusive: only 13% see measurable performance improvement at the company level. The gap between “I feel faster” and “the business is measurably better” is where botsitting lives.
In practical terms, botsitting is not a niche behavior confined to technical teams. It is increasingly a cross-functional reality for knowledge workers who must:
- Translate messy real-world goals into machine-readable prompts
- Re-check AI-generated text, analysis, or code for accuracy and compliance
- Reconcile outputs across tools (chat assistants, document systems, CRM platforms, analytics dashboards)
- Patch failures manually when AI lacks context, permissions, or reliable data
The result is a new kind of operational drag—one that is easy to miss in budgets, but hard to ignore in daily experience.
Why individual AI productivity gains aren’t translating into enterprise performance
The study’s most consequential signal may be the disconnect between widespread AI usage and limited organizational uplift. This resembles a familiar pattern in technology history: the productivity paradox, where transformative tools proliferate before measurement systems, workflows, and management practices adapt enough to capture the gains.
Botsitting helps explain why. AI tools can accelerate discrete tasks—drafting an email, summarizing a meeting, generating a first-pass analysis—while simultaneously introducing integration overhead that cancels out benefits at scale. Three dynamics stand out:
- Hidden integration costs: AI systems rarely arrive “plug-and-play” inside real enterprise environments. Workers become the glue between incomplete data, inconsistent taxonomies, and fragmented systems. The labor of preparing context and cleaning outputs becomes a recurring cost center, not a one-time setup.
- Human-machine feedback loops without infrastructure: AI improves when organizations systematically capture corrections, track error patterns, and share best practices. Without mechanisms like error-tracking dashboards, prompt libraries, or quality-control workflows, each employee repeats the same trial-and-error cycle—raising frustration and lowering trust.
- Platform fragmentation: The proliferation of specialized AI services—language models, workflow automation, search, analytics—creates a “many tools, many interfaces” environment. In the absence of coherent architecture, employees serve as de facto integration layers, manually transferring context from one system to another.
For executives, the implication is not that AI “doesn’t work,” but that AI benefits are being absorbed by coordination costs. The organization may be moving faster in pockets while still failing to improve end-to-end throughput, customer outcomes, or cycle times—metrics that ultimately define enterprise performance.
The talent and cost implications: when AI oversight becomes a retention risk
Perhaps the most strategically urgent finding is human, not technical: employees who do more botsitting are 73% more likely to explore other job opportunities. That is a stark indicator that AI, when poorly operationalized, can shift from a productivity enhancer to a morale hazard.
This risk compounds in roles where quality and accountability are non-negotiable—customer support, sales operations, finance, HR, legal, and regulated industries—because these functions often require:
- High verification effort (checking for hallucinations, policy violations, or incorrect numbers)
- Auditability (documenting why a decision was made and what sources were used)
- Consistency (ensuring outputs match brand voice, legal standards, or contractual terms)
Economically, botsitting creates at least three layers of cost that many organizations are not yet accounting for:
- Inflated labor costs: If 6.4 hours per week are spent on AI supervision and correction, the ROI of AI licenses can be overstated unless those hours are measured and reduced.
- Opportunity cost: Time spent managing AI is time not spent on high-value work—innovation, client relationships, strategic planning, and creative problem-solving.
- Retention and wage pressure: If botsitting is experienced as drudgery, organizations may face higher turnover, rehiring costs, and wage inflation—especially for employees who can move to firms with better AI operating models.
In other words, botsitting is not merely a workflow nuisance; it is a workforce experience issue with direct implications for competitiveness and continuity.
What enterprise leaders should do next: from AI tool rollout to AI operating model
The report’s underlying message is that deploying more AI is not the same as building an AI-capable organization. The next phase of enterprise AI will likely be defined less by model selection and more by operational design—how work is structured, governed, measured, and improved.
Several strategic moves emerge as high-leverage:
- Invest in AI enablement infrastructure, not just licenses
Prioritize shared assets that reduce repeated botsitting, such as prompt playbooks, approved context templates, evaluation rubrics, and issue-tracking systems that turn individual fixes into organizational learning.
- Build digital dexterity at scale
Training should extend beyond “how to use a chatbot” into data literacy, prompt engineering, workflow orchestration, and risk-aware verification—skills that reduce rework and raise confidence.
- Establish governance and standards for AI-augmented work
Define when AI outputs can be trusted, how exceptions are escalated, and how accountability is maintained. Clear protocols also support compliance as regulators increasingly focus on AI transparency, labor impacts, and decision accountability.
- Measure the hidden labor explicitly
Introducing internal metrics—effectively “botsitting KPIs”—can shift investment toward reducing friction. What gets measured gets managed, and what gets managed becomes scalable.
The organizations that capture durable value from AI will be those that treat botsitting as a design flaw to be engineered down—not a new normal to be endured. That shift, from novelty-driven adoption to disciplined operating model transformation, is where AI’s promise can finally move from individual acceleration to enterprise advantage.
By
By
By
By
By
By
