75% of workers ask AI instead of colleagues, and that convenience has a social cost
When a new analyst needs a quick answer, the path of least resistance is changing. MyIQ reported in July that 75% of people who use AI now ask chatbots instead of colleagues (MyIQ, July, methodology not published). That simple stat points to a bigger shift: AI assistants are becoming the first stop for routine work questions. That saves time, but it also changes how people learn, who gets credit for expertise, and how teams build trust.
What the numbers show, clear benefits, murky long-term trade-offs
Multiple surveys back up the headline: employees are turning to AI for answers and reporting big productivity and confidence gains. MyIQ (July, methodology not published) found 71% of respondents felt more self-sufficient and 62% felt more comfortable making decisions on their own than a year earlier. Workday (May, methodology not published) reported that as many as 86% of surveyed workers felt more productive because of AI tools. Adobe for Business (July, methodology not published) found 68% of workers would be more comfortable asking a chatbot an obvious question than a coworker, and reported a 35% dip in imposter syndrome and mental load for some remote workers attributed to agentic AI.
Terminology check: “agentic AI” refers to AI assistants or agents that take actions or answer questions on behalf of users, think internal chatbots, assistant agents that draft responses, or tools that perform basic triage.
These are self-reported outcomes, not randomized, long-term causal studies. Survey respondents tend to be early adopters or people who use new tools regularly, so treat the numbers as directional evidence rather than definitive proof.
“A question could solve an immediate problem, but it could also reveal who had expertise, how another person approached uncertainty, and which colleagues were willing to help. AI can preserve the practical answer while removing much of that social information.”, MyIQ report
Why leaders should care
Informal question-and-answer interactions do more than move a task forward. They transmit tacit knowledge, signal who has domain credibility, and create low-friction relationships that turn into mentorship and sponsorship. When those interactions move to an AI assistant, the practical answer survives, but many of those social signals disappear.
The risk sits on top of existing trends. The Centers for Disease Control and Prevention estimates about one in three adults feels lonely, and Gallup estimates one in five people worldwide feel lonely at work. A recent Science study linked post‑COVID remote work patterns to increased loneliness. Gallup also estimates that low employee engagement cost the global economy roughly $10 trillion in lost productivity last year (Gallup, methodology not published here).
Experts warn that short-term productivity gains can mask a slow erosion of talent development and connection. Adam Mendler, a leadership expert and UCLA instructor, said leaders must intentionally create opportunities for human connection to avoid weakening recruitment, retention, and development efforts.
Voices from the workplace
Data and expert commentary give one perspective; lived experience gives another. Maria Goyal, a 25-year-old marketing agency employee in Colorado, summed up the anxiety:
“It prevents me from like learning more in depth rather than just having like a robot do it for me.”, Maria Goyal
At the same time, people who use AI daily say it helps, but it doesn’t replace deliberate human touch. Allison Clair, founder and president of AJC Communications & Advisory, described using AI regularly while still scheduling time to check in with her team.
“I’ve definitely noticed that I do need to set up those times to chat with the people who work for me and get that face time in, and ask about their weeks and see how their weekends have been.”, Allison Clair
Practical steps that actually work
High-level advice like monitoring social impact, creating guidelines, and designing AI to prompt human contact is useful. Below are specific, operational actions leaders can pilot this quarter.
1. Instrument question flows (fast diagnostics)
- Log and tag question routing: record whether answers come from AI or a human, and tag the question type (how‑to/operational, judgment call, client-facing, career development).
- Set a 90‑day baseline: measure peer‑contact rate for onboarding and recurring “how‑to” queries. If peer‑routed questions drop more than 30% from baseline, trigger a deeper review.
- Track complementary outcomes: new‑hire time‑to‑first independent task, frequency of mentor touchpoints per 30 days, re‑hand‑off rate (how often AI answers escalate to a human), and changes in engagement and voluntary turnover.
- Privacy and legal caveat: anonymize logs, get HR and legal sign‑off, and be transparent with staff about what is being measured and why.
2. Create simple, enforceable rules
- Define escalation triggers: mandate human escalation for questions involving client relationships, judgment calls, promotion or compensation topics, and sensitive HR matters.
- Protect onboarding mentorship: require scheduled mentor check‑ins (for example, weekly for the first 90 days) regardless of how many questions the new hire routes to AI.
- Require attribution and “who to ask” prompts: when an AI answer affects a relationship, career outcome, or client deliverable, the assistant should include a recommended colleague and why that person is relevant.
3. Design AI to pull people back in
Product and platform teams, internal or vendor, can add simple patterns that nudge human contact without removing AI’s convenience.
- Use confidence scores and simple rules: if model confidence is low, the assistant suggests a short call with a named expert instead of a canned reply. Not all vendors expose confidence, so fall back to rules: repeated similar queries from a new hire or triggers like “promotion” or “contract” should prompt escalation.
- Hybrid replies: present a concise operational answer plus a “Talk to an expert” CTA that links to a booking page or shows availability for a 15‑minute sync.
- Automated learning handoffs: when a junior employee repeatedly asks the same question, flag it for a subject-matter expert and schedule a brief knowledge-transfer session to convert tribal knowledge into documented practice.
4. Monitor equity and bias risks
Who benefits from AI routing, and who loses visibility? Ask whether AI answers are replacing interactions that helped underrepresented employees build networks and sponsorship. If so, add guardrails that preserve human touch where it matters for career mobility.
Balancing trade-offs, not banning, but redesigning work
AI assistants reduce friction: fewer interruptions, faster routine answers, and reported drops in imposter feelings for some remote workers. Adobe for Business noted such reductions in mental load for certain remote staff. Yet replacing casual interactions with AI can hollow out informal learning and relationship building over time. Constance Noonan Hadley of the Institute for Life at Work put it plainly:
“I think it’s an underestimation of how much those moments matter, and it’s hard to convince people they matter until they don’t have them anymore.”, Constance Noonan Hadley
The sensible path is design: treat AI as a tool to make work cleaner, not as a replacement for the social architecture that builds capability. Choose policies and product behaviors that keep mentees, sponsors, and expertise pathways visible.
Questions leaders should ask, and short, honest answers
-
Are employees actually asking AI instead of colleagues?
Yes. MyIQ reported 75% of AI users ask chatbots instead of coworkers (MyIQ, July, methodology not published). Adobe and Workday surveys show similar patterns of comfort and uptake, but all are self‑reported snapshots rather than longitudinal, representative studies.
-
Does AI improve productivity?
Workers report improvements: Workday found up to 86% felt more productive, and MyIQ users reported increased self-sufficiency. These are self-reported effects; leaders should pair perceptions with objective metrics (throughput, error rates, customer satisfaction) before declaring lasting productivity gains.
-
Is there a social downside?
Yes, experts warn that routing questions to AI removes social information and casual exchanges that build expertise, trust, and mentorship. Broader loneliness and remote-work trends amplify that risk. The causal chain is plausible but not definitively proven; monitor indicators rather than assume inevitability.
-
What metrics should I pilot to spot problems early?
Start with a 90‑day baseline: peer‑contact rate for onboarding tasks, new‑hire time‑to‑first independent task, mentor touchpoints per 30 days, re‑hand‑off rate from AI to humans, and engagement/turnover trends. Trigger a review if peer-contact or mentor touchpoints decline by more than ~30% from baseline.
-
How should AI be governed inside the company?
Create clear, targeted policies: define when AI is appropriate, require human escalation for judgment or career-impacting issues, mandate mentor check‑ins for new hires, and make AI recommend human contacts for contextual answers. Keep governance light, measurable, and privacy‑aware.
Leader’s short game
AI agents are now a first stop for workplace questions. That change brings speed and confidence, but it can also erode the informal pathways that develop talent and trust. The answer is not prohibition, it is intentional design. Instrument question flows, set simple governance rules, require mentor touchpoints during onboarding, and make your AI nudges point people to people when the stakes are social, relational, or career‑building.
Start with a 90‑day pilot: log question routing, set a clear threshold for review, and implement one product change that nudges employees toward a human expert (for example, a “Talk to an expert” CTA). Those low-friction moves let you harvest AI’s benefits without trading away the social capital that supports long-term performance and retention.